From 87abd327bfcf0a7c66bfe79d6ce40ae2fd24c969 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 20:52:33 +0800 Subject: [PATCH 01/19] docs: standardize tutorial pages - Card components, blob URLs, i18n fixes - Replaced tags with Mintlify CardGroup+Card components - Added Cloud+Download workflow cards with UTM params - Fixed resolve/main to blob/main URLs - Removed em dashes across tutorials - Synced zh/ja/ko translations to match EN structure --- ja/tutorials/3d/hunyuan3D-2.mdx | 39 +++- ja/tutorials/3d/triposplat.mdx | 10 +- ja/tutorials/audio/ace-step/ace-step-v1-5.mdx | 10 +- ja/tutorials/audio/ace-step/ace-step-v1.mdx | 36 ++- .../audio/stable-audio/stable-audio-1.mdx | 4 +- .../audio/stable-audio/stable-audio-3.mdx | 8 +- ja/tutorials/basic/inpaint.mdx | 4 +- ja/tutorials/basic/outpaint.mdx | 2 +- ja/tutorials/controlnet/controlnet.mdx | 4 +- ja/tutorials/controlnet/depth-controlnet.mdx | 2 +- ja/tutorials/controlnet/depth-t2i-adapter.mdx | 2 +- .../controlnet/mixing-controlnets.mdx | 6 +- .../controlnet/pose-controlnet-2-pass.mdx | 4 +- ja/tutorials/flux/flux-1-controlnet.mdx | 40 ++-- ja/tutorials/flux/flux-1-fill-dev.mdx | 32 ++- ja/tutorials/flux/flux-1-kontext-dev.mdx | 15 +- ja/tutorials/flux/flux-1-text-to-image.mdx | 70 ++++-- ja/tutorials/flux/flux-1-uso.mdx | 28 +-- ja/tutorials/flux/flux-2-dev.mdx | 6 +- ja/tutorials/flux/flux-2-klein.mdx | 12 +- ja/tutorials/flux/flux1-krea-dev.mdx | 25 +- ja/tutorials/image/anima/anima.mdx | 8 +- ja/tutorials/image/boogu/boogu-image-0.1.mdx | 14 +- .../image/cosmos/cosmos-predict2-t2i.mdx | 6 +- .../image/ernie-image/ernie-image.mdx | 16 +- ja/tutorials/image/hidream/hidream-e1.mdx | 24 +- ja/tutorials/image/hidream/hidream-i1.mdx | 45 ++-- ja/tutorials/image/hidream/hidream-o1.mdx | 22 +- ja/tutorials/image/ideogram/ideogram-v4.mdx | 10 +- ja/tutorials/image/krea/krea-2.mdx | 6 +- ja/tutorials/image/lens/lens.mdx | 8 +- .../newbie-image/newbie-image-exp-0-1.mdx | 23 +- ja/tutorials/image/omnigen/omnigen2.mdx | 24 +- ja/tutorials/image/ovis/ovis-image.mdx | 21 +- ja/tutorials/image/pixeldit/pixeldit.mdx | 4 +- ja/tutorials/image/qwen/qwen-image-2512.mdx | 22 +- .../image/qwen/qwen-image-edit-2511.mdx | 23 +- ja/tutorials/image/qwen/qwen-image-edit.mdx | 23 +- .../image/qwen/qwen-image-layered.mdx | 24 +- ja/tutorials/image/qwen/qwen-image.mdx | 81 +++---- ja/tutorials/image/z-image/z-image-turbo.mdx | 8 +- ja/tutorials/image/z-image/z-image.mdx | 6 +- ja/tutorials/llm/gemma4/gemma4.mdx | 4 +- ja/tutorials/llm/qwen/qwen3.mdx | 6 +- ja/tutorials/llm/qwen/qwen3_5.mdx | 6 +- ja/tutorials/partner-nodes/google/gemini.mdx | 12 +- .../kling/kling-motion-control.mdx | 5 +- .../moonvalley-video-generation.mdx | 24 +- ja/tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 +- .../partner-nodes/runway/video-generation.mdx | 12 +- .../partner-nodes/tripo/model-generation.mdx | 24 +- ja/tutorials/utility/depth-anything-3.mdx | 8 +- .../utility/face-detection/mediapipe.mdx | 2 +- ja/tutorials/utility/moge.mdx | 4 +- .../utility/pose-detection-sdpose.mdx | 6 +- .../utility/remove-background-birefnet.mdx | 2 +- ja/tutorials/utility/video-segment-sam3.mdx | 2 +- .../utility/void-video-inpainting.mdx | 12 +- ja/tutorials/video/bytedance/bernini-r.mdx | 8 +- .../cosmos/cosmos-predict2-video2world.mdx | 47 ++-- .../video/hunyuan/hunyuan-video-1-5.mdx | 10 +- ja/tutorials/video/hunyuan/hunyuan-video.mdx | 14 +- ja/tutorials/video/kandinsky/kandinsky-5.mdx | 64 ++++-- ja/tutorials/video/ltxv.mdx | 32 ++- ja/tutorials/video/wan/fun-camera.mdx | 98 +++++--- ja/tutorials/video/wan/fun-control.mdx | 10 +- ja/tutorials/video/wan/fun-inp.mdx | 12 +- ja/tutorials/video/wan/vace.mdx | 10 +- ja/tutorials/video/wan/wan-ati.mdx | 10 +- ja/tutorials/video/wan/wan-causal-forcing.mdx | 8 +- ja/tutorials/video/wan/wan-dancer.mdx | 12 +- ja/tutorials/video/wan/wan-flf.mdx | 10 +- ja/tutorials/video/wan/wan-move.mdx | 42 ++-- ja/tutorials/video/wan/wan-video.mdx | 14 +- ja/tutorials/video/wan/wan2-2-animate.mdx | 56 +++-- ja/tutorials/video/wan/wan2-2-fun-camera.mdx | 53 ++++- ja/tutorials/video/wan/wan2-2-fun-control.mdx | 72 +++--- ja/tutorials/video/wan/wan2-2-fun-inp.mdx | 49 ++-- ja/tutorials/video/wan/wan2-2-s2v.mdx | 72 ++++-- ja/tutorials/video/wan/wan2_2.mdx | 82 +++---- ja/tutorials/video/zai/scail2.mdx | 14 +- ko/tutorials/3d/hunyuan3D-2.mdx | 39 +++- ko/tutorials/3d/triposplat.mdx | 10 +- ko/tutorials/audio/ace-step/ace-step-v1-5.mdx | 10 +- ko/tutorials/audio/ace-step/ace-step-v1.mdx | 36 ++- .../audio/stable-audio/stable-audio-1.mdx | 4 +- .../audio/stable-audio/stable-audio-3.mdx | 8 +- ko/tutorials/basic/inpaint.mdx | 4 +- ko/tutorials/basic/outpaint.mdx | 2 +- ko/tutorials/controlnet/controlnet.mdx | 4 +- ko/tutorials/controlnet/depth-controlnet.mdx | 2 +- ko/tutorials/controlnet/depth-t2i-adapter.mdx | 2 +- .../controlnet/mixing-controlnets.mdx | 6 +- .../controlnet/pose-controlnet-2-pass.mdx | 4 +- ko/tutorials/flux/flux-1-controlnet.mdx | 40 ++-- ko/tutorials/flux/flux-1-fill-dev.mdx | 32 ++- ko/tutorials/flux/flux-1-kontext-dev.mdx | 21 +- ko/tutorials/flux/flux-1-text-to-image.mdx | 70 ++++-- ko/tutorials/flux/flux-1-uso.mdx | 28 +-- ko/tutorials/flux/flux-2-dev.mdx | 6 +- ko/tutorials/flux/flux-2-klein.mdx | 12 +- ko/tutorials/flux/flux1-krea-dev.mdx | 25 +- ko/tutorials/image/anima/anima.mdx | 8 +- ko/tutorials/image/boogu/boogu-image-0.1.mdx | 14 +- .../image/cosmos/cosmos-predict2-t2i.mdx | 6 +- .../image/ernie-image/ernie-image.mdx | 16 +- ko/tutorials/image/hidream/hidream-e1.mdx | 24 +- ko/tutorials/image/hidream/hidream-i1.mdx | 45 ++-- ko/tutorials/image/hidream/hidream-o1.mdx | 22 +- ko/tutorials/image/ideogram/ideogram-v4.mdx | 10 +- ko/tutorials/image/krea/krea-2.mdx | 6 +- ko/tutorials/image/lens/lens.mdx | 8 +- .../newbie-image/newbie-image-exp-0-1.mdx | 23 +- ko/tutorials/image/omnigen/omnigen2.mdx | 22 +- ko/tutorials/image/ovis/ovis-image.mdx | 21 +- ko/tutorials/image/pixeldit/pixeldit.mdx | 4 +- ko/tutorials/image/qwen/qwen-image-2512.mdx | 22 +- .../image/qwen/qwen-image-edit-2511.mdx | 23 +- ko/tutorials/image/qwen/qwen-image-edit.mdx | 23 +- .../image/qwen/qwen-image-layered.mdx | 22 +- ko/tutorials/image/qwen/qwen-image.mdx | 81 +++---- ko/tutorials/image/z-image/z-image-turbo.mdx | 8 +- ko/tutorials/image/z-image/z-image.mdx | 6 +- ko/tutorials/llm/gemma4/gemma4.mdx | 4 +- ko/tutorials/llm/qwen/qwen3.mdx | 6 +- ko/tutorials/llm/qwen/qwen3_5.mdx | 6 +- ko/tutorials/partner-nodes/google/gemini.mdx | 12 +- .../kling/kling-motion-control.mdx | 5 +- .../moonvalley-video-generation.mdx | 24 +- ko/tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 +- .../partner-nodes/runway/video-generation.mdx | 12 +- .../partner-nodes/tripo/model-generation.mdx | 24 +- ko/tutorials/utility/depth-anything-3.mdx | 8 +- .../utility/face-detection/mediapipe.mdx | 2 +- ko/tutorials/utility/moge.mdx | 4 +- .../utility/pose-detection-sdpose.mdx | 6 +- .../utility/remove-background-birefnet.mdx | 2 +- ko/tutorials/utility/video-segment-sam3.mdx | 2 +- .../utility/void-video-inpainting.mdx | 12 +- ko/tutorials/video/bytedance/bernini-r.mdx | 8 +- .../cosmos/cosmos-predict2-video2world.mdx | 47 ++-- .../video/hunyuan/hunyuan-video-1-5.mdx | 10 +- ko/tutorials/video/hunyuan/hunyuan-video.mdx | 14 +- ko/tutorials/video/kandinsky/kandinsky-5.mdx | 64 ++++-- ko/tutorials/video/ltxv.mdx | 32 ++- ko/tutorials/video/wan/fun-camera.mdx | 98 +++++--- ko/tutorials/video/wan/fun-control.mdx | 10 +- ko/tutorials/video/wan/fun-inp.mdx | 12 +- ko/tutorials/video/wan/vace.mdx | 10 +- ko/tutorials/video/wan/wan-ati.mdx | 10 +- ko/tutorials/video/wan/wan-causal-forcing.mdx | 8 +- ko/tutorials/video/wan/wan-dancer.mdx | 12 +- ko/tutorials/video/wan/wan-flf.mdx | 10 +- ko/tutorials/video/wan/wan-move.mdx | 42 ++-- ko/tutorials/video/wan/wan-video.mdx | 14 +- ko/tutorials/video/wan/wan2-2-animate.mdx | 56 +++-- ko/tutorials/video/wan/wan2-2-fun-camera.mdx | 53 ++++- ko/tutorials/video/wan/wan2-2-fun-control.mdx | 70 ++++-- ko/tutorials/video/wan/wan2-2-fun-inp.mdx | 49 ++-- ko/tutorials/video/wan/wan2-2-s2v.mdx | 68 ++++-- ko/tutorials/video/wan/wan2_2.mdx | 82 +++---- ko/tutorials/video/zai/scail2.mdx | 14 +- tutorials/3d/hunyuan3D-2.mdx | 39 +++- tutorials/flux/flux-1-controlnet.mdx | 22 +- tutorials/flux/flux-1-fill-dev.mdx | 32 ++- tutorials/flux/flux-1-kontext-dev.mdx | 21 +- tutorials/flux/flux-1-text-to-image.mdx | 70 ++++-- tutorials/flux/flux-1-uso.mdx | 28 +-- tutorials/flux/flux1-krea-dev.mdx | 25 +- tutorials/image/hidream/hidream-e1.mdx | 26 ++- tutorials/image/hidream/hidream-i1.mdx | 45 ++-- .../newbie-image/newbie-image-exp-0-1.mdx | 15 +- tutorials/image/omnigen/omnigen2.mdx | 16 +- tutorials/image/ovis/ovis-image.mdx | 15 +- tutorials/image/qwen/qwen-image-2512.mdx | 25 +- tutorials/image/qwen/qwen-image-edit-2511.mdx | 23 +- tutorials/image/qwen/qwen-image-edit.mdx | 23 +- tutorials/image/qwen/qwen-image-layered.mdx | 24 +- tutorials/image/qwen/qwen-image.mdx | 80 ++++--- tutorials/partner-nodes/google/gemini.mdx | 12 +- tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 +- tutorials/video/bytedance/bernini-r.mdx | 44 ++-- .../cosmos/cosmos-predict2-video2world.mdx | 52 +++-- tutorials/video/hunyuan/hunyuan-video-1-5.mdx | 170 ++++++++++++-- tutorials/video/hunyuan/hunyuan-video.mdx | 58 +++-- tutorials/video/kandinsky/kandinsky-5.mdx | 64 ++++-- tutorials/video/ltx/ltx-2.mdx | 39 ++-- tutorials/video/ltxv.mdx | 32 ++- tutorials/video/wan/fun-camera.mdx | 105 ++++++--- tutorials/video/wan/fun-control.mdx | 126 ++++++---- tutorials/video/wan/fun-inp.mdx | 66 ++++-- tutorials/video/wan/vace.mdx | 215 +++++++++++++----- tutorials/video/wan/wan-alpha.mdx | 96 +++++++- tutorials/video/wan/wan-ati.mdx | 92 +++++--- tutorials/video/wan/wan-causal-forcing.mdx | 51 +++-- tutorials/video/wan/wan-flf.mdx | 68 ++++-- tutorials/video/wan/wan-move.mdx | 42 ++-- tutorials/video/wan/wan-video.mdx | 108 ++++++--- tutorials/video/wan/wan2-2-animate.mdx | 69 +++--- tutorials/video/wan/wan2-2-fun-camera.mdx | 57 +++-- tutorials/video/wan/wan2-2-fun-control.mdx | 81 ++++--- tutorials/video/wan/wan2-2-fun-inp.mdx | 59 +++-- tutorials/video/wan/wan2-2-s2v.mdx | 68 ++++-- tutorials/video/wan/wan2_2.mdx | 148 ++++++++---- tutorials/video/zai/scail2.mdx | 8 +- zh/tutorials/3d/hunyuan3D-2.mdx | 39 +++- zh/tutorials/3d/triposplat.mdx | 10 +- zh/tutorials/audio/ace-step/ace-step-v1-5.mdx | 10 +- zh/tutorials/audio/ace-step/ace-step-v1.mdx | 36 ++- .../audio/stable-audio/stable-audio-1.mdx | 4 +- .../audio/stable-audio/stable-audio-3.mdx | 8 +- zh/tutorials/basic/inpaint.mdx | 4 +- zh/tutorials/basic/outpaint.mdx | 2 +- zh/tutorials/controlnet/controlnet.mdx | 4 +- zh/tutorials/controlnet/depth-controlnet.mdx | 2 +- zh/tutorials/controlnet/depth-t2i-adapter.mdx | 2 +- .../controlnet/mixing-controlnets.mdx | 6 +- .../controlnet/pose-controlnet-2-pass.mdx | 4 +- zh/tutorials/flux/flux-1-controlnet.mdx | 40 ++-- zh/tutorials/flux/flux-1-fill-dev.mdx | 34 ++- zh/tutorials/flux/flux-1-kontext-dev.mdx | 21 +- zh/tutorials/flux/flux-1-text-to-image.mdx | 70 ++++-- zh/tutorials/flux/flux-1-uso.mdx | 28 +-- zh/tutorials/flux/flux-2-dev.mdx | 6 +- zh/tutorials/flux/flux-2-klein.mdx | 12 +- zh/tutorials/flux/flux1-krea-dev.mdx | 25 +- zh/tutorials/image/anima/anima.mdx | 8 +- zh/tutorials/image/boogu/boogu-image-0.1.mdx | 14 +- .../image/cosmos/cosmos-predict2-t2i.mdx | 6 +- .../image/ernie-image/ernie-image.mdx | 16 +- zh/tutorials/image/hidream/hidream-e1.mdx | 24 +- zh/tutorials/image/hidream/hidream-i1.mdx | 45 ++-- zh/tutorials/image/hidream/hidream-o1.mdx | 22 +- zh/tutorials/image/ideogram/ideogram-v4.mdx | 10 +- zh/tutorials/image/krea/krea-2.mdx | 6 +- zh/tutorials/image/lens/lens.mdx | 8 +- .../newbie-image/newbie-image-exp-0-1.mdx | 23 +- zh/tutorials/image/omnigen/omnigen2.mdx | 22 +- zh/tutorials/image/ovis/ovis-image.mdx | 21 +- zh/tutorials/image/pixeldit/pixeldit.mdx | 4 +- zh/tutorials/image/qwen/qwen-image-2512.mdx | 22 +- .../image/qwen/qwen-image-edit-2511.mdx | 23 +- zh/tutorials/image/qwen/qwen-image-edit.mdx | 23 +- .../image/qwen/qwen-image-layered.mdx | 24 +- zh/tutorials/image/qwen/qwen-image.mdx | 81 +++---- zh/tutorials/image/z-image/z-image-turbo.mdx | 8 +- zh/tutorials/image/z-image/z-image.mdx | 6 +- zh/tutorials/llm/gemma4/gemma4.mdx | 4 +- zh/tutorials/llm/qwen/qwen3.mdx | 6 +- zh/tutorials/llm/qwen/qwen3_5.mdx | 6 +- zh/tutorials/partner-nodes/google/gemini.mdx | 12 +- .../kling/kling-motion-control.mdx | 5 +- .../moonvalley-video-generation.mdx | 24 +- zh/tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 +- .../partner-nodes/runway/video-generation.mdx | 12 +- .../partner-nodes/tripo/model-generation.mdx | 24 +- zh/tutorials/utility/depth-anything-3.mdx | 8 +- .../utility/face-detection/mediapipe.mdx | 2 +- zh/tutorials/utility/moge.mdx | 4 +- .../utility/pose-detection-sdpose.mdx | 6 +- .../utility/remove-background-birefnet.mdx | 2 +- zh/tutorials/utility/video-segment-sam3.mdx | 2 +- .../utility/void-video-inpainting.mdx | 12 +- zh/tutorials/video/bytedance/bernini-r.mdx | 8 +- .../cosmos/cosmos-predict2-video2world.mdx | 57 +++-- .../video/hunyuan/hunyuan-video-1-5.mdx | 10 +- zh/tutorials/video/hunyuan/hunyuan-video.mdx | 14 +- zh/tutorials/video/kandinsky/kandinsky-5.mdx | 64 ++++-- zh/tutorials/video/ltxv.mdx | 32 ++- zh/tutorials/video/wan/fun-camera.mdx | 98 +++++--- zh/tutorials/video/wan/fun-control.mdx | 10 +- zh/tutorials/video/wan/fun-inp.mdx | 12 +- zh/tutorials/video/wan/vace.mdx | 10 +- zh/tutorials/video/wan/wan-ati.mdx | 10 +- zh/tutorials/video/wan/wan-causal-forcing.mdx | 8 +- zh/tutorials/video/wan/wan-dancer.mdx | 12 +- zh/tutorials/video/wan/wan-flf.mdx | 10 +- zh/tutorials/video/wan/wan-move.mdx | 42 ++-- zh/tutorials/video/wan/wan-video.mdx | 14 +- zh/tutorials/video/wan/wan2-2-animate.mdx | 57 +++-- zh/tutorials/video/wan/wan2-2-fun-camera.mdx | 57 +++-- zh/tutorials/video/wan/wan2-2-fun-control.mdx | 72 +++--- zh/tutorials/video/wan/wan2-2-fun-inp.mdx | 53 +++-- zh/tutorials/video/wan/wan2-2-s2v.mdx | 67 ++++-- zh/tutorials/video/wan/wan2_2.mdx | 82 +++---- zh/tutorials/video/zai/scail2.mdx | 14 +- 290 files changed, 4839 insertions(+), 2871 deletions(-) diff --git a/ja/tutorials/3d/hunyuan3D-2.mdx b/ja/tutorials/3d/hunyuan3D-2.mdx index 35f163849..d815f9748 100644 --- a/ja/tutorials/3d/hunyuan3D-2.mdx +++ b/ja/tutorials/3d/hunyuan3D-2.mdx @@ -57,9 +57,14 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して - -

Run on Comfy Cloud

- + + + Comfy Cloud でこのワークフローをすぐに実行 + + + ワークフロー JSON ファイルをダウンロード + + ### 1. ワークフロー @@ -81,7 +86,7 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます ``` ComfyUI/ @@ -104,9 +109,14 @@ ComfyUI/ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを使用して 3D モデルを生成します。このモデルは Hunyuan3D-2mv のステップ蒸留バージョンで、より高速な 3D モデル生成を可能にします。このバージョンのワークフローでは、`cfg` を 1.0 に設定し、`flux guidance` ノードを追加して `distilled cfg` 生成を制御します。 - -

Run on Comfy Cloud

- + + + Comfy Cloud でこのワークフローをすぐに実行 + + + ワークフロー JSON ファイルをダウンロード + + ### 1. ワークフロー @@ -125,7 +135,7 @@ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます ``` ComfyUI/ @@ -146,9 +156,14 @@ ComfyUI/ Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D モデルを生成します。このモデルはマルチビューモデルではありません。このワークフローでは、`Hunyuan3Dv2ConditioningMultiView` ノードの代わりに `Hunyuan3Dv2Conditioning` ノードを使用します。 - -

Run on Comfy Cloud

- + + + Comfy Cloud でこのワークフローをすぐに実行 + + + ワークフロー JSON ファイルをダウンロード + + ### 1. ワークフロー @@ -163,7 +178,7 @@ Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます ``` ComfyUI/ diff --git a/ja/tutorials/3d/triposplat.mdx b/ja/tutorials/3d/triposplat.mdx index 3c1dbffce..0db4907a4 100644 --- a/ja/tutorials/3d/triposplat.mdx +++ b/ja/tutorials/3d/triposplat.mdx @@ -106,23 +106,23 @@ TripoSplat は **フィードフォワードアーキテクチャ** を使用し TripoSplat モデルと必要なファイルをダウンロードします。対応する `models/` サブディレクトリに配置してください。 - + triposplat_fp16.safetensors — TripoSplat 拡散モデルチェックポイント - + triposplat_vae_decoder_fp16.safetensors — VAE デコーダー - + flux2-vae.safetensors — Flux.2 VAE、潜在表現エンコード用 - + dino_v3_vit_h.safetensors — CLIP ビジョンエンコーダー（DINOv2） - + birefnet.safetensors — 前処理用の背景除去モデル diff --git a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx index 97dd4f880..72d03af91 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -45,7 +45,7 @@ AIO（All-in-One）版は、すべてのモデルを単一のチェックポイ ### AIOモデルのダウンロード - + オールインワンチェックポイントファイル（大多数のユーザーに推奨）。 @@ -74,19 +74,19 @@ AIO（All-in-One）版は、すべてのモデルを単一のチェックポイ ### 分割モデルのダウンロード - + 拡散モデル（Diffusion Model）。 - + テキストエンコーダー（0.6B）。 - + テキストエンコーダー（1.7B）。 - + VAEモデル。 diff --git a/ja/tutorials/audio/ace-step/ace-step-v1.mdx b/ja/tutorials/audio/ace-step/ace-step-v1.mdx index d558882bf..1e36359f5 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1.mdx @@ -33,9 +33,15 @@ ACE-Step は、中国のチーム StepFun と ACE Studio が共同開発した 以下のボタンをクリックして、対応するワークフローファイルをダウンロードしてください。ダウンロードしたファイルを ComfyUI にドラッグ＆ドロップすることで、ワークフロー情報が読み込まれます。このワークフローには、モデルのダウンロード情報も含まれています。 - -

JSON 形式のワークフローファイルをダウンロード

- + + + JSON 形式のワークフローファイルをダウンロード + + また、[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors) を手動でダウンロードし、`ComfyUI/models/checkpoints` フォルダーに保存することもできます。 @@ -58,15 +64,27 @@ ACE-Step は、中国のチーム StepFun と ACE Studio が共同開発した 以下のボタンをクリックして、対応するワークフローファイルをダウンロードしてください。ダウンロードしたファイルを ComfyUI にドラッグ＆ドロップすることで、ワークフロー情報が読み込まれます。 - -

JSON 形式のワークフローファイルをダウンロード

- + + + JSON 形式のワークフローファイルをダウンロード + + 以下の音声ファイルを入力音声としてダウンロードしてください： - -

入力用のサンプル音声ファイルをダウンロード

- + + + 入力用のサンプル音声ファイルをダウンロード + + ### 2. ワークフローをステップごとに実行 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx index a0c3bcfa4..258bd2f9f 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -41,7 +41,7 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" ### チェックポイント - + 2.3GB。models/checkpoints/ に配置 @@ -56,7 +56,7 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" ### テキストエンコーダー - + プロンプト処理用テキストエンコーダー。models/text_encoders/ に配置 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx index e638d2d56..c21d44138 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -77,11 +77,11 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 ### チェックポイント - + Medium ワークフロー用。models/checkpoints/ に配置 - + Medium Base ワークフロー用。models/checkpoints/ に配置 @@ -97,11 +97,11 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 ### テキストエンコーダー - + すべての Stable Audio 3 ワークフローで必要。models/text_encoders/ に配置 - + Medium ワークフローで必要（Qwen リプロンプト）。models/text_encoders/ に配置 diff --git a/ja/tutorials/basic/inpaint.mdx b/ja/tutorials/basic/inpaint.mdx index d97d65da2..2d4a7298f 100644 --- a/ja/tutorials/basic/inpaint.mdx +++ b/ja/tutorials/basic/inpaint.mdx @@ -35,7 +35,7 @@ AIによる画像生成において、全体としては満足できる画像が #### 1. モデルのインストール 以下のファイルをダウンロードし、`ComfyUI/models/checkpoints` フォルダに保存してください： -[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) +[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) #### 2. 局部再描画用の入力画像 @@ -69,7 +69,7 @@ AIによる画像生成において、全体としては満足できる画像が ![SD1.5 インペインティング結果](/images/tutorial/basic/inpaint/inpaint_sd1.5_pruned_emaonly.png) -一方、[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) モデルを用いた結果は、より高品質なインペインティング効果と、自然な境界のトランジションが得られます。 +一方、[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) モデルを用いた結果は、より高品質なインペインティング効果と、自然な境界のトランジションが得られます。 これは、当該モデルが**インペインティング専用に設計・最適化**されているためであり、生成領域をより正確に制御でき、結果として優れた再描画品質を実現します。 先ほどご紹介した「画家」というアナロジーを思い出してください。異なるモデルは、それぞれ得意分野や限界を持つ「画家」のような存在です。適切なモデルを選択することで、より理想的な生成結果を得ることが可能になります。 diff --git a/ja/tutorials/basic/outpaint.mdx b/ja/tutorials/basic/outpaint.mdx index 9a96f5004..4d09d7120 100644 --- a/ja/tutorials/basic/outpaint.mdx +++ b/ja/tutorials/basic/outpaint.mdx @@ -34,7 +34,7 @@ AI画像生成においては、既存の画像の構図が優れているもの #### 1. モデルのインストール 以下のモデルファイルをダウンロードし、`ComfyUI/models/checkpoints` ディレクトリに保存してください： -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) #### 2. 入力画像 diff --git a/ja/tutorials/controlnet/controlnet.mdx b/ja/tutorials/controlnet/controlnet.mdx index 42acac85e..bacd4e2ec 100644 --- a/ja/tutorials/controlnet/controlnet.mdx +++ b/ja/tutorials/controlnet/controlnet.mdx @@ -76,8 +76,8 @@ ControlNet の登場により、追加の条件を導入することで画像生 - [dreamCreationVirtual3DECommerce_v10.safetensors](https://civitai.com/api/download/models/731340?type=Model&format=SafeTensor&size=full&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/depth-controlnet.mdx b/ja/tutorials/controlnet/depth-controlnet.mdx index 19637bc3d..ea0da6d71 100644 --- a/ja/tutorials/controlnet/depth-controlnet.mdx +++ b/ja/tutorials/controlnet/depth-controlnet.mdx @@ -54,7 +54,7 @@ Depth ControlNet は、深度マップの情報を理解・活用するために - [architecturerealmix_v11.safetensors](https://civitai.com/api/download/models/431755?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) +- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/depth-t2i-adapter.mdx b/ja/tutorials/controlnet/depth-t2i-adapter.mdx index f6ae26c4c..2541de169 100644 --- a/ja/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/ja/tutorials/controlnet/depth-t2i-adapter.mdx @@ -76,7 +76,7 @@ ComfyUI における T2I Adapter の使用方法は、インターフェース - [interiordesignsuperm_v2.safetensors](https://civitai.com/api/download/models/93152?type=Model&format=SafeTensor&size=full&fp=fp16) -- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/resolve/main/models/t2iadapter_depth_sd15v2.pth?download=true) +- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_depth_sd15v2.pth?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/mixing-controlnets.mdx b/ja/tutorials/controlnet/mixing-controlnets.mdx index a9990bd85..7f65a43fd 100644 --- a/ja/tutorials/controlnet/mixing-controlnets.mdx +++ b/ja/tutorials/controlnet/mixing-controlnets.mdx @@ -54,9 +54,9 @@ AI による画像生成において、単一の制御条件では複雑なシ - [awpainting_v14.safetensors](https://civitai.com/api/download/models/624939?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx b/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx index aacef9020..b4dca8f4d 100644 --- a/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -49,10 +49,10 @@ AI による画像生成において、OpenPose で生成された骨格構造 ネットワーク環境によっては、対応するモデルの自動ダウンロードが失敗する場合があります。その場合は、以下のモデルを手動でダウンロードし、指定されたディレクトリに配置してください： -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) - [majicmixRealistic_v7.safetensors](https://civitai.com/api/download/models/176425?type=Model&format=SafeTensor&size=pruned&fp=fp16) - [japaneseStyleRealistic_v20.safetensors](https://civitai.com/api/download/models/85426?type=Model&format=SafeTensor&size=pruned&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/flux/flux-1-controlnet.mdx b/ja/tutorials/flux/flux-1-controlnet.mdx index 39c15df7e..9448ed209 100644 --- a/ja/tutorials/flux/flux-1-controlnet.mdx +++ b/ja/tutorials/flux/flux-1-controlnet.mdx @@ -49,9 +49,14 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 ## FLUX.1-Canny-dev 完全版ワークフロー - -

Comfy Cloud で実行

- + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Canny」を検索してください + + + Comfy Cloud で開く + + ### 1. ワークフローおよび関連アセット @@ -72,10 +77,10 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） ファイルの保存先ディレクトリ構成： ``` @@ -115,9 +120,14 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora ワークフロー - -

Comfy Cloud で実行

- + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Depth LoRA」を検索してください + + + Comfy Cloud で開く + + LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを追加したものであり、[Flux ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image) と比較して、対応する LoRA モデルを読み込むためのノードが追加されています。 @@ -138,11 +148,11 @@ LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) ファイルの保存先ディレクトリ構成： ``` diff --git a/ja/tutorials/flux/flux-1-fill-dev.mdx b/ja/tutorials/flux/flux-1-fill-dev.mdx index 9e8b9f12e..b6bb401b4 100644 --- a/ja/tutorials/flux/flux-1-fill-dev.mdx +++ b/ja/tutorials/flux/flux-1-fill-dev.mdx @@ -39,10 +39,10 @@ Inpainting や Outpainting のワークフローについてまだご存じな ![Flux Agreement](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) ファイルの保存場所： ``` @@ -61,13 +61,14 @@ ComfyUI/ ### 1. Inpainting ワークフローおよび関連アセット - -

ワークフロー画像をダウンロード

- - - -

Comfy Cloud で実行

- + + + Download JSON or search "flux_fill_inpaint" in Template Library + + + Open in Comfy Cloud + + 以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 ![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -96,6 +97,15 @@ ComfyUI/ ### 1. Outpainting ワークフローおよび関連アセット + + + Download JSON or search "flux_fill_outpaint" in Template Library + + + Open in Comfy Cloud + + + 以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 ![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/ja/tutorials/flux/flux-1-kontext-dev.mdx b/ja/tutorials/flux/flux-1-kontext-dev.mdx index b9537bda5..5933115ec 100644 --- a/ja/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ja/tutorials/flux/flux-1-kontext-dev.mdx @@ -51,7 +51,7 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： **Diffusion Model（拡散モデル）** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) オリジナルの重み（weights）を使用したい場合は、Black Forest Labs の関連リポジトリからオリジナルモデルの重みを取得・利用できます。 @@ -62,7 +62,7 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： **Text Encoder（テキストエンコーダー）** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) または [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) または [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) モデルの保存先 @@ -80,9 +80,14 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： ## Flux.1 Kontext Dev ワークフロー - -

Comfy Cloud で実行

- + + + JSONをダウンロードするか、テンプレートライブラリで「Flux Kontext Dev」を検索してください + + + Comfy Cloud で開く + + このワークフローでは、編集対象の画像を読み込むために `Load Image(from output)` ノードを採用しており、編集後の画像を容易に取得・再利用できるため、複数回の反復編集がよりスムーズに行えます。 diff --git a/ja/tutorials/flux/flux-1-text-to-image.mdx b/ja/tutorials/flux/flux-1-text-to-image.mdx index 51b98dd27..1e19bd6f7 100644 --- a/ja/tutorials/flux/flux-1-text-to-image.mdx +++ b/ja/tutorials/flux/flux-1-text-to-image.mdx @@ -50,25 +50,30 @@ Flux は、優れた画像品質と高い柔軟性で知られており、高品 #### 1. ワークフロー・ファイル + + + Comfy Cloud でこのワークフローを実行 + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Dev」を検索 + + + 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Dev オリジナルバージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - -

Comfy Cloud で実行

- - #### 2. モデルの手動インストール - `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) のライセンス契約に同意する必要があります。 -- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 +- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 以下のモデルファイルをダウンロードしてください： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) 保存先のディレクトリ構成： ``` @@ -103,14 +108,19 @@ Flux の優れたプロンプト追従能力により、負のプロンプト（ #### 1. ワークフロー・ファイル + + + Comfy Cloud でこのワークフローを実行 + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Schnell」を検索 + + + 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Schnell バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) - -

Comfy Cloud で実行

- - #### 2. モデルの手動インストール @@ -120,10 +130,10 @@ Flux の優れたプロンプト追従能力により、負のプロンプト（ 完全なモデルファイル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) ファイルの保存先ディレクトリ構成： ``` @@ -156,24 +166,38 @@ FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したも ### Flux.1 Dev + + + Comfy Cloud でこのワークフローを実行 + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Dev FP8」を検索 + + + 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Dev fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) - -

Comfy Cloud で実行

- - -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 +[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 対応する `Load Checkpoint` ノードが `flux1-dev-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 ### Flux.1 Schnell + + + Comfy Cloud でこのワークフローを実行 + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Schnell FP8」を検索 + + + 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Schnell fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 +[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 対応する `Load Checkpoint` ノードが `flux1-schnell-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 diff --git a/ja/tutorials/flux/flux-1-uso.mdx b/ja/tutorials/flux/flux-1-uso.mdx index 9e8790c9a..81c03b2c7 100644 --- a/ja/tutorials/flux/flux-1-uso.mdx +++ b/ja/tutorials/flux/flux-1-uso.mdx @@ -31,18 +31,14 @@ USO は以下の3つの主要なアプローチをサポートします： ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + ワークフロー JSON をダウンロードし、ComfyUI にドラッグ＆ドロップしてください + + + Comfy Cloud でこのワークフローを実行 + + 以下の画像を入力画像として使用します。 @@ -52,18 +48,18 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', **checkpoints** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) **loras** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **model_patches** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **clip_visions** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) 上記すべてのモデルをダウンロードし、以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/flux/flux-2-dev.mdx b/ja/tutorials/flux/flux-2-dev.mdx index 653f78180..2667de564 100644 --- a/ja/tutorials/flux/flux-2-dev.mdx +++ b/ja/tutorials/flux/flux-2-dev.mdx @@ -65,15 +65,15 @@ FLUX.2 Dev を用いた基本的なテキストから画像への生成ワーク **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) **モデルの保存先** diff --git a/ja/tutorials/flux/flux-2-klein.mdx b/ja/tutorials/flux/flux-2-klein.mdx index c12f18487..c1a373245 100644 --- a/ja/tutorials/flux/flux-2-klein.mdx +++ b/ja/tutorials/flux/flux-2-klein.mdx @@ -45,19 +45,19 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 ## Flux.2 Klein 4B モデルのダウンロード - + 4B モデル用のテキストエンコーダーです。 - + 拡散モデル（4B Base 版）。 - + 拡散モデル（4B 蒸留版）。 - + 4B モデル用のVAEです。 @@ -103,11 +103,11 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 拡散モデル（9B 蒸留版）。 - + 9B モデル用のテキストエンコーダーです。 - + 9B モデル用のVAEです。 diff --git a/ja/tutorials/flux/flux1-krea-dev.mdx b/ja/tutorials/flux/flux1-krea-dev.mdx index 827438849..7ba34a270 100644 --- a/ja/tutorials/flux/flux1-krea-dev.mdx +++ b/ja/tutorials/flux/flux1-krea-dev.mdx @@ -31,13 +31,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下の画像または JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 ![Flux Krea Dev ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Comfy Cloud でこのワークフローを実行 + + + JSON をダウンロードするか、テンプレート ライブラリで「Flux.1 Krea Dev」を検索してください + + #### 2. 手動によるモデルのインストール @@ -50,7 +51,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' より高品質な出力を求め、かつ十分な VRAM をお持ちの場合、オリジナルの重みファイルも試すことができます： -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) の利用規約に同意する必要があります。 @@ -59,12 +60,12 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以前に Flux 関連のワークフローをご利用済みの場合、以下のモデルは既に存在するため、再ダウンロードの必要はありません。 **テキストエンコーダー** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) — VRAM が 32GB を超える環境で推奨 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) — 低 VRAM 環境向け +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true): VRAM が 32GB を超える環境で推奨 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors): 低 VRAM 環境向け **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) ファイルの保存先： ``` diff --git a/ja/tutorials/image/anima/anima.mdx b/ja/tutorials/image/anima/anima.mdx index e78fb3dca..e10df10db 100644 --- a/ja/tutorials/image/anima/anima.mdx +++ b/ja/tutorials/image/anima/anima.mdx @@ -82,15 +82,15 @@ Anima は 2 つのワークフローを提供しています——標準的な すべてのモデルファイルは Hugging Face の [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima) から入手できます。 - + Anima Base v1 用拡散モデル（2B）。 - + 両方のワークフローで共有されるテキストエンコーダー（Qwen-3 0.6B）。 - + 両方のワークフローで共有される VAE。 @@ -111,7 +111,7 @@ Anima は 2 つのワークフローを提供しています——標準的な Preview ワークフローを使用する場合は、代わりに以下のプレビュー拡散モデルをダウンロードしてください： - + Anima Preview 用拡散モデル（2B）。 diff --git a/ja/tutorials/image/boogu/boogu-image-0.1.mdx b/ja/tutorials/image/boogu/boogu-image-0.1.mdx index 7b75c26ed..b15572a40 100644 --- a/ja/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ja/tutorials/image/boogu/boogu-image-0.1.mdx @@ -50,19 +50,19 @@ Boogu-Image-0.1-Turbo ワークフローは、拡散、テキストエンコー ### Boogu-Image-0.1-Turbo モデルのダウンロード - + Boogu-Image-0.1-Turbo 用の拡散モデル。 - + Boogu-Image-0.1-Turbo 用のテキストエンコーダー。 - + Boogu-Image-0.1-Turbo 用の VAE。 - + Boogu-Image-0.1-Turbo 用の LoRA モジュール (rank-128)。 @@ -100,15 +100,15 @@ Boogu-Image-0.1-Turbo ワークフローは、拡散、テキストエンコー ### Boogu-Image-0.1-Edit モデルのダウンロード - + Boogu-Image-0.1-Edit 用の拡散モデル。 - + Boogu-Image-0.1-Edit 用のテキストエンコーダー。 - + Boogu-Image-0.1-Edit 用の VAE。 diff --git a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index 7b1f7a47c..b4d4f2486 100644 --- a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -43,17 +43,17 @@ Hugging Face: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos **Diffusion モデル** -- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors) +- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_t2i.safetensors) その他の重みファイルについては、[Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) からダウンロードしてください。 **テキストエンコーダー** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) ファイルの保存場所 diff --git a/ja/tutorials/image/ernie-image/ernie-image.mdx b/ja/tutorials/image/ernie-image/ernie-image.mdx index b02fa0f45..023097979 100644 --- a/ja/tutorials/image/ernie-image/ernie-image.mdx +++ b/ja/tutorials/image/ernie-image/ernie-image.mdx @@ -55,19 +55,19 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' リパッケージされたすべてのモデルファイルは、Hugging Face の [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) で入手できます。 - + ERNIE-Image 用拡散モデル。 - + ERNIE-Image 用テキストエンコーダー。 - + ERNIE-Image 用プロンプトエンハンサーテキストエンコーダー。 - + ERNIE-Image 用 VAE。 @@ -99,19 +99,19 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### ERNIE-Image-Turbo モデルのダウンロード - + ERNIE-Image-Turbo 用拡散モデル。 - + ERNIE-Image-Turbo 用テキストエンコーダー。 - + ERNIE-Image-Turbo 用プロンプトエンハンサーテキストエンコーダー。 - + ERNIE-Image-Turbo 用 VAE。 diff --git a/ja/tutorials/image/hidream/hidream-e1.mdx b/ja/tutorials/image/hidream/hidream-e1.mdx index 47798c64a..3707adfc3 100644 --- a/ja/tutorials/image/hidream/hidream-e1.mdx +++ b/ja/tutorials/image/hidream/hidream-e1.mdx @@ -40,8 +40,8 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン **Diffusion モデル** 両方のモデルを同時にダウンロードする必要はありません。E1.1 は E1 をベースとした改良版であり、実際のテスト結果から、品質およびパフォーマンスの両面で E1 を大幅に上回ることが確認されています。 -- [hidream_e1_1_bf16.safetensors（推奨）](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors（推奨）](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **テキストエンコーダー**： @@ -74,6 +74,15 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン ## HiDream E1.1 の ComfyUI ネイティブ ワークフローの例 + + + Comfy Cloud で開く + + + JSON をダウンロード、またはテンプレートライブラリで "HiDream E1.1" を検索 + + + E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メガピクセル解像度** をサポートしています。ワークフローでは `Scale Image to Total Pixels` ノードを用いて、入力画像を自動的に 100万ピクセルにスケーリングします。 @@ -118,9 +127,14 @@ E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メ ## HiDream E1 の ComfyUI ネイティブ ワークフローの例 - -

Comfy Cloud で実行

- + + + Comfy Cloud で開く + + + JSON をダウンロード、またはテンプレートライブラリで "HiDream E1 Full" を検索 + + E1 は 2025年4月28日にリリースされたモデルで、**768×768 の固定解像度のみ** をサポートします。 diff --git a/ja/tutorials/image/hidream/hidream-i1.mdx b/ja/tutorials/image/hidream/hidream-i1.mdx index 2e7c76364..ed3432e68 100644 --- a/ja/tutorials/image/hidream/hidream-i1.mdx +++ b/ja/tutorials/image/hidream/hidream-i1.mdx @@ -101,16 +101,21 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Full バージョンのワークフロー - -

Comfy Cloud で実行

- + + + 設定不要ですぐに Comfy Cloud で実行 + + + ワークフローの JSON ファイルをダウンロード + + #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード @@ -139,16 +144,21 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Dev バージョンのワークフロー - -

Comfy Cloud で実行

- + + + 設定不要ですぐに Comfy Cloud で実行 + + + ワークフローの JSON ファイルをダウンロード + + #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード @@ -177,16 +187,21 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Fast バージョンのワークフロー - -

Comfy Cloud で実行

- + + + 設定不要ですぐに Comfy Cloud で実行 + + + ワークフローの JSON ファイルをダウンロード + + #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード diff --git a/ja/tutorials/image/hidream/hidream-o1.mdx b/ja/tutorials/image/hidream/hidream-o1.mdx index 6295428f3..875b0d90c 100644 --- a/ja/tutorials/image/hidream/hidream-o1.mdx +++ b/ja/tutorials/image/hidream/hidream-o1.mdx @@ -51,19 +51,19 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` **チェックポイント** — 再パッケージおよび量子化済み。すべてのバージョンで最悪の外れ値は bf16 で保持され、未使用の deepstack 層は削除されています： -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量子化版 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 フル精度版（最大ファイル） +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量子化版 +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 フル精度版（最大ファイル） **テキストエンコーダ（プロンプト補強）** — 全バージョン共通： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA（オプション）** — Dev 蒸留は LoRA として Full モデルにも適用でき、蒸留強度を調整できます（[Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — フルランク -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — プルーニング版 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 代替 Checkpoint ベースの蒸留 +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — フルランク +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — プルーニング版 +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 代替 Checkpoint ベースの蒸留 ``` 📂 ComfyUI/ @@ -107,13 +107,13 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` **チェックポイント（Dev）** — 再パッケージおよび量子化済み。すべてのバージョンで最悪の外れ値は bf16 で保持され、未使用の deepstack 層は削除されています： -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量子化版 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 フル精度版（最大ファイル） +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量子化版 +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 フル精度版（最大ファイル） **テキストエンコーダ（プロンプト補強）** — 全バージョン共通： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/ja/tutorials/image/ideogram/ideogram-v4.mdx b/ja/tutorials/image/ideogram/ideogram-v4.mdx index 2c44d5ea6..6f2a5ec82 100644 --- a/ja/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ja/tutorials/image/ideogram/ideogram-v4.mdx @@ -46,23 +46,23 @@ Ideogram 4.0 は、Ideogram がオープンソースモデルとして公開し Hugging Face の [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) ですべての再パッケージ化されたモデルファイルを見つけることができます。 - + Ideogram 4.0 拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 - + Ideogram 4.0 条件なし拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 - + Ideogram 4.0 テキストエンコーダー（~8 GB）。models/text_encoders/ に配置 - + Ideogram 4.0 テキストエンコーダー（~2 GB）。models/text_encoders/ に配置 - + Ideogram 4.0 VAE（~335 MB）。models/vae/ に配置 diff --git a/ja/tutorials/image/krea/krea-2.mdx b/ja/tutorials/image/krea/krea-2.mdx index 32a6f8bd9..88aa59318 100644 --- a/ja/tutorials/image/krea/krea-2.mdx +++ b/ja/tutorials/image/krea/krea-2.mdx @@ -108,13 +108,13 @@ Krea 2向けのスタイルLoRAコレクションもKreaから公開されてい ローカルで使用するには、[Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2) からComfyUI最適化モデルファイルをダウンロードしてください。 - + krea2_turbo_fp8_scaled.safetensors: Turbo FP8 (推奨、ほとんどのユーザー向け) - + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B テキストエンコーダー - + qwen_image_vae.safetensors diff --git a/ja/tutorials/image/lens/lens.mdx b/ja/tutorials/image/lens/lens.mdx index edd441dec..90dea6b56 100644 --- a/ja/tutorials/image/lens/lens.mdx +++ b/ja/tutorials/image/lens/lens.mdx @@ -94,19 +94,19 @@ Lens Turbo は蒸留版で、より少ないサンプリングステップで画 すべてのモデルファイルは Hugging Face の [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens) にあります。 - + Lens 用拡散モデル (BF16) - + Lens Turbo 用拡散モデル (BF16) - + Lens と Lens Turbo で共通のテキストエンコーダー (GPT-OSS-20B) - + Lens と Lens Turbo で共通の VAE (FLUX.2) diff --git a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 02b2594ef..d607b2bc5 100644 --- a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -30,13 +30,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## NewBie-image 文生成画像ワークフロー - -

JSON ワークフローファイルをダウンロード

- - - -

ComfyUI Cloud で実行

- +| +| +| JSONをダウンロードするか、テンプレートライブラリで「NewBie-image」を検索してください +| +| +| クラウドで開く +| +| @@ -44,16 +45,16 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/omnigen/omnigen2.mdx b/ja/tutorials/image/omnigen/omnigen2.mdx index f9c0a2d72..b0c305f56 100644 --- a/ja/tutorials/image/omnigen/omnigen2.mdx +++ b/ja/tutorials/image/omnigen/omnigen2.mdx @@ -1,6 +1,6 @@ --- title: "ComfyUI OmniGen2 ネイティブワークフローの例" -description: "ComfyUI OmniGen2 ネイティブワークフローの例 — 文字から画像生成、画像編集、および複数画像の合成を統合したモデル。" +description: "ComfyUI OmniGen2 ネイティブワークフローの例 - 文字から画像生成、画像編集、および複数画像の合成を統合したモデル。" sidebarTitle: "OmniGen2" translationSourceHash: b061ff8c translationFrom: tutorials/image/omnigen/omnigen2.mdx @@ -42,13 +42,13 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 本記事では複数のワークフローを取り扱うため、対応するモデルファイルおよびインストール先は以下の通りです。各ワークフロー内にも、該当するモデルファイルのダウンロード情報が記載されています。 **拡散モデル（Diffusion Models）** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **テキストエンコーダー（Text Encoders）** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) ファイル保存先： ``` @@ -66,9 +66,11 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 ### 1. ワークフローファイルのダウンロード - -

Comfy Cloud で実行

- + + + Comfy Cloud で実行 + + ![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -94,9 +96,11 @@ OmniGen2 は豊富な画像編集機能を備えており、画像へのテキ ### 1. ワークフローファイルのダウンロード - -

Comfy Cloud で実行

- + + + Comfy Cloud で実行 + + ![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/ja/tutorials/image/ovis/ovis-image.mdx b/ja/tutorials/image/ovis/ovis-image.mdx index ecbad6885..78f8b80e1 100644 --- a/ja/tutorials/image/ovis/ovis-image.mdx +++ b/ja/tutorials/image/ovis/ovis-image.mdx @@ -22,13 +22,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Ovis-Image のテキストから画像を生成するワークフロー - -

JSONワークフローファイルをダウンロード

- - - -

ComfyUI Cloud 上で実行

- + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで「Ovis image」を検索してください + + @@ -36,15 +37,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders（テキストエンコーダ）** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models（拡散モデル）** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/pixeldit/pixeldit.mdx b/ja/tutorials/image/pixeldit/pixeldit.mdx index 4cc67f103..01869d6de 100644 --- a/ja/tutorials/image/pixeldit/pixeldit.mdx +++ b/ja/tutorials/image/pixeldit/pixeldit.mdx @@ -62,11 +62,11 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' PixelDiT はテキストエンコーダーと拡散モデルの 2 つのモデルファイルを使用します。 - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT テキストエンコーダー - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 拡散モデル diff --git a/ja/tutorials/image/qwen/qwen-image-2512.mdx b/ja/tutorials/image/qwen/qwen-image-2512.mdx index a144f0632..1c5c6f755 100644 --- a/ja/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ja/tutorials/image/qwen/qwen-image-2512.mdx @@ -43,9 +43,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - + + Comfy Cloud で実行 - + + + JSON をダウンロードするか、テンプレートライブラリで "Qwen-Image-2512" を検索してください + + ### 1. ワークフローファイル @@ -55,28 +60,25 @@ ComfyUI を更新した後、テンプレートからワークフローファイ - **Text to Image (Qwen-Image 2512)**: 標準的な50ステップ生成 - **Text to Image (Qwen-Image 2512 4steps)**: Lightning LoRA を用いた高速4ステップ生成 - -

JSON ワークフローをダウンロード

- ### 2. モデルのダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（任意：4ステップ Lightning 加速用）** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **拡散モデル** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（ほとんどのユーザーに推奨） -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（VRAM が十分に確保でき、より高品質な出力を求めている場合） +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（ほとんどのユーザーに推奨） +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（VRAM が十分に確保でき、より高品質な出力を求めている場合） **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx index 946f5c718..7d1768c29 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -33,31 +33,32 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ComfyUI を更新した後、テンプレートからワークフローファイルを取得できます。あるいは、以下のワークフローを ComfyUI にドラッグ＆ドロップして読み込むこともできます。 - -

JSON 形式ワークフローをダウンロード

- - - -

ComfyUI Cloud 上で実行

- + + + ComfyUI Cloud 上で実行 + + + JSON 形式ワークフローをダウンロード + + ### 2. モデルのダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（任意：4 ステップ Lightning 加速用）** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **拡散モデル** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image-edit.mdx b/ja/tutorials/image/qwen/qwen-image-edit.mdx index 01cbdb893..bd7f93969 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit.mdx @@ -47,13 +47,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ComfyUI を更新後、テンプレートからワークフローファイルを取得するか、下記のワークフローを ComfyUI へドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

JSON形式ワークフローをダウンロード

- - - -

ComfyUI Cloud 上で実行

- + + + JSON形式ワークフローをダウンロードするか、テンプレートライブラリで"image_qwen_image_edit"を検索してください + + + ComfyUI Cloud 上でこのワークフローを実行（ゼロセットアップ） + + 以下の画像を入力画像としてダウンロードしてください。 ![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -64,19 +65,19 @@ ComfyUI を更新後、テンプレートからワークフローファイルを **Diffusion モデル** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **テキストエンコーダ** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) モデルの保存場所 diff --git a/ja/tutorials/image/qwen/qwen-image-layered.mdx b/ja/tutorials/image/qwen/qwen-image-layered.mdx index b2c05499a..28749f22e 100644 --- a/ja/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ja/tutorials/image/qwen/qwen-image-layered.mdx @@ -30,13 +30,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Qwen-Image-Layered ワークフロー - -

JSON ワークフローファイルをダウンロード

- - - -

ComfyUI Cloud で実行

- +| +| +| JSON ワークフローファイルをダウンロード +| +| +| +| ComfyUI Cloud で実行 +| +| @@ -44,15 +46,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **モデルの保存場所** @@ -87,4 +89,4 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### プロンプト（任意） -テキストプロンプトは、入力画像全体の内容を記述することを目的としています——たとえば、前景オブジェクトの後ろに隠れている文字など、部分的に遮蔽されている要素も含めて指定可能です。ただし、個々のレイヤーの意味的内容を明示的に制御するためのものではありません。 +テキストプロンプトは、入力画像全体の内容を記述することを目的としています。たとえば、前景オブジェクトの後ろに隠れている文字など、部分的に遮蔽されている要素も含めて指定可能です。ただし、個々のレイヤーの意味的内容を明示的に制御するためのものではありません。 diff --git a/ja/tutorials/image/qwen/qwen-image.mdx b/ja/tutorials/image/qwen/qwen-image.mdx index 7fb6475cc..bb4f04855 100644 --- a/ja/tutorials/image/qwen/qwen-image.mdx +++ b/ja/tutorials/image/qwen/qwen-image.mdx @@ -58,9 +58,12 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - - Comfy Cloudで実行 - + + + + + + 本ドキュメントに添付されたワークフローでは、以下の3種類の異なるモデルが使用されています： 1. Qwen-Imageオリジナルモデル（fp8_e4m3fn） @@ -82,14 +85,11 @@ GPU：RTX4090D（24GB） ComfyUIを更新後、テンプレートからワークフローファイルを検索するか、以下のワークフローをComfyUIにドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像へ変換するワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

Qwen-Image公式モデル用ワークフロー（JSON形式）をダウンロード

- + 蒸留版 - -

蒸留モデル用ワークフロー（JSON形式）をダウンロード

- + + ### 2. モデルのダウンロード @@ -103,12 +103,12 @@ ComfyUIを更新後、テンプレートからワークフローファイルを **拡散モデル** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill（蒸留版） -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 蒸留版のオリジナル作者は、CFG値1.0で15ステップでの使用を推奨しています。 @@ -117,15 +117,15 @@ Qwen_image_distill（蒸留版） **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** @@ -164,18 +164,19 @@ Qwen_image_distill（蒸留版） これはControlNetモデルであるため、通常のControlNetとして使用できます。 - - Comfy Cloudで実行 - + + + + + + ### 1. ワークフローおよび入力画像 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

JSON形式ワークフローをダウンロード

- + 以下の画像を入力としてダウンロードしてください ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -183,7 +184,7 @@ Qwen_image_distill（蒸留版） 1. InstantX ControlNet -[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)をダウンロードし、`ComfyUI/models/controlnet/`フォルダーに保存してください +[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)をダウンロードし、`ComfyUI/models/controlnet/`フォルダーに保存してください 2. **Lotus Depthモデル** @@ -191,11 +192,11 @@ Qwen_image_distill（蒸留版） **拡散モデル** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAEモデル** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) または任意のSD1.5互換VAE +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) または任意のSD1.5互換VAE ``` ComfyUI/ @@ -219,9 +220,12 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNetsモデルパッチワークフロー - - Comfy Cloudで実行 - + + + + + + このモデルは実際にはControlNetではなく、Canny、Depth、Inpaintの3種類の異なる制御モードをサポートする「モデルパッチ」です。 @@ -234,9 +238,7 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして対応するワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

JSON形式ワークフローをダウンロード

- + 以下の画像を入力としてダウンロードしてください： @@ -246,9 +248,9 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http その他のモデルはQwen-Image基本ワークフローと同一です。以下のモデルのみをダウンロードし、`ComfyUI/models/model_patches`フォルダーに保存してください。 -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. ワークフローの使用方法 @@ -290,9 +292,12 @@ Inpaintモデルでは、[マスクエディター](/ja/interface/maskeditor)を ## Qwen Image Union ControlNet LoRAワークフロー - - Comfy Cloudで実行 - + + + + + + オリジナルモデルのURL：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org再ホストURL：[qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors)：Canny、Depth、Pose、Lineart、Softedge、Normal、Openposeをサポートする画像構造制御用LoRA @@ -301,9 +306,7 @@ Comfy Org再ホストURL：[qwen_image_union_diffsynth_lora.safetensors](https:/ 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

JSON形式ワークフローをダウンロード

- + 以下の画像を入力としてダウンロードしてください diff --git a/ja/tutorials/image/z-image/z-image-turbo.mdx b/ja/tutorials/image/z-image/z-image-turbo.mdx index fe7d24df0..e77bfd842 100644 --- a/ja/tutorials/image/z-image/z-image-turbo.mdx +++ b/ja/tutorials/image/z-image/z-image-turbo.mdx @@ -46,15 +46,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### Z-Image-Turbo モデルのダウンロード - + Z-Image-Turbo 専用のテキストエンコーダーです。 - + Z-Image-Turbo 専用の拡散モデルです。 - + Z-Image-Turbo 専用のVAE（変分オートエンコーダー）です。 @@ -81,7 +81,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### ControlNet 用の追加モデル - + Z-Image-Turbo 専用のControlNetモデルパッチです。 diff --git a/ja/tutorials/image/z-image/z-image.mdx b/ja/tutorials/image/z-image/z-image.mdx index 733137ed1..a25bc4c77 100644 --- a/ja/tutorials/image/z-image/z-image.mdx +++ b/ja/tutorials/image/z-image/z-image.mdx @@ -37,15 +37,15 @@ Z-Image（Base）は、コミュニティ主導のファインチューニング ## Z-Image モデルのダウンロード - + Z-Image 用テキストエンコーダー。 - + Z-Image 用拡散モデル。 - + Z-Image 用 VAE。 diff --git a/ja/tutorials/llm/gemma4/gemma4.mdx b/ja/tutorials/llm/gemma4/gemma4.mdx index 0234a4b0e..f48575928 100644 --- a/ja/tutorials/llm/gemma4/gemma4.mdx +++ b/ja/tutorials/llm/gemma4/gemma4.mdx @@ -74,11 +74,11 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' Gemma 4 モデルは ComfyUI ではテキストエンコーダー（text encoder）として読み込まれます。該当するモデルファイルをダウンロードし、正しいディレクトリに配置してください： - + 高速・軽量、コンシューマー GPU に最適。 - + バランスの取れた性能。ワークフローのデフォルトモデル。 diff --git a/ja/tutorials/llm/qwen/qwen3.mdx b/ja/tutorials/llm/qwen/qwen3.mdx index 0a2e148a1..874bbebbb 100644 --- a/ja/tutorials/llm/qwen/qwen3.mdx +++ b/ja/tutorials/llm/qwen/qwen3.mdx @@ -71,15 +71,15 @@ Qwen 3.0 は、ComfyUI ワークフロー内で構造化テキスト生成とイ Qwen 3.0 モデルは ComfyUI でテキストエンコーダーとして読み込まれます。モデルファイルは Qwen3.5 と共有されています。ハードウェアに合わせて適切なバージョンを選択してください： - + 軽量版、約 4.5 GB。低 VRAM 環境や高速ダウンロードに最適。 - + サイズと品質のバランス型。ほとんどのコンシューマー GPU に推奨。 - + 最大版、約 19 GB。より高品質な出力、より多くの VRAM が必要。 diff --git a/ja/tutorials/llm/qwen/qwen3_5.mdx b/ja/tutorials/llm/qwen/qwen3_5.mdx index 0a8733d93..12f65e4c5 100644 --- a/ja/tutorials/llm/qwen/qwen3_5.mdx +++ b/ja/tutorials/llm/qwen/qwen3_5.mdx @@ -73,15 +73,15 @@ Qwen3.5 は、視覚的理解とテキスト生成の組み合わせが ComfyUI Qwen3.5 モデルは ComfyUI でテキストエンコーダーとして読み込まれます。ハードウェアに合わせて適切なバージョンを選択してください： - + 軽量版、約 4.5 GB。低 VRAM 環境や高速ダウンロードに最適。 - + サイズと品質のバランス型。ほとんどのコンシューマー GPU に推奨。 - + 最大版、約 19 GB。より高品質な出力、より多くの VRAM が必要。 diff --git a/ja/tutorials/partner-nodes/google/gemini.mdx b/ja/tutorials/partner-nodes/google/gemini.mdx index f1a6c2ec1..ab669a618 100644 --- a/ja/tutorials/partner-nodes/google/gemini.mdx +++ b/ja/tutorials/partner-nodes/google/gemini.mdx @@ -22,13 +22,11 @@ Google Gemini は、Google が開発した強力な AI モデルであり、対 以下の JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式のワークフローファイルをダウンロード

- + + + JSON 形式のワークフローファイルをダウンロード + + ### 2. ワークフロー実行の手順 diff --git a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx index ef05651b6..c2217ede4 100644 --- a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -51,9 +51,8 @@ Kling 2.6 Motion Control は、快手（Kuaishou）社が開発した専用の ## Kling 2.6 Motion Control のワークフロー - -

JSON形式のワークフローファイルをダウンロード

- + + ## 入力要件 diff --git a/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 951239804..5c3863523 100644 --- a/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -54,9 +54,11 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_text_to_video.mp4" > - -

JSON 形式のワークフロー ファイルをダウンロード

- + + + JSON 形式のワークフロー ファイルをダウンロード + + ### 2. ワークフロー実行手順 @@ -80,9 +82,11 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_image_to_video.mp4" > - -

JSON 形式のワークフロー ファイルをダウンロード

- + + + JSON 形式のワークフロー ファイルをダウンロード + + 以下の画像を入力画像としてダウンロードしてください。 @@ -112,9 +116,11 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video.mp4" > - -

JSON 形式のワークフロー ファイルをダウンロード

- + + + JSON 形式のワークフロー ファイルをダウンロード + + 以下の動画を入力動画としてダウンロードしてください： diff --git a/ja/tutorials/partner-nodes/openai/chat.mdx b/ja/tutorials/partner-nodes/openai/chat.mdx index 2d195caf5..12400d75f 100644 --- a/ja/tutorials/partner-nodes/openai/chat.mdx +++ b/ja/tutorials/partner-nodes/openai/chat.mdx @@ -22,13 +22,9 @@ OpenAI は生成AIに特化した企業であり、強力な対話機能を提 以下の JSON ファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式のワークフローファイルをダウンロード

- +| +| JSON 形式のワークフローファイルをダウンロード +| ### 2. ワークフロー実行手順（ステップごと） diff --git a/ja/tutorials/partner-nodes/rodin/model-generation.mdx b/ja/tutorials/partner-nodes/rodin/model-generation.mdx index 99704e747..17009c0bb 100644 --- a/ja/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ja/tutorials/partner-nodes/rodin/model-generation.mdx @@ -32,13 +32,9 @@ ComfyUI は現在、Rodin のモデル生成 API をネイティブ統合して 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式ワークフローファイルをダウンロード

- + + 単一視点モデル生成 (JSON形式) + 以下の画像を入力画像としてダウンロードしてください。 @@ -66,13 +62,9 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式ワークフローファイルをダウンロード

- + + 複数視点モデル生成 (JSON形式) + 以下の画像を入力画像としてダウンロードしてください。 diff --git a/ja/tutorials/partner-nodes/runway/video-generation.mdx b/ja/tutorials/partner-nodes/runway/video-generation.mdx index e85367f8e..5befee90e 100644 --- a/ja/tutorials/partner-nodes/runway/video-generation.mdx +++ b/ja/tutorials/partner-nodes/runway/video-generation.mdx @@ -37,9 +37,7 @@ Runway は、生成AIに特化した企業であり、強力な動画生成機 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen3a_turbo_image_to_video/runway_image_to_video_gen3a_turbo.mp4" > - -

JSON形式ワークフローファイルをダウンロード

- + 以下の画像を入力画像としてダウンロードしてください。 @@ -67,9 +65,7 @@ Runway は、生成AIに特化した企業であり、強力な動画生成機 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen4_turbo_image_to_video/runway_gen4_turo_image_to_video.mp4" > - -

JSON形式ワークフローファイルをダウンロード

- + 以下の画像を入力画像としてダウンロードしてください。 @@ -97,9 +93,7 @@ Runway は、生成AIに特化した企業であり、強力な動画生成機 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/first_last_frame_to_video/runway_first_last_frame.mp4" > - -

JSON形式ワークフローファイルをダウンロード

- + 以下の画像を入力画像としてダウンロードしてください。 diff --git a/ja/tutorials/partner-nodes/tripo/model-generation.mdx b/ja/tutorials/partner-nodes/tripo/model-generation.mdx index c9a3bb055..b71b44788 100644 --- a/ja/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ja/tutorials/partner-nodes/tripo/model-generation.mdx @@ -35,9 +35,11 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - -

JSON形式ワークフローファイルをダウンロード

- + + + テキストからモデル生成ワークフロー + + ### 2. ワークフロー実行手順（ステップ・バイ・ステップ） @@ -58,9 +60,11 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - -

JSON形式ワークフローファイルをダウンロード

- + + + 画像からモデル生成ワークフロー + + 以下の画像を入力画像としてダウンロードしてください： @@ -85,9 +89,11 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - -

JSON形式ワークフローファイルをダウンロード

- + + + 複数視点からのモデル生成ワークフロー + + 以下の画像を入力画像としてダウンロードしてください： diff --git a/ja/tutorials/utility/depth-anything-3.mdx b/ja/tutorials/utility/depth-anything-3.mdx index d49627577..36a32d873 100644 --- a/ja/tutorials/utility/depth-anything-3.mdx +++ b/ja/tutorials/utility/depth-anything-3.mdx @@ -39,10 +39,10 @@ ComfyUI は Depth Anything 3 ノードをネイティブサポートしていま Depth Anything 3 チェックポイントをダウンロードし、対応する ComfyUI フォルダに保存します： -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — 軽量で高速な推論 -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — バランスの取れた性能 -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 単眼深度に最適（空検出対応） -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — メートル単位の物理深度（空検出対応） +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — 軽量で高速な推論 +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — バランスの取れた性能 +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 単眼深度に最適（空検出対応） +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — メートル単位の物理深度（空検出対応） ``` ComfyUI/ diff --git a/ja/tutorials/utility/face-detection/mediapipe.mdx b/ja/tutorials/utility/face-detection/mediapipe.mdx index 0044222f6..0743c3501 100644 --- a/ja/tutorials/utility/face-detection/mediapipe.mdx +++ b/ja/tutorials/utility/face-detection/mediapipe.mdx @@ -55,7 +55,7 @@ ComfyUI を最新バージョンにアップデートし、`Workflow` → `Brows MediaPipe Face Detection モデルは [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe) でホストされています。 -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) 以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/utility/moge.mdx b/ja/tutorials/utility/moge.mdx index e3e7397fa..075dc59f1 100644 --- a/ja/tutorials/utility/moge.mdx +++ b/ja/tutorials/utility/moge.mdx @@ -50,8 +50,8 @@ ComfyUI は MoGe ノードをネイティブサポートしています。始め MoGe チェックポイントをダウンロードし、ComfyUI の該当フォルダに保存します： -- **MoGe-2（推奨）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1（ベースライン）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2（推奨）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1（ベースライン）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/utility/pose-detection-sdpose.mdx b/ja/tutorials/utility/pose-detection-sdpose.mdx index a63be80a2..ba69183f9 100644 --- a/ja/tutorials/utility/pose-detection-sdpose.mdx +++ b/ja/tutorials/utility/pose-detection-sdpose.mdx @@ -87,11 +87,11 @@ ComfyUIを最新バージョンにアップデートし、`Workflow` → `Browse SDPoseとRT-DETRv4のモデルチェックポイントは、[Comfy-Org SDPose モデルリポジトリ](https://huggingface.co/Comfy-Org/SDPose) で公開されています。 **checkpoints** (SDPoseモデル): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models** (RT-DETRv4検出器): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (推奨) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (完全精度、サイズ大) +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (推奨) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (完全精度、サイズ大) 以下のディレクトリ構成に配置してください: diff --git a/ja/tutorials/utility/remove-background-birefnet.mdx b/ja/tutorials/utility/remove-background-birefnet.mdx index c0f4cc3ff..7f1295b3a 100644 --- a/ja/tutorials/utility/remove-background-birefnet.mdx +++ b/ja/tutorials/utility/remove-background-birefnet.mdx @@ -48,7 +48,7 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` BiRefNet モデルは [Comfy-Org BiRefNet モデルリポジトリ](https://huggingface.co/Comfy-Org/BiRefNet) でホストされています。 -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) 以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/utility/video-segment-sam3.mdx b/ja/tutorials/utility/video-segment-sam3.mdx index 6ac27d505..3295028ed 100644 --- a/ja/tutorials/utility/video-segment-sam3.mdx +++ b/ja/tutorials/utility/video-segment-sam3.mdx @@ -68,7 +68,7 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` SAM 3.1 モデルは [Comfy-Org SAM 3.1 モデルリポジトリ](https://huggingface.co/Comfy-Org/sam3.1) でホストされています。 -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) 以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/utility/void-video-inpainting.mdx b/ja/tutorials/utility/void-video-inpainting.mdx index 57a3967e5..a512579a3 100644 --- a/ja/tutorials/utility/void-video-inpainting.mdx +++ b/ja/tutorials/utility/void-video-inpainting.mdx @@ -65,24 +65,24 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` **拡散モデル** — 中核となる2パス修復モデル： -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 精錬パス、時間的安定性に優れる -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 一次パス +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — 精錬パス、時間的安定性に優れる +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — 一次パス **VAE：** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) **オプティカルフロー：** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 セグメンテーション：** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **テキストエンコーダ：** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ diff --git a/ja/tutorials/video/bytedance/bernini-r.mdx b/ja/tutorials/video/bytedance/bernini-r.mdx index 13d6b0383..6485b68d0 100644 --- a/ja/tutorials/video/bytedance/bernini-r.mdx +++ b/ja/tutorials/video/bytedance/bernini-r.mdx @@ -49,16 +49,16 @@ ComfyUI は Bernini-R ノードをネイティブサポートしています。 必要なモデルウェイトをダウンロードし、対応する ComfyUI フォルダに保存します： **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 5871cea65..32d050845 100644 --- a/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -14,17 +14,27 @@ Cosmos-Predict2 は、NVIDIA によって開発された次世代の物理世界 Cosmos-Predict2 は、テキストから画像（Text2Image）や動画から世界へ（Video2World）など、さまざまな生成方法をサポートしており、産業シミュレーション、自動運転、都市計画、科学研究などの分野で広く使用されています。 これは、インテリジェントビジョンと物理世界の深い統合を促進するための重要な基礎ツールです。 -GitHub:[Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) -huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) + + + Cosmos-Predict2 ソースコードとドキュメント + + + Cosmos-Predict2 モデルコレクション + + このガイドでは、ComfyUI での **Video2World** 生成の完了までの手順を説明します。 テキストから画像のセクションについては、以下の部分を参照してください。 - - Cosmos-Predict2 を使用したテキストから画像の生成 - -{/* + + + Cosmos-Predict2 を使用したテキストから画像の生成 + + + 強力な GPU で Comfy Cloud 上で Cosmos-Predict2 ワークフローを実行 + + ## Cosmos Predict2 Video2World ワークフロー @@ -41,9 +51,14 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - -

Json 形式ワークフローファイルをダウンロード

- + + + JSON 形式ワークフローファイルをダウンロード + + + Comfy Cloud でこのワークフローを実行（モデルプリインストール済み） + + 入力として以下の画像をダウンロードしてください。 @@ -55,17 +70,23 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **Diffusion model** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) + + cosmos_predict2_2B_video2world_480p_16fps.safetensors + 他の重みについては、[Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) にアクセスしてダウンロードしてください **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) + + oldt5_xxl_fp8_e4m3fn_scaled.safetensors + **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + wan_2.1_vae.safetensors + ファイル保存場所 ``` @@ -92,4 +113,4 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- 6. (オプション) `ClipTextEncode` ノードでプロンプトを変更できます 7. (オプション) `CosmosPredict2ImageToVideoLatent` ノードでサイズとフレーム数を変更してください 8. `Run` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行してください -9. 生成が完了すると、動画は自動的に `ComfyUI/output/` ディレクトリに保存されます。`save video` ノードでプレビューすることもできます */} \ No newline at end of file +9. 生成が完了すると、動画は自動的に `ComfyUI/output/` ディレクトリに保存されます。`save video` ノードでプレビューすることもできます \ No newline at end of file diff --git a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 5ca22c6eb..7c6b936e5 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -29,17 +29,17 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## モデルリンク **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **diffusion_models** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **vae** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) モデルの保存場所 diff --git a/ja/tutorials/video/hunyuan/hunyuan-video.mdx b/ja/tutorials/video/hunyuan/hunyuan-video.mdx index 61bee272e..77788802f 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video.mdx @@ -46,9 +46,9 @@ Hunyuan Video シリーズは [Tencent](https://huggingface.co/tencent) によ 以下のモデルは、テキストから動画および画像から動画の両方のワークフローで使用されます。ダウンロードして、指定されたディレクトリに保存してください： -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors?download=true) +- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) 保存場所： @@ -73,7 +73,7 @@ Hunyuan Text-to-Video は 2024 年 12 月にオープンソース化され、中 ### 2. モデルの手動インストール -[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models` フォルダに保存してください。 +[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models` フォルダに保存してください。 これらのモデルファイルがすべて正しい場所に存在することを確認してください： @@ -126,7 +126,7 @@ Hunyuan Image-to-Video モデルは 2025 年 3 月 6 日にオープンソース ### v1 および v2 バージョンで共通のモデル 以下のファイルをダウンロードし、`ComfyUI/models/clip_vision` ディレクトリに保存してください： -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) +- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) ### V1"concat"画像から動画ワークフロー @@ -140,7 +140,7 @@ Hunyuan Image-to-Video モデルは 2025 年 3 月 6 日にオープンソース #### 2. 関連モデルの手動インストール -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) これらのモデルファイルがすべて正しい場所に存在することを確認してください： @@ -185,7 +185,7 @@ v2 ワークフローは本質的に v1 ワークフローと同じです。**re #### 2. 関連モデルの手動インストール -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) これらのモデルファイルがすべて正しい場所に存在することを確認してください： diff --git a/ja/tutorials/video/kandinsky/kandinsky-5.mdx b/ja/tutorials/video/kandinsky/kandinsky-5.mdx index bb14be791..a6eb738f9 100644 --- a/ja/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/ja/tutorials/video/kandinsky/kandinsky-5.mdx @@ -51,21 +51,39 @@ Kandinsky 5.0 は、Flow Matching を備えた潜在拡散パイプラインを ComfyUI を最新バージョンに更新し、メニュー `ワークフロー` -> `テンプレートを表示` -> `動画` から "Kandinsky 5.0 T2V" を見つけてワークフローを読み込んでください。 - -

JSON ワークフローファイルをダウンロード

- + + + T2V ワークフローをダウンロードしてローカルで使用 + + + Comfy Cloud で開く + + ### 2. モデルの手動ダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B テキストエンコーダー (FP8) + + + CLIP-L テキストエンコーダー + + **拡散モデル** -- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) + + + Kandinsky 5.0 T2V Lite SFT 拡散モデル (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ @@ -85,21 +103,39 @@ ComfyUI/ ComfyUI を最新バージョンに更新し、メニュー `ワークフロー` -> `テンプレートを表示` -> `動画` から "Kandinsky 5.0 I2V" を見つけてワークフローを読み込んでください。 - -

JSON ワークフローファイルをダウンロード

- + + + I2V ワークフローをダウンロードしてローカルで使用 + + + Comfy Cloud で開く + + ### 2. モデルの手動ダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B テキストエンコーダー (FP8) + + + CLIP-L テキストエンコーダー + + **拡散モデル** -- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) + + + Kandinsky 5.0 I2V Lite 拡散モデル (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ diff --git a/ja/tutorials/video/ltxv.mdx b/ja/tutorials/video/ltxv.mdx index 8d047bc2e..74c84afe6 100644 --- a/ja/tutorials/video/ltxv.mdx +++ b/ja/tutorials/video/ltxv.mdx @@ -26,9 +26,17 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 最初の [フレーム画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png) を使用して動画を制御できます。 - -

Comfy Cloud で実行

- + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "LTX-Video" を検索 + + + このワークフローのサンプル入力画像を入手 + + LTX-Video 画像から動画ワークフロー @@ -38,6 +46,15 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## テキストから動画 + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "LTX-Video" を検索 + + + LTX-Video テキストから動画ワークフロー @@ -48,8 +65,13 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 以下のモデルをダウンロードし、下記に指定された場所に配置してください： -- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true) + + ダウンロードして ComfyUI/models/checkpoints/ に配置 + + + + ダウンロードして ComfyUI/models/text_encoders/ に配置 + ``` ├── checkpoints/ diff --git a/ja/tutorials/video/wan/fun-camera.mdx b/ja/tutorials/video/wan/fun-camera.mdx index adfb6689a..d4c4eecaa 100644 --- a/ja/tutorials/video/wan/fun-camera.mdx +++ b/ja/tutorials/video/wan/fun-camera.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Performance Reference": 32425486 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Wan2.1 Fun Camera について @@ -36,21 +35,49 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下に示すすべてのモデルは、[Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) から入手できます。 -**Diffusion モデル**（1.3B または 14B のいずれかを選択）: -- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) -- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) +### Diffusion モデル + +1.3B または 14B のいずれかを選択: + + + + Wan2.1 Fun Camera 1.3B 拡散モデル + + + Wan2.1 Fun Camera 14B 拡散モデル + + 以前に Wan2.1 関連のモデルをご利用になったことがある場合、以下のモデルは既にご所有である可能性があります。万が一不足している場合は、それぞれダウンロードしてください。 -**Text Encoders**（いずれか1つを選択）: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +### Text Encoders + +いずれか1つを選択: + + + + フル精度テキストエンコーダ + + + FP8 量子化テキストエンコーダ（低 VRAM 推奨） + + + +### VAE -**VAE**: -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + + Wan2.1 VAE モデル + + -**CLIP Vision**: -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +### CLIP Vision + + + + CLIP ビジョンエンコーダ + + ファイルの保存場所: @@ -70,9 +97,16 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 1.3B ネイティブワークフロー例 -### 1. ワークフロー関連ファイルのダウンロード +### 1. ワークフローのダウンロード -#### 1.1 ワークフローファイル + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "Wan 2.1 Fun Camera 1.3B" を検索 + + 以下の動画をダウンロードし、ComfyUI にドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます： @@ -82,21 +116,19 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > - -

JSON ワークフローファイルをダウンロード

- - 14B バージョンをご利用になりたい場合は、単にモデルファイルを 14B バージョンに置き換えてください。ただし、VRAM の要件にご注意ください。 -#### 1.2 入力画像のダウンロード - -以下の画像をダウンロードし、これを開始フレームとして使用します： +### 2. 入力素材のダウンロード -![入力参照画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) + + + 以下の画像をダウンロードし、1.3B ワークフローの開始フレームとして使用します + + -### 2. ワークフローをステップ・バイ・ステップで完了させる +### 3. ワークフローをステップ・バイ・ステップで完了させる ![Wan2.1 Fun Camera ワークフロー手順](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -113,18 +145,30 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 14B ワークフローおよび入力画像 +### 1. ワークフローのダウンロード + + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "Wan 2.1 Fun Camera 14B" を検索 + + + - -

JSON ワークフローファイルをダウンロード

- +### 2. 入力素材のダウンロード -**入力画像** -![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) + + + 以下の画像をダウンロードし、14B ワークフローの開始フレームとして使用します + + ## パフォーマンスの参考値 diff --git a/ja/tutorials/video/wan/fun-control.mdx b/ja/tutorials/video/wan/fun-control.mdx index c1001c086..fe5402c18 100644 --- a/ja/tutorials/video/wan/fun-control.mdx +++ b/ja/tutorials/video/wan/fun-control.mdx @@ -55,18 +55,18 @@ ComfyUI は現在、Wan2.1 Fun Control モデルを**ネイティブサポート 対応するリンクをクリックしてダウンロードしてください。以前に Wan 関連のワークフローを使用したことがある場合は、**Diffusion models** のみをダウンロードする必要があります。 **Diffusion models** - 1.3B または 14B を選択。14B バージョンはファイルサイズが大きく（32GB）、VRAM 要件も高くなります： -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true)：ダウンロード後に `Wan2.1-Fun-14B-Control.safetensors` にリネームしてください **Text encoders** - 以下のモデルのいずれかを選択（fp16 精度はサイズが大きく、パフォーマンス要件が高くなります）： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存場所： ``` diff --git a/ja/tutorials/video/wan/fun-inp.mdx b/ja/tutorials/video/wan/fun-inp.mdx index ccae9db7b..181f518ce 100644 --- a/ja/tutorials/video/wan/fun-inp.mdx +++ b/ja/tutorials/video/wan/fun-inp.mdx @@ -53,18 +53,18 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 以下のモデルは [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) と [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) で見つかります。 **Diffusion models** - 1.3B または 14B を選択してください。14B バージョンはファイルサイズが大きく (32GB)、VRAM 要件も高くなります： -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): ダウンロード後、`Wan2.1-Fun-14B-InP.safetensors` にリネームしてください +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true): ダウンロード後、`Wan2.1-Fun-14B-InP.safetensors` にリネームしてください **Text encoders** - 以下のモデルのいずれかを選択してください（fp16 精度はサイズが大きく、パフォーマンス要件も高くなります）： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存場所： ``` diff --git a/ja/tutorials/video/wan/vace.mdx b/ja/tutorials/video/wan/vace.mdx index 67687320f..e7127c367 100644 --- a/ja/tutorials/video/wan/vace.mdx +++ b/ja/tutorials/video/wan/vace.mdx @@ -61,18 +61,18 @@ VACE 14B は、アリババ Tongyi Wanxiang チームが公開したオープン ### モデルのダウンロード **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 以前に Wan Video 関連のワークフローをご利用になったことがある場合、以下のモデルファイルはすでにダウンロード済みです。 **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **Text encoders** からいずれか 1 つのバージョンを選択してダウンロードしてください： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) ファイルの保存先： ``` diff --git a/ja/tutorials/video/wan/wan-ati.mdx b/ja/tutorials/video/wan/wan-ati.mdx index d239223b3..87176883f 100644 --- a/ja/tutorials/video/wan/wan-ati.mdx +++ b/ja/tutorials/video/wan/wan-ati.mdx @@ -46,17 +46,17 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ワークフローからモデルファイルを正常にダウンロードできていない場合、以下のリンクから手動でダウンロードしてみてください。 **Diffusionモデル** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **テキストエンコーダー**（以下のいずれか1つを選択） -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) ファイル保存先 ``` diff --git a/ja/tutorials/video/wan/wan-causal-forcing.mdx b/ja/tutorials/video/wan/wan-causal-forcing.mdx index 4b1420be4..66fbf617d 100644 --- a/ja/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ja/tutorials/video/wan/wan-causal-forcing.mdx @@ -83,10 +83,10 @@ Wan2.1 I2V モデルと必要なファイルをダウンロードします。対 ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B チェックポイント - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B チェックポイント（最小 8GB VRAM） @@ -94,10 +94,10 @@ Wan2.1 I2V モデルと必要なファイルをダウンロードします。対 ### CLIP と VAE - + google-bert/bert-base-uncased — CLIP テキストエンコーダー - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/ja/tutorials/video/wan/wan-dancer.mdx b/ja/tutorials/video/wan/wan-dancer.mdx index 6ea08b40f..8a6fb2fdc 100644 --- a/ja/tutorials/video/wan/wan-dancer.mdx +++ b/ja/tutorials/video/wan/wan-dancer.mdx @@ -61,20 +61,20 @@ ComfyUI を最新バージョンに更新し、ワークフローファイルを ### 3. 手動でのモデルのダウンロード **拡散モデル** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **テキストエンコーダ** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP ビジョン** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan-flf.mdx b/ja/tutorials/video/wan/wan-flf.mdx index 7e4a45626..184c61a11 100644 --- a/ja/tutorials/video/wan/wan-flf.mdx +++ b/ja/tutorials/video/wan/wan-flf.mdx @@ -56,7 +56,7 @@ Wan FLF2V（First-Last Frame Video Generation：始終フレーム動画生成 本ガイドで使用するすべてのモデルは、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)から入手できます。 **diffusion_models**：ご使用のハードウェア環境に応じて、以下のいずれかのバージョンを選択してください。 -- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8: [wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -64,14 +64,14 @@ Wan FLF2V（First-Last Frame Video Generation：始終フレーム動画生成 **Text encoders**：以下のいずれか1つのバージョンをダウンロードしてください。 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存先 diff --git a/ja/tutorials/video/wan/wan-move.mdx b/ja/tutorials/video/wan/wan-move.mdx index 02d87136c..8b950e59d 100644 --- a/ja/tutorials/video/wan/wan-move.mdx +++ b/ja/tutorials/video/wan/wan-move.mdx @@ -28,37 +28,37 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Wan-Move 画像から動画へのワークフロー - -

JSON ワークフローファイルをダウンロード

- + + ワークフローをダウンロード + - -

ComfyUI Cloud で実行

- + + クラウドで開く + ## モデルリンク -**text_encoders** + + **text_encoders** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors + -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + **clip_vision** -- clip_vision_h.safetensors + -**clip_vision** + + **loras** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors + -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) + + **diffusion_models** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors + -**loras** - -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) - -**diffusion_models** - -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) - -**vae** - -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + **vae** -- wan_2.1_vae.safetensors + **モデル保存場所** diff --git a/ja/tutorials/video/wan/wan-video.mdx b/ja/tutorials/video/wan/wan-video.mdx index 73df834b7..a4cc007ed 100644 --- a/ja/tutorials/video/wan/wan-video.mdx +++ b/ja/tutorials/video/wan/wan-video.mdx @@ -41,14 +41,14 @@ Wan2.1 Video シリーズは、アリババ社が 2025 年 2 月に [Apache 2.0 このガイドで言及されるすべてのモデルは、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files) から入手できます。以下は、このガイドのサンプルで使用する共通のモデルであり、事前にダウンロードしておくことを推奨します： **Text encoders** からいずれか 1 つのバージョンを選択してダウンロードしてください： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイルの保存先ディレクトリ構成： ``` @@ -70,7 +70,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル ## Wan2.1 テキスト→動画（T2V）ワークフロー -ワークフローを開始する前に、[wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +ワークフローを開始する前に、[wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 > 他の T2V 精度バージョンが必要な場合は、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) からダウンロードしてください。 @@ -107,7 +107,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル ![Wan2.1 画像→動画ワークフロー（14B、480P）の入力画像サンプル](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) #### 2. モデルのダウンロード -[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 #### 3. ワークフローをステップごとに実行 @@ -135,7 +135,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル #### 2. モデルのダウンロード -[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 #### 3. ワークフローをステップごとに実行 diff --git a/ja/tutorials/video/wan/wan2-2-animate.mdx b/ja/tutorials/video/wan/wan2-2-animate.mdx index 314d5e51b..410410b6e 100644 --- a/ja/tutorials/video/wan/wan2-2-animate.mdx +++ b/ja/tutorials/video/wan/wan2-2-animate.mdx @@ -55,13 +55,14 @@ Wan-Animate は、WAN チームが開発した人物アニメーションおよ 以下のワークフローファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして読み込んでください。 - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Comfy Cloud で実行 + + + JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Animate" を検索 + + 以下の素材を入力としてダウンロードしてください： @@ -77,21 +78,40 @@ Wan-Animate は、WAN チームが開発した人物アニメーションおよ ### 2. モデルのダウンロードリンク -**diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) — Kijai のリポジトリから提供されるモデル -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) — 元のモデル重み +**Diffusion Models** + + + + Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: Kijai のリポジトリから提供されるスケーリング FP8 モデル + + + wan2.2_animate_14B_bf16.safetensors: 元の bf16 モデル重み + + + +**CLIP Vision** + + + clip_vision_h.safetensors: CLIP Vision エンコーダ + + +**LoRAs** + + + lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4 ステップ高速化 LoRA + -**clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +**VAE** -**loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) — 4ステップ高速化対応の LoRA + + wan_2.1_vae.safetensors: エンコードおよびデコード用の Wan2.1 VAE + -**vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +**Text Encoders** -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: スケーリング FP8 テキストエンコーダ + ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-fun-camera.mdx b/ja/tutorials/video/wan/wan2-2-fun-camera.mdx index 2636c306a..8ba4abf19 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -46,31 +46,60 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

JSON ワークフローをダウンロード

- + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Fun Camera" を検索 + + 以下の画像をダウンロードし、入力として使用します。 -![入力開始画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) + + + 動画生成の開始フレーム。この画像をダウンロードして使用するか、ご自身の画像に置き換えてください。 + + ### 2. モデルのダウンロードリンク 以下のモデルは、[Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 **Diffusion モデル** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) + + + + Wan2.2 Fun Camera 用高ノイズ拡散モデル + + + Wan2.2 Fun Camera 用低ノイズ拡散モデル + + **Wan2.2-Lightning LoRA（オプション：高速化用）** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + 高ノイズモデル用 4ステップ加速 LoRA + + + 低ノイズモデル用 4ステップ加速 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**テキストエンコーダー** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + エンコード/デコード用 Wan2.1 VAE + + +**テキストエンコーダー** + + + FP8 スケーリング版テキストエンコーダー + ファイル保存先のディレクトリ構成： @@ -112,4 +141,4 @@ ComfyUI/ - **Width/Height（幅／高さ）**：動画の解像度を設定 - **Length（長さ）**：動画のフレーム数を設定（デフォルトは81フレーム） - **Speed（速度）**：動画の再生速度を設定（デフォルトは1.0） -8. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合は Cmd）+ Enter` を押して動画生成を実行してください。 \ No newline at end of file +8. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合は Cmd）+ Enter` を押して動画生成を実行してください。 diff --git a/ja/tutorials/video/wan/wan2-2-fun-control.mdx b/ja/tutorials/video/wan/wan2-2-fun-control.mdx index 87fd9ec0c..9bf54688c 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-control.mdx @@ -55,27 +55,27 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### 1. ワークフローと素材のダウンロード -以下の動画または JSON ファイルをダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください +ComfyUI を最新バージョンに更新し、ワークフローファイルをダウンロードして ComfyUI にドラッグするか、テンプレートライブラリの `Workflow` → `Browse Templates` → `Video` から "Wan2.2 Fun Control" を見つけてください。 - - - -

JSON ワークフローをダウンロード

- + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Fun Control" を検索 + + 入力素材として以下の画像および動画をダウンロードしてください。 -![入力開始画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) - - + + + 動画生成の開始フレームです。この画像をダウンロードして使用するか、ご自身の画像に置き換えてください。 + + + 前処理済みのポーズ制御動画です。この動画をダウンロードして使用するか、ご自身の動画に置き換えてください。 + + > ここでは前処理済みの動画を使用しています。 @@ -83,19 +83,39 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下のモデルは [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) で見つかります -**Diffusion Model** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) +**Diffusion Models** + + + + wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors：高ノイズ拡散モデル + + + wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors：低ノイズ拡散モデル + + -**Wan2.2-Lightning LoRA (オプション、加速用)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +**Wan2.2-Lightning LoRA（オプション、加速用）** + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors：高ノイズ 4 ステップ加速 LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors：低ノイズ 4 ステップ加速 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors：エンコード/デコード用 Wan2.1 VAE + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors：スケーリング FP8 テキストエンコーダ + ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx index b00d7fa5d..8fb36d964 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -61,13 +61,14 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem または、ComfyUI を最新版に更新した上で、以下のリンクからワークフローファイルをダウンロードし、ComfyUI の画面にドラッグ＆ドロップして読み込んでください。 - -

JSON形式ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Fun Inp" を検索してください + + + Comfy Cloud で開く + + 以下の画像を開始フレームおよび終了フレームの素材としてご使用ください。 @@ -77,18 +78,38 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem ### 2. モデルの準備 **Diffusion モデル** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) + + + + wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: 首尾フレーム修復用の高ノイズ拡散モデル + + + wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: 首尾フレーム修復用の低ノイズ拡散モデル + + **Lightning LoRA（任意：高速化用）** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 高ノイズモデル用の 4 ステップ高速化 LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 低ノイズモデル用の 4 ステップ高速化 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**テキストエンコーダー** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors: エンコードおよびデコード用の Wan2.1 VAE + + +**テキストエンコーダー** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: スケーリングされた FP8 テキストエンコーダー + ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-s2v.mdx b/ja/tutorials/video/wan/wan2-2-s2v.mdx index 872e49120..179e2aa60 100644 --- a/ja/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ja/tutorials/video/wan/wan2-2-s2v.mdx @@ -35,38 +35,58 @@ Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 S2V" を検索 + + 以下の画像および音声ファイルを入力としてダウンロードしてください: -![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - - -

入力音声をダウンロード

- + + + デフォルトの入力画像をダウンロードするか、ご自身の画像をお使いください。 + + + デフォルトの入力音声をダウンロードするか、ご自身の音声をお使いください。 + + ### 2. モデルのダウンロードリンク すべてのモデルは、[当社のリポジトリ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 -**diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +**diffusion_models** + + + + FP8 scaled diffusion model。ComfyUI/models/diffusion_models/ に配置 + + + BF16 diffusion model。ComfyUI/models/diffusion_models/ に配置 + + + +**audio_encoders** + + + Audio encoder model。ComfyUI/models/audio_encoders/ に配置 + + +**vae** -**audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) + + Wan2.1 VAE model。ComfyUI/models/vae/ に配置 + -**vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +**text_encoders** -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + FP8 scaled text encoder。ComfyUI/models/text_encoders/ に配置 + ``` @@ -94,8 +114,14 @@ ComfyUI/ 両方のモデルは、[こちらのページ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) から入手可能です： -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) + + + FP8 scaled diffusion model + + + BF16 diffusion model + + 本テンプレートでは `wan2.2_s2v_14B_fp8_scaled.safetensors` を使用しており、VRAM 使用量が少ないのが特徴です。ただし、品質劣化を抑えるために `wan2.2_s2v_14B_bf16.safetensors` を試すことも可能です。 diff --git a/ja/tutorials/video/wan/wan2_2.mdx b/ja/tutorials/video/wan/wan2_2.mdx index 7805a5892..2d439d858 100644 --- a/ja/tutorials/video/wan/wan2_2.mdx +++ b/ja/tutorials/video/wan/wan2_2.mdx @@ -102,24 +102,25 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - -

JSON ワークフローファイルをダウンロード

- - - -

Run on Comfy Cloud

- + + + JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 5B」を検索 + + + Comfy Cloud で開く + + ### 2. モデルの手動ダウンロード **Diffusion Model** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) +- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) +- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan2.2_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -157,25 +158,26 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - -

JSON ワークフローファイルをダウンロード

- - - -

Run on Comfy Cloud

- + + + JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 14B T2V」を検索 + + + Comfy Cloud で開く + + ### 2. モデルの手動ダウンロード **Diffusion Model** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -214,13 +216,14 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - -

JSON ワークフローファイルをダウンロード

- - - -

Run on Comfy Cloud

- + + + JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 14B I2V」を検索 + + + Comfy Cloud で開く + + 以下の画像を入力として使用できます： ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -228,14 +231,14 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows ### 2. モデルの手動ダウンロード **Diffusion Model** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) +- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -273,13 +276,14 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - -

JSON ワークフローをダウンロード

- - - -

Run on Comfy Cloud

- + + + JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 14B FLF2V」を検索 + + + Comfy Cloud で開く + + 以下の画像を入力素材としてダウンロードしてください： diff --git a/ja/tutorials/video/zai/scail2.mdx b/ja/tutorials/video/zai/scail2.mdx index 0e91e1098..0ad537457 100644 --- a/ja/tutorials/video/zai/scail2.mdx +++ b/ja/tutorials/video/zai/scail2.mdx @@ -105,23 +105,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### 必要なモデル **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders**（いずれか） -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### ファイル保存場所 diff --git a/ko/tutorials/3d/hunyuan3D-2.mdx b/ko/tutorials/3d/hunyuan3D-2.mdx index a91bcb675..1841d79c9 100644 --- a/ko/tutorials/3d/hunyuan3D-2.mdx +++ b/ko/tutorials/3d/hunyuan3D-2.mdx @@ -57,9 +57,14 @@ Hunyuan3D-2mv 워크플로우에서는 다중뷰 이미지를 사용해 3D 모 - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 이 워크플로우를 즉시 실행 + + + 워크플로우 JSON 파일 다운로드 + + ### 1. 워크플로우 @@ -81,7 +86,7 @@ Hunyuan3D-2mv 워크플로우에서는 다중뷰 이미지를 사용해 3D 모 아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv.safetensors`로 이름을 변경할 수 있습니다. ``` ComfyUI/ @@ -104,9 +109,14 @@ ComfyUI/ Hunyuan3D-2mv-turbo 워크플로우에서는 Hunyuan3D-2mv-turbo 모델을 사용해 3D 모델을 생성합니다. 이 모델은 Hunyuan3D-2mv의 단계 증류 버전으로, 더 빠른 3D 모델 생성을 가능하게 합니다. 이번 버전의 워크플로우에서는 `cfg`를 1.0으로 설정하고, `flux guidance` 노드를 추가해 `증류된 cfg` 생성을 제어합니다. - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 이 워크플로우를 즉시 실행 + + + 워크플로우 JSON 파일 다운로드 + + ### 1. 워크플로우 @@ -125,7 +135,7 @@ Hunyuan3D-2mv-turbo 워크플로우에서는 Hunyuan3D-2mv-turbo 모델을 사 아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv-turbo.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv-turbo.safetensors`로 이름을 변경할 수 있습니다. ``` ComfyUI/ @@ -146,9 +156,14 @@ ComfyUI/ Hunyuan3D-2 워크플로우에서는 Hunyuan3D-2 모델을 사용해 3D 모델을 생성합니다. 이 모델은 다중뷰 모델이 아닙니다. 이번 워크플로우에서는 `Hunyuan3Dv2ConditioningMultiView` 노드 대신 `Hunyuan3Dv2Conditioning` 노드를 사용합니다. - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 이 워크플로우를 즉시 실행 + + + 워크플로우 JSON 파일 다운로드 + + ### 1. 워크플로우 @@ -163,7 +178,7 @@ Hunyuan3D-2 워크플로우에서는 Hunyuan3D-2 모델을 사용해 3D 모델 아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2.safetensors`로 이름을 변경할 수 있습니다. ``` ComfyUI/ diff --git a/ko/tutorials/3d/triposplat.mdx b/ko/tutorials/3d/triposplat.mdx index 18e78fdb3..d8b14a750 100644 --- a/ko/tutorials/3d/triposplat.mdx +++ b/ko/tutorials/3d/triposplat.mdx @@ -106,23 +106,23 @@ TripoSplat은 단일 RGB 이미지를 입력으로 받아 3D 가우시안 프리 TripoSplat 모델과 필요한 파일을 다운로드하세요. 해당 `models/` 하위 디렉토리에 배치하세요. - + triposplat_fp16.safetensors — TripoSplat 확산 모델 체크포인트 - + triposplat_vae_decoder_fp16.safetensors — VAE 디코더 - + flux2-vae.safetensors — Flux.2 VAE, 잠재적 인코딩용 - + dino_v3_vit_h.safetensors — CLIP 비전 인코더 (DINOv2) - + birefnet.safetensors — 전처리용 배경 제거 모델 diff --git a/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx index 4dfcd158a..d0fd07c1d 100644 --- a/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -45,7 +45,7 @@ AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공 ### AIO 모델 다운로드 - + 올인원 체크포인트 파일 (대부분의 사용자에게 권장). @@ -74,19 +74,19 @@ AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공 ### 분할 모델 다운로드 - + 디퓨전 모델. - + 텍스트 인코더 (0.6B). - + 텍스트 인코더 (1.7B). - + VAE 모델. diff --git a/ko/tutorials/audio/ace-step/ace-step-v1.mdx b/ko/tutorials/audio/ace-step/ace-step-v1.mdx index 9e80a4df1..b8646e8e4 100644 --- a/ko/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ko/tutorials/audio/ace-step/ace-step-v1.mdx @@ -33,9 +33,15 @@ ACE-Step은 중국 팀 StepFun과 ACE Studio가 공동 개발한 오픈소스 아래 버튼을 클릭해 해당 워크플로 파일을 다운로드하세요. 이를 ComfyUI로 드래그해 워크플로 정보를 로드하세요. 워크플로에는 모델 다운로드 정보도 포함되어 있습니다. - -

Json 형식 워크플로 파일 다운로드

- + + + Json 형식 워크플로 파일 다운로드 + + 또는 [ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors)를 수동으로 다운로드해 `ComfyUI/models/checkpoints` 폴더에 저장할 수도 있습니다. @@ -58,15 +64,27 @@ ACE-Step은 중국 팀 StepFun과 ACE Studio가 공동 개발한 오픈소스 아래 버튼을 클릭해 해당 워크플로우 파일을 다운로드하세요. 이를 ComfyUI로 드래그해 워크플로우 정보를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + Json 형식 워크플로우 파일 다운로드 + + 다음 오디오 파일을 입력 오디오로 다운로드하세요: - -

입력용 예시 오디오 파일 다운로드

- + + + 입력용 예시 오디오 파일 다운로드 + + ### 2. 워크플로우 단계별 완료하기 diff --git a/ko/tutorials/audio/stable-audio/stable-audio-1.mdx b/ko/tutorials/audio/stable-audio/stable-audio-1.mdx index a732bad3d..b27712c45 100644 --- a/ko/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ko/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -41,7 +41,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### 체크포인트 - + 2.3GB. models/checkpoints/ 폴더에 배치하세요. @@ -56,7 +56,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### 텍스트 인코더 - + 프롬프트 조건부 설정을 위한 텍스트 인코더. models/text_encoders/ 폴더에 배치하세요. diff --git a/ko/tutorials/audio/stable-audio/stable-audio-3.mdx b/ko/tutorials/audio/stable-audio/stable-audio-3.mdx index bc5d2baa4..fd318f6f1 100644 --- a/ko/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ko/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -77,11 +77,11 @@ Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화 ### 체크포인트 - + Medium 워크플로우용. models/checkpoints/에 배치하세요 - + Medium Base 워크플로우용. models/checkpoints/에 배치하세요 @@ -97,11 +97,11 @@ Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화 ### 텍스트 인코더 - + 모든 Stable Audio 3 워크플로우에 필수. models/text_encoders/에 배치하세요 - + Medium 워크플로우용 (Qwen 리프롬프트). models/text_encoders/에 배치하세요 diff --git a/ko/tutorials/basic/inpaint.mdx b/ko/tutorials/basic/inpaint.mdx index d542c9610..51c611b17 100644 --- a/ko/tutorials/basic/inpaint.mdx +++ b/ko/tutorials/basic/inpaint.mdx @@ -34,7 +34,7 @@ AI 이미지 생성 과정에서 우리는 종종 전체적인 이미지는 만 #### 1. 모델 설치 -[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) 파일을 다운로드하여 `ComfyUI/models/checkpoints` 폴더에 넣으세요: +[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) 파일을 다운로드하여 `ComfyUI/models/checkpoints` 폴더에 넣으세요: #### 2. 인페인팅 자산 @@ -68,7 +68,7 @@ AI 이미지 생성 과정에서 우리는 종종 전체적인 이미지는 만 ![SD1.5 인페인팅 결과](/images/tutorial/basic/inpaint/inpaint_sd1.5_pruned_emaonly.png) -[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) 모델로 생성한 결과가 더 나은 인페인팅 효과와 자연스러운 전환을 보여줍니다. 이는 해당 모델이 인페인팅에 특화되어 있어 생성 영역을 더 잘 제어할 수 있고, 결과적으로 인페인팅 효과가 개선되기 때문입니다. +[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) 모델로 생성한 결과가 더 나은 인페인팅 효과와 자연스러운 전환을 보여줍니다. 이는 해당 모델이 인페인팅에 특화되어 있어 생성 영역을 더 잘 제어할 수 있고, 결과적으로 인페인팅 효과가 개선되기 때문입니다. 앞서 사용했던 비유를 기억하시나요? 서로 다른 모델은 각각 다른 능력을 가진 예술가와 같으며, 각 예술가는 자신의 한계를 가지고 있습니다. 적합한 모델을 선택하면 더 나은 생성 결과를 얻을 수 있습니다. diff --git a/ko/tutorials/basic/outpaint.mdx b/ko/tutorials/basic/outpaint.mdx index fb246ba21..8211bd46e 100644 --- a/ko/tutorials/basic/outpaint.mdx +++ b/ko/tutorials/basic/outpaint.mdx @@ -34,7 +34,7 @@ AI 이미지 생성 과정에서 종종 기존 이미지의 구도는 좋지만 #### 1. 모델 설치 다음 모델 파일을 다운로드하여 `ComfyUI/models/checkpoints` 디렉토리에 저장하세요: -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) #### 2. 입력 이미지 diff --git a/ko/tutorials/controlnet/controlnet.mdx b/ko/tutorials/controlnet/controlnet.mdx index ca0a1ca24..a1ab868dc 100644 --- a/ko/tutorials/controlnet/controlnet.mdx +++ b/ko/tutorials/controlnet/controlnet.mdx @@ -73,8 +73,8 @@ ControlNet의 등장으로 우리는 추가적인 조건을 도입해 이미지 - [dreamCreationVirtual3DECommerce_v10.safetensors](https://civitai.com/api/download/models/731340?type=Model&format=SafeTensor&size=full&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/depth-controlnet.mdx b/ko/tutorials/controlnet/depth-controlnet.mdx index eded209da..d9241f202 100644 --- a/ko/tutorials/controlnet/depth-controlnet.mdx +++ b/ko/tutorials/controlnet/depth-controlnet.mdx @@ -54,7 +54,7 @@ Depth ControlNet은 깊이 맵 정보를 이해하고 활용하도록 특별히 - [architecturerealmix_v11.safetensors](https://civitai.com/api/download/models/431755?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) +- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/depth-t2i-adapter.mdx b/ko/tutorials/controlnet/depth-t2i-adapter.mdx index 413009f36..ffb32bc4d 100644 --- a/ko/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/ko/tutorials/controlnet/depth-t2i-adapter.mdx @@ -76,7 +76,7 @@ ComfyUI에서 T2I 어댑터를 사용하는 방법은 [ControlNet](/ko/tutorials - [interiordesignsuperm_v2.safetensors](https://civitai.com/api/download/models/93152?type=Model&format=SafeTensor&size=full&fp=fp16) -- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/resolve/main/models/t2iadapter_depth_sd15v2.pth?download=true) +- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_depth_sd15v2.pth?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/mixing-controlnets.mdx b/ko/tutorials/controlnet/mixing-controlnets.mdx index 2063e94ab..375237151 100644 --- a/ko/tutorials/controlnet/mixing-controlnets.mdx +++ b/ko/tutorials/controlnet/mixing-controlnets.mdx @@ -54,9 +54,9 @@ AI 이미지 생성에서 단일 제어 조건은 종종 복잡한 장면의 요 - [awpainting_v14.safetensors](https://civitai.com/api/download/models/624939?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx b/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx index 16a13ffcf..e7701c918 100644 --- a/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -47,10 +47,10 @@ AI 이미지 생성에서 OpenPose로 생성된 골격 구조 맵은 ControlNet 네트워크 환경상 해당 모델의 자동 다운로드가 성공적으로 이루어지지 않는다면, 아래 모델을 수동으로 다운로드해 지정된 디렉토리에 배치해주세요: -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) - [majicmixRealistic_v7.safetensors](https://civitai.com/api/download/models/176425?type=Model&format=SafeTensor&size=pruned&fp=fp16) - [japaneseStyleRealistic_v20.safetensors](https://civitai.com/api/download/models/85426?type=Model&format=SafeTensor&size=pruned&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/flux/flux-1-controlnet.mdx b/ko/tutorials/flux/flux-1-controlnet.mdx index 42ab13e6c..13217fd92 100644 --- a/ko/tutorials/flux/flux-1-controlnet.mdx +++ b/ko/tutorials/flux/flux-1-controlnet.mdx @@ -49,9 +49,14 @@ Depth 버전은 깊이 맵 추출 기법을 통해 원본 이미지의 공간적 ## FLUX.1-Canny-dev 전체 버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Canny" 검색 + + + Comfy Cloud에서 열기 + + ### 1. 워크플로우 및 자산 @@ -73,10 +78,10 @@ Depth 버전은 깊이 맵 추출 기법을 통해 원본 이미지의 공간적 완전한 모델 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true) (해당 리포지토리의 이용 약관에 동의했는지 확인해 주세요) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true) (해당 리포지토리의 이용 약관에 동의했는지 확인해 주세요) 파일 저장 위치: ``` @@ -116,9 +121,14 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora 워크플로우 - -

Comfy Cloud에서 실행

- + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Depth LoRA" 검색 + + + Comfy Cloud에서 열기 + + LoRA 버전 워크플로우는 완전한 버전을 기반으로 LoRA 모델을 추가한 것입니다. [Flux 워크플로우의 전체 버전](/ko/tutorials/flux/flux-1-text-to-image)과 비교해, 해당 LoRA 모델을 로드하고 사용하는 노드가 추가되었습니다. @@ -139,11 +149,11 @@ LoRA 버전 워크플로우는 완전한 버전을 기반으로 LoRA 모델을 완전한 모델 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/flux/flux-1-fill-dev.mdx b/ko/tutorials/flux/flux-1-fill-dev.mdx index 36a6de4e8..9a6398e02 100644 --- a/ko/tutorials/flux/flux-1-fill-dev.mdx +++ b/ko/tutorials/flux/flux-1-fill-dev.mdx @@ -40,10 +40,10 @@ Flux.1 fill dev의 주요 특징: ![Flux 약정](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 완성된 모델 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) 파일 저장 위치: ``` @@ -62,13 +62,14 @@ ComfyUI/ ### 1. 인페인팅 워크플로우 및 자산 - -

워크플로우 이미지 다운로드

- - - -

Comfy Cloud에서 실행

- + + + Download JSON or search "flux_fill_inpaint" in Template Library + + + Open in Comfy Cloud + + 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![ComfyUI Flux.1 인페인팅](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -97,6 +98,15 @@ ComfyUI/ ### 1. 아웃페인팅 워크플로우 및 자산 + + + Download JSON or search "flux_fill_outpaint" in Template Library + + + Open in Comfy Cloud + + + 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![ComfyUI Flux.1 아웃페인팅](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/ko/tutorials/flux/flux-1-kontext-dev.mdx b/ko/tutorials/flux/flux-1-kontext-dev.mdx index b74cc5546..7435bcd27 100644 --- a/ko/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ko/tutorials/flux/flux-1-kontext-dev.mdx @@ -37,9 +37,9 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 ### 버전 정보 -- **[FLUX.1 Kontext [pro]** - 상용 버전, 빠른 반복 편집에 중점 -- **FLUX.1 Kontext [max]** - 더 강력한 프롬프트 적합성을 갖춘 실험용 버전 -- **FLUX.1 Kontext [dev]** - 오픈소스 버전 (본 튜토리얼에서 사용), 120억 파라미터, 주로 연구용 +- **[FLUX.1 Kontext [pro]** — 상용 버전, 빠른 반복 편집에 중점 +- **FLUX.1 Kontext [max]** — 더 강력한 프롬프트 적합성을 갖춘 실험용 버전 +- **FLUX.1 Kontext [dev]** — 오픈소스 버전 (본 튜토리얼에서 사용), 120억 파라미터, 주로 연구용 현재 ComfyUI에서는 이 모든 버전을 사용할 수 있으며, [Pro 및 Max 버전](/ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext)은 파트너 노드를 통해 호출할 수 있고, Dev 오픈소스 버전은 본 가이드의 지침을 참고하세요. @@ -51,7 +51,7 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 **확산 모델** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) 원본 가중치를 사용하고 싶다면 블랙 포레스트 랩스의 관련 리포지토리를 방문해 원본 모델 가중치를 받아 사용할 수 있습니다. @@ -62,7 +62,7 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 **텍스트 인코더** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) 또는 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) 또는 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) 모델 저장 위치 @@ -80,9 +80,14 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 ## Flux.1 Kontext Dev 워크플로우 - -

Comfy Cloud에서 실행하기

- + + + JSON을 다운로드하거나 템플릿 라이브러리에서 "Flux Kontext Dev"를 검색하세요 + + + Comfy Cloud에서 열기 + + 이 워크플로우는 `Load Image(from output)` 노드를 사용해 편집할 이미지를 불러오므로, 여러 차례의 편집을 위해 편집된 이미지에 더욱 편리하게 접근할 수 있습니다. diff --git a/ko/tutorials/flux/flux-1-text-to-image.mdx b/ko/tutorials/flux/flux-1-text-to-image.mdx index 88611043c..2cd435ecc 100644 --- a/ko/tutorials/flux/flux-1-text-to-image.mdx +++ b/ko/tutorials/flux/flux-1-text-to-image.mdx @@ -50,25 +50,30 @@ Flux는 뛰어난 이미지 품질과 유연성으로 고화질의 다양한 이 #### 1. 워크플로우 파일 + + + Comfy Cloud에서 이 워크플로우 실행 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Dev" 검색 + + + 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Dev 원본 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - -

Comfy Cloud에서 실행

- - #### 2. 수동 모델 설치 - `flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) 계약에 동의해야 합니다. -- VRAM이 부족한 경우, [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true)를 사용해 `t5xxl_fp16.safetensors` 파일을 대체해보세요. +- VRAM이 부족한 경우, [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true)를 사용해 `t5xxl_fp16.safetensors` 파일을 대체해보세요. 다음 모델 파일을 다운로드하세요: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) 저장 위치: ``` @@ -103,14 +108,19 @@ Flux의 뛰어난 프롬프트 추종 능력 덕분에 부정적인 프롬프트 #### 1. 워크플로우 파일 + + + Comfy Cloud에서 이 워크플로우 실행 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Schnell" 검색 + + + 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Schnell 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) - -

Comfy Cloud에서 실행

- - #### 2. 수동 모델 설치 @@ -120,10 +130,10 @@ Flux의 뛰어난 프롬프트 추종 능력 덕분에 부정적인 프롬프트 완전한 모델 파일 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) 파일 저장 위치: ``` @@ -156,24 +166,38 @@ fp8 버전은 원본 Flux.1 fp16 버전의 양자화된 버전입니다. ### Flux.1 Dev + + + Comfy Cloud에서 이 워크플로우 실행 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Dev FP8" 검색 + + + 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Dev fp8 체크포인트 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) - -

Comfy Cloud에서 실행

- - -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. +[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. 해당 `Load Checkpoint` 노드가 `flux1-dev-fp8.safetensors`를 로드하도록 설정했는지 확인하고, 워크플로우를 실행해보세요. ### Flux.1 Schnell + + + Comfy Cloud에서 이 워크플로우 실행 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Schnell FP8" 검색 + + + 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Schnell fp8 체크포인트 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. +[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. 해당 `Load Checkpoint` 노드가 `flux1-schnell-fp8.safetensors`를 로드하도록 설정했는지 확인하고, 워크플로우를 실행해보세요. diff --git a/ko/tutorials/flux/flux-1-uso.mdx b/ko/tutorials/flux/flux-1-uso.mdx index 93f60cba9..ddb733350 100644 --- a/ko/tutorials/flux/flux-1-uso.mdx +++ b/ko/tutorials/flux/flux-1-uso.mdx @@ -31,18 +31,14 @@ USO는 세 가지 주요 방식을 지원합니다: ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - -

JSON 워크플로우 다운로드

- - - -

Comfy Cloud에서 실행

- + + + 워크플로우 JSON을 다운로드하고 ComfyUI로 드래그하세요 + + + Comfy Cloud에서 이 워크플로우 실행 + + 아래 이미지를 입력 이미지로 사용하세요. @@ -53,19 +49,19 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', **체크포인트** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) **로라** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **모델 패치** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **클립 비전** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) 모든 모델을 다운로드해 다음 디렉토리에 배치하세요: diff --git a/ko/tutorials/flux/flux-2-dev.mdx b/ko/tutorials/flux/flux-2-dev.mdx index 617406994..99e2d074e 100644 --- a/ko/tutorials/flux/flux-2-dev.mdx +++ b/ko/tutorials/flux/flux-2-dev.mdx @@ -65,15 +65,15 @@ FLUX.2 Dev를 사용하여 단일 이미지를 생성하는 기본적인 텍스 **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/flux/flux-2-klein.mdx b/ko/tutorials/flux/flux-2-klein.mdx index 69d31c6db..fff51e9d2 100644 --- a/ko/tutorials/flux/flux-2-klein.mdx +++ b/ko/tutorials/flux/flux-2-klein.mdx @@ -45,19 +45,19 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 ## Flux.2 Klein 4B 모델 다운로드 - + 4B 모델용 텍스트 인코더입니다. - + 확산 모델(4B 베이스). - + 확산 모델(4B 정제). - + 4B 모델용 VAE입니다. @@ -103,11 +103,11 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 확산 모델(9B 정제). - + 9B 모델용 텍스트 인코더입니다. - + 9B 모델용 VAE입니다. diff --git a/ko/tutorials/flux/flux1-krea-dev.mdx b/ko/tutorials/flux/flux1-krea-dev.mdx index f1cba3b61..8a63c8b7c 100644 --- a/ko/tutorials/flux/flux1-krea-dev.mdx +++ b/ko/tutorials/flux/flux1-krea-dev.mdx @@ -31,13 +31,14 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 아래 이미지 또는 JSON 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![Flux Krea Dev 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

JSON 워크플로우 다운로드

- - - -

Comfy Cloud에서 실행하기

- + + + Comfy Cloud에서 이 워크플로우 실행 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Krea Dev" 검색 + + #### 2. 수동 모델 설치 @@ -48,7 +49,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 더 높은 품질을 원하고 VRAM이 충분하다면 원본 모델 가중치를 사용해 보실 수 있습니다. -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) `flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 약정에 동의해야 합니다. @@ -57,12 +58,12 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 이전에 Flux 관련 워크플로우를 사용한 적이 있다면, 다음 모델들은 동일하므로 다시 다운로드할 필요가 없습니다. **텍스트 인코더** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/image/anima/anima.mdx b/ko/tutorials/image/anima/anima.mdx index 038e74d36..84ec6b714 100644 --- a/ko/tutorials/image/anima/anima.mdx +++ b/ko/tutorials/image/anima/anima.mdx @@ -84,15 +84,15 @@ Anima는 두 가지 워크플로를 제공합니다. 일반 사용을 위한 베 모델 파일은 Hugging Face의 [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima)에서 확인할 수 있습니다. - + Anima Base v1용 확산 모델(20억). - + 두 워크플로 공유 텍스트 인코더(Qwen-3 0.6B). - + 두 워크플로 공유 VAE. @@ -113,7 +113,7 @@ Anima는 두 가지 워크플로를 제공합니다. 일반 사용을 위한 베 미리보기 워크플로를 사용한다면, 대신 미리보기 확산 모델을 다운로드하세요: - + Anima Preview용 확산 모델(20억). diff --git a/ko/tutorials/image/boogu/boogu-image-0.1.mdx b/ko/tutorials/image/boogu/boogu-image-0.1.mdx index ae1e7aa94..043e48c95 100644 --- a/ko/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ko/tutorials/image/boogu/boogu-image-0.1.mdx @@ -50,19 +50,19 @@ Boogu-Image-0.1-Turbo 워크플로는 확산, 텍스트 인코딩 및 VAE 디코 ### Boogu-Image-0.1-Turbo 모델 다운로드 - + Boogu-Image-0.1-Turbo용 확산 모델. - + Boogu-Image-0.1-Turbo용 텍스트 인코더. - + Boogu-Image-0.1-Turbo용 VAE. - + Boogu-Image-0.1-Turbo용 LoRA 모듈 (rank-128). @@ -100,15 +100,15 @@ Boogu-Image-0.1-Turbo 워크플로는 확산, 텍스트 인코딩 및 VAE 디코 ### Boogu-Image-0.1-Edit 모델 다운로드 - + Boogu-Image-0.1-Edit용 확산 모델. - + Boogu-Image-0.1-Edit용 텍스트 인코더. - + Boogu-Image-0.1-Edit용 VAE. diff --git a/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index e72a01eb8..6f0fad7c3 100644 --- a/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -44,17 +44,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **디퓨전 모델** -- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors) +- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_t2i.safetensors) 기타 가중치는 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged)에서 다운로드해 주세요. **텍스트 인코더** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) 파일 저장 위치 diff --git a/ko/tutorials/image/ernie-image/ernie-image.mdx b/ko/tutorials/image/ernie-image/ernie-image.mdx index 8c72a09e3..079f9e47e 100644 --- a/ko/tutorials/image/ernie-image/ernie-image.mdx +++ b/ko/tutorials/image/ernie-image/ernie-image.mdx @@ -52,19 +52,19 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" 모든 재포장된 모델 파일은 Hugging Face의 [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image)에서 확인할 수 있습니다. - + ERNIE-Image용 디퓨전 모델입니다. - + ERNIE-Image용 텍스트 인코더입니다. - + ERNIE-Image용 프롬프트 향상기 텍스트 인코더입니다. - + ERNIE-Image용 VAE입니다. @@ -96,19 +96,19 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### ERNIE-Image-Turbo 모델 다운로드 - + ERNIE-Image-Turbo용 diffusion 모델입니다. - + ERNIE-Image-Turbo용 텍스트 인코더입니다. - + ERNIE-Image-Turbo용 프롬프트 향상기 텍스트 인코더입니다. - + ERNIE-Image-Turbo용 VAE입니다. diff --git a/ko/tutorials/image/hidream/hidream-e1.mdx b/ko/tutorials/image/hidream/hidream-e1.mdx index 86a1e2e38..a326ec40f 100644 --- a/ko/tutorials/image/hidream/hidream-e1.mdx +++ b/ko/tutorials/image/hidream/hidream-e1.mdx @@ -42,8 +42,8 @@ HiDream-E1은 HiDream-ai가 공식적으로 오픈소스로 배포한 대규모 **Diffusion 모델** 두 모델을 모두 다운로드할 필요는 없습니다. E1.1은 E1을 기반으로 한 반복 버전이므로, 우리의 테스트 결과에 따르면 E1보다 품질과 성능이 크게 향상되었습니다. -- [hidream_e1_1_bf16.safetensors (권장)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors (권장)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **텍스트 인코더**: @@ -76,6 +76,15 @@ HiDream-E1은 HiDream-ai가 공식적으로 오픈소스로 배포한 대규모 ## HiDream E1.1 ComfyUI 네이티브 워크플로우 예시 + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "HiDream E1.1" 검색 + + + E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전은 동적 1메가픽셀 해상도를 지원하며, 워크플로우에서는 `Scale Image to Total Pixels` 노드를 사용해 입력 이미지를 동적으로 100만 픽셀로 조정합니다. @@ -119,9 +128,14 @@ E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전 ## HiDream E1 ComfyUI 네이티브 워크플로우 예시 - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "HiDream E1 Full" 검색 + + E1은 2025년 4월 28일에 출시된 모델입니다. 이 모델은 768*768 해상도만 지원합니다. diff --git a/ko/tutorials/image/hidream/hidream-i1.mdx b/ko/tutorials/image/hidream/hidream-i1.mdx index 2b27ec88d..1042b02c6 100644 --- a/ko/tutorials/image/hidream/hidream-i1.mdx +++ b/ko/tutorials/image/hidream/hidream-i1.mdx @@ -98,16 +98,21 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 풀버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + 설정 없이 Comfy Cloud에서 이 워크플로우 실행 + + + 워크플로우 JSON 파일 다운로드 + + #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 @@ -136,15 +141,20 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Dev 버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + 설정 없이 Comfy Cloud에서 이 워크플로우 실행 + + + 워크플로우 JSON 파일 다운로드 + + #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. @@ -172,15 +182,20 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Fast 버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + 설정 없이 Comfy Cloud에서 이 워크플로우 실행 + + + 워크플로우 JSON 파일 다운로드 + + #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. diff --git a/ko/tutorials/image/hidream/hidream-o1.mdx b/ko/tutorials/image/hidream/hidream-o1.mdx index 4741dcbbe..cfa00d520 100644 --- a/ko/tutorials/image/hidream/hidream-o1.mdx +++ b/ko/tutorials/image/hidream/hidream-o1.mdx @@ -50,19 +50,19 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 **체크포인트** — 재패키징되고 양자화되었습니다. 모든 모델은 최악의 이상치에 대해 bf16을 사용하며, 사용되지 않는 딥스택 레이어는 제거되었습니다: -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 양자화된 변형 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 양자화된 변형 +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) **텍스트 인코더**(프롬프트 강화) — 모든 버전에서 공유됩니다: -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA(선택사항)** — Dev 디스틸레이션은 풀 모델에도 LoRA로 적용할 수 있으며, 디스틸레이션 강도를 조정할 수 있습니다([Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 제공): -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 전체 랭크 -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 프룬드 변형 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 대안적인 체크포인트 기반 디스틸레이션 +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 전체 랭크 +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 프룬드 변형 +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 대안적인 체크포인트 기반 디스틸레이션 ``` 📂 ComfyUI/ @@ -105,13 +105,13 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 **체크포인트(Dev)** — 재패키징되고 양자화되었습니다. 모든 모델은 최악의 이상치에 대해 bf16을 사용하며, 사용되지 않는 딥스택 레이어는 제거되었습니다: -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 양자화된 변형 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 양자화된 변형 +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) **텍스트 인코더**(프롬프트 강화) — 모든 버전에서 공유됩니다: -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/ko/tutorials/image/ideogram/ideogram-v4.mdx b/ko/tutorials/image/ideogram/ideogram-v4.mdx index 9ff8cc452..717de0dbd 100644 --- a/ko/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ko/tutorials/image/ideogram/ideogram-v4.mdx @@ -46,23 +46,23 @@ Ideogram 4.0은 Ideogram에서 출시한 최신 텍스트 기반 이미지 생 Hugging Face의 [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4)에서 모든 재포장된 모델 파일을 확인할 수 있습니다. - + Ideogram 4.0용 디퓨전 모델 (~13.8 GB). models/diffusion_models/에 저장하세요 - + Ideogram 4.0용 비조건 디퓨전 모델 (~13.8 GB). models/diffusion_models/에 저장하세요 - + Ideogram 4.0용 텍스트 인코더 (~8 GB). models/text_encoders/에 저장하세요 - + Ideogram 4.0용 텍스트 인코더 (~2 GB). models/text_encoders/에 저장하세요 - + Ideogram 4.0용 VAE (~335 MB). models/vae/에 저장하세요 diff --git a/ko/tutorials/image/krea/krea-2.mdx b/ko/tutorials/image/krea/krea-2.mdx index 78aaa2437..bf2cf4a7b 100644 --- a/ko/tutorials/image/krea/krea-2.mdx +++ b/ko/tutorials/image/krea/krea-2.mdx @@ -107,13 +107,13 @@ Krea는 또한 Krea 2용 스타일 LoRA 컬렉션을 출시했습니다. **Custo 로컬에서 사용하려면 [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2)에서 ComfyUI 최적화 모델 파일을 다운로드하세요. - + krea2_turbo_fp8_scaled.safetensors: Turbo FP8 (대부분의 사용자에게 권장) - + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B 텍스트 인코더 - + qwen_image_vae.safetensors diff --git a/ko/tutorials/image/lens/lens.mdx b/ko/tutorials/image/lens/lens.mdx index 8ceae12be..7d28b9d1c 100644 --- a/ko/tutorials/image/lens/lens.mdx +++ b/ko/tutorials/image/lens/lens.mdx @@ -94,19 +94,19 @@ Lens Turbo는 추출된 변형으로, 더 적은 샘플링 단계로 이미지 모든 모델 파일은 Hugging Face의 [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens)에서 찾을 수 있습니다. - + Lens용 디퓨전 모델(BF16). - + Lens Turbo용 디퓨전 모델(BF16). - + Lens와 Lens Turbo가 공유하는 텍스트 인코더(GPT-OSS-20B). - + Lens와 Lens Turbo가 공유하는 VAE(FLUX.2). diff --git a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 798943559..2c53a6f10 100644 --- a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -30,13 +30,14 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## NewBie-image 텍스트 기반 이미지 생성 워크플로 - -

JSON 워크플로 파일 다운로드

- - - -

ComfyUI 클라우드에서 실행하기

- +| +| +| JSON을 다운로드하거나 템플릿 라이브러리에서 'NewBie-image' 검색 +| +| +| 클라우드에서 열기 +| +| @@ -44,16 +45,16 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/omnigen/omnigen2.mdx b/ko/tutorials/image/omnigen/omnigen2.mdx index 78ed966cf..dcc09b6e3 100644 --- a/ko/tutorials/image/omnigen/omnigen2.mdx +++ b/ko/tutorials/image/omnigen/omnigen2.mdx @@ -42,13 +42,13 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 본 문서에는 다양한 워크플로우가 포함되어 있으므로, 해당 모델 파일과 설치 위치는 다음과 같습니다. 모델 파일의 다운로드 정보는 각 워크플로우에도 포함되어 있습니다. **확산 모델** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) 파일 저장 위치: ``` @@ -66,9 +66,11 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 ### 1. 워크플로우 파일 다운로드 - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 실행 + + ![텍스트 기반 이미지 생성 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -94,9 +96,11 @@ OmniGen2는 풍부한 이미지 편집 기능을 갖추고 있으며, 이미지 ### 1. 워크플로우 파일 다운로드 - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 실행 + + ![이미지 편집 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/ko/tutorials/image/ovis/ovis-image.mdx b/ko/tutorials/image/ovis/ovis-image.mdx index 6586e2039..5456cf94b 100644 --- a/ko/tutorials/image/ovis/ovis-image.mdx +++ b/ko/tutorials/image/ovis/ovis-image.mdx @@ -22,13 +22,14 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Ovis-Image 텍스트 기반 이미지 생성 워크플로우 - -

JSON 워크플로우 파일 다운로드

- - - -

ComfyUI 클라우드에서 실행하기

- + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Ovis image" 검색 + + @@ -36,15 +37,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/pixeldit/pixeldit.mdx b/ko/tutorials/image/pixeldit/pixeldit.mdx index bdd451a41..acda9a892 100644 --- a/ko/tutorials/image/pixeldit/pixeldit.mdx +++ b/ko/tutorials/image/pixeldit/pixeldit.mdx @@ -62,11 +62,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" PixelDiT는 두 개의 모델 파일을 사용합니다: 텍스트 인코더와 확산 모델입니다. - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 텍스트 인코더 - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 확산 모델 diff --git a/ko/tutorials/image/qwen/qwen-image-2512.mdx b/ko/tutorials/image/qwen/qwen-image-2512.mdx index e8b9a4dc4..04da73481 100644 --- a/ko/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ko/tutorials/image/qwen/qwen-image-2512.mdx @@ -43,9 +43,14 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - + + Comfy Cloud에서 실행하기 - + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Qwen-Image-2512" 검색 + + ### 1. 워크플로우 파일 @@ -55,28 +60,25 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 - **텍스트 기반 이미지 생성 (Qwen-Image 2512)**: 표준 50단계 생성 - **텍스트 기반 이미지 생성 (Qwen-Image 2512 4단계)**: Lightning LoRA를 사용한 가속화된 4단계 생성 - -

JSON 워크플로우 다운로드

- ### 2. 모델 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (선택사항 - 4단계 Lightning 가속화용)** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **디퓨전 모델** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (대부분의 사용자에게 권장됨) -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (충분한 VRAM을 보유하고 더 높은 품질을 원하는 경우) +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (대부분의 사용자에게 권장됨) +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (충분한 VRAM을 보유하고 더 높은 품질을 원하는 경우) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx index 8f3152cb1..3061dce3e 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -33,31 +33,32 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그하여 불러올 수 있습니다. - -

JSON 워크플로우 다운로드

- - - -

ComfyUI 클라우드에서 실행

- + + + ComfyUI 클라우드에서 실행 + + + JSON 워크플로우 다운로드 + + ### 2. 모델 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (선택사항 - 4단계 라이트닝 가속화용)** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **디퓨전 모델** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image-edit.mdx b/ko/tutorials/image/qwen/qwen-image-edit.mdx index 23f22ef51..132119b81 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit.mdx @@ -47,13 +47,14 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. ![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

JSON 워크플로우 다운로드

- - - -

ComfyUI 클라우드에서 실행

- + + + JSON 워크플로우를 다운로드하거나 템플릿 라이브러리에서 "image_qwen_image_edit" 검색 + + + ComfyUI 클라우드에서 이 워크플로우 실행 (제로 설정) + + 아래 이미지를 입력으로 다운로드하세요 ![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -64,19 +65,19 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 **디퓨전 모델** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) 모델 저장 위치 diff --git a/ko/tutorials/image/qwen/qwen-image-layered.mdx b/ko/tutorials/image/qwen/qwen-image-layered.mdx index 51aae2791..8da4e5dca 100644 --- a/ko/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ko/tutorials/image/qwen/qwen-image-layered.mdx @@ -30,13 +30,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 워크플로우 - -

JSON 워크플로우 파일 다운로드

- - - -

ComfyUI 클라우드에서 실행하기

- +| +| +| JSON 워크플로우 파일 다운로드 +| +| +| +| ComfyUI 클라우드에서 실행하기 +| +| @@ -44,15 +46,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image.mdx b/ko/tutorials/image/qwen/qwen-image.mdx index 1a4c7c333..c06554198 100644 --- a/ko/tutorials/image/qwen/qwen-image.mdx +++ b/ko/tutorials/image/qwen/qwen-image.mdx @@ -58,9 +58,12 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - - Comfy Cloud에서 실행하기 - + + + + + + 이 문서에 첨부된 워크플로우에는 세 가지 다른 모델이 사용됩니다: 1. Qwen-Image 원본 모델 fp8_e4m3fn @@ -82,14 +85,11 @@ GPU: RTX4090D 24GB ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. ![Qwen-image Text-to-Image 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

Qwen-Image 공식 모델용 워크플로우 다운로드

- + 증류 버전 - -

증류 모델용 워크플로우 다운로드

- + + ### 2. 모델 다운로드 @@ -103,12 +103,12 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 **디퓨전 모델** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 증류 버전의 원저자는 cfg 1.0에서 15단계 사용을 권장합니다. @@ -117,15 +117,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** @@ -164,18 +164,19 @@ Qwen_image_distill 이것은 ControlNet 모델이므로 일반 ControlNet처럼 사용할 수 있습니다. - - Comfy Cloud에서 실행하기 - + + + + + + ### 1. 워크플로우 및 입력 이미지 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

JSON 형식 워크플로우 다운로드

- + 아래 이미지를 입력으로 다운로드하세요. ![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -184,7 +185,7 @@ Qwen_image_distill 1. InstantX Controlnet -[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)를 다운로드해 `ComfyUI/models/controlnet/` 폴더에 저장하세요. +[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)를 다운로드해 `ComfyUI/models/controlnet/` 폴더에 저장하세요. 2. **Lotus Depth 모델** @@ -192,11 +193,11 @@ Qwen_image_distill **디퓨전 모델** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAE 모델** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) 또는 어떤 SD1.5 VAE도 가능합니다. +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) 또는 어떤 SD1.5 VAE도 가능합니다. ``` ComfyUI/ @@ -220,9 +221,12 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets 모델 패치 워크플로우 - - Comfy Cloud에서 실행하기 - + + + + + + 이 모델은 실제로 ControlNet이 아니라, 캐니, 딥스, 인페인트 등 세 가지 다른 제어 모드를 지원하는 모델 패치입니다. @@ -235,9 +239,7 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( 아래 이미지를 다운로드해 ComfyUI로 드래그해 해당 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

JSON 형식 워크플로우 다운로드

- + 아래 이미지를 입력으로 다운로드하세요: @@ -247,9 +249,9 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( 다른 모델은 Qwen-Image 기본 워크플로우와 동일합니다. 아래 모델만 다운로드해 `ComfyUI/models/model_patches` 폴더에 저장하면 됩니다. -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. 워크플로우 사용 지침 @@ -292,9 +294,12 @@ ControlNet 관련 워크플로우를 처음 사용한다면, 제어 이미지는 ## Qwen Image Union ControlNet LoRA 워크플로우 - - Comfy Cloud에서 실행하기 - + + + + + + 원본 모델 주소: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 캐니, 딥스, 포즈, 라인아트, 소프트엣지, 노말, 오픈포즈 지원 이미지 구조 제어 LoRA @@ -303,9 +308,7 @@ Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](htt 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

JSON 형식 워크플로우 다운로드

- + 아래 이미지를 입력으로 다운로드하세요. diff --git a/ko/tutorials/image/z-image/z-image-turbo.mdx b/ko/tutorials/image/z-image/z-image-turbo.mdx index f67002698..220f1c488 100644 --- a/ko/tutorials/image/z-image/z-image-turbo.mdx +++ b/ko/tutorials/image/z-image/z-image-turbo.mdx @@ -46,15 +46,15 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### Z-Image-Turbo 모델 다운로드 - + Z-Image-Turbo용 텍스트 인코더입니다. - + Z-Image-Turbo용 디퓨전 모델입니다. - + Z-Image-Turbo용 VAE입니다. @@ -81,7 +81,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### ControlNet용 추가 모델 - + Z-Image-Turbo용 ControlNet 모델 패치입니다. diff --git a/ko/tutorials/image/z-image/z-image.mdx b/ko/tutorials/image/z-image/z-image.mdx index 127f2ee20..34a2ec450 100644 --- a/ko/tutorials/image/z-image/z-image.mdx +++ b/ko/tutorials/image/z-image/z-image.mdx @@ -37,15 +37,15 @@ Z-Image (Base)는 커뮤니티 주도의 미세조정 및 맞춤형 개발을 ## Z-Image 모델 다운로드 - + Z-Image용 텍스트 인코더 - + Z-Image용 확산 모델 - + Z-Image용 VAE diff --git a/ko/tutorials/llm/gemma4/gemma4.mdx b/ko/tutorials/llm/gemma4/gemma4.mdx index 81eff1335..41acd9e86 100644 --- a/ko/tutorials/llm/gemma4/gemma4.mdx +++ b/ko/tutorials/llm/gemma4/gemma4.mdx @@ -74,11 +74,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" Gemma 4 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 관련 모델 파일을 다운로드하여 올바른 디렉토리에 저장하세요: - + 빠르고 가벼움. 소비자용 GPU에 권장됩니다. - + 균형 잡힌 성능. 워크플로의 기본 모델입니다. diff --git a/ko/tutorials/llm/qwen/qwen3.mdx b/ko/tutorials/llm/qwen/qwen3.mdx index f80db86ee..89eb6630f 100644 --- a/ko/tutorials/llm/qwen/qwen3.mdx +++ b/ko/tutorials/llm/qwen/qwen3.mdx @@ -71,15 +71,15 @@ Qwen 3.0은 ComfyUI 워크플로우 내에서 구조화된 텍스트 생성과 Qwen 3.0 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 모델 파일은 Qwen3.5와 공유되며, 하드웨어에 가장 적합한 변형을 다운로드하세요: - + 경량형, 약 4.5GB. 낮은 VRAM 환경과 빠른 다운로드에 최적입니다. - + 균형 잡힌 크기와 품질. 대부분의 소비자용 GPU에 권장됩니다. - + 가장 큰 변형, 약 19GB. 더 높은 품질의 출력을 제공하며, 더 많은 VRAM이 필요합니다. diff --git a/ko/tutorials/llm/qwen/qwen3_5.mdx b/ko/tutorials/llm/qwen/qwen3_5.mdx index 098918a6c..84d0dde5e 100644 --- a/ko/tutorials/llm/qwen/qwen3_5.mdx +++ b/ko/tutorials/llm/qwen/qwen3_5.mdx @@ -73,15 +73,15 @@ Qwen3.5는 시각적 이해와 텍스트 생성을 결합해 ComfyUI 워크플 Qwen3.5 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 하드웨어에 가장 적합한 변형을 선택하세요: - + 경량형, 약 4.5GB. 낮은 VRAM 환경과 빠른 다운로드에 최적입니다. - + 균형 잡힌 크기와 품질. 대부분의 소비자용 GPU에 권장됩니다. - + 가장 큰 변형, 약 19GB. 더 높은 품질의 출력을 제공하며, 더 많은 VRAM을 요구합니다. diff --git a/ko/tutorials/partner-nodes/google/gemini.mdx b/ko/tutorials/partner-nodes/google/gemini.mdx index a3e14d274..72dc0b71b 100644 --- a/ko/tutorials/partner-nodes/google/gemini.mdx +++ b/ko/tutorials/partner-nodes/google/gemini.mdx @@ -22,13 +22,11 @@ Google Gemini는 구글이 개발한 강력한 AI 모델로, 대화 및 텍스 아래 Json 파일을 다운로드한 후, ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + Json 형식 워크플로우 파일 다운로드 + + ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx index af7f71b93..b288a04fa 100644 --- a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -51,9 +51,8 @@ Kling 2.6 모션 컨트롤은 Kuaishou가 개발한 특수 다중모달 모델 ## Kling 2.6 모션 컨트롤 작업 흐름 - -

워크플로우 파일을 JSON 형식으로 다운로드하세요

- + + ## 입력 요구사항 diff --git a/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 36c342eb2..902c51f3a 100644 --- a/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -54,9 +54,11 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_text_to_video.mp4" > - -

워크플로우 파일(JSON 형식) 다운로드하기

- + + + 워크플로우 파일(JSON 형식) 다운로드하기 + + ### 2. 워크플로우 실행 단계 따라하기 @@ -80,9 +82,11 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_image_to_video.mp4" > - -

워크플로우 파일(JSON 형식) 다운로드하기

- + + + 워크플로우 파일(JSON 형식) 다운로드하기 + + 아래 이미지를 입력 이미지로 다운로드하세요 @@ -112,9 +116,11 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video.mp4" > - -

워크플로우 파일(JSON 형식) 다운로드하기

- + + + 워크플로우 파일(JSON 형식) 다운로드하기 + + 아래 비디오를 입력 비디오로 다운로드하세요: diff --git a/ko/tutorials/partner-nodes/openai/chat.mdx b/ko/tutorials/partner-nodes/openai/chat.mdx index e431f97a4..62be63ccc 100644 --- a/ko/tutorials/partner-nodes/openai/chat.mdx +++ b/ko/tutorials/partner-nodes/openai/chat.mdx @@ -22,13 +22,9 @@ OpenAI는 생성형 AI에 중점을 둔 회사로, 강력한 대화 기능을 아래 Json 파일을 다운로드한 후, ComfyUI로 드래그하여 해당 워크플로우를 불러오세요. - -

Json 형식 워크플로우 파일 다운로드

- +| +| Json 형식 워크플로우 파일 다운로드 +| ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/rodin/model-generation.mdx b/ko/tutorials/partner-nodes/rodin/model-generation.mdx index ad386d281..c0d95fe25 100644 --- a/ko/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ko/tutorials/partner-nodes/rodin/model-generation.mdx @@ -32,13 +32,9 @@ ComfyUI는 이제 해당 Rodin 모델 생성 API를 기본적으로 통합하여 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + 단일뷰 모델 생성 (Json 형식) + 입력 이미지로 아래 이미지를 다운로드하세요. @@ -66,13 +62,9 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + 다중뷰 모델 생성 (Json 형식) + 입력 이미지로 아래 이미지를 다운로드하세요. diff --git a/ko/tutorials/partner-nodes/runway/video-generation.mdx b/ko/tutorials/partner-nodes/runway/video-generation.mdx index aab91e761..479ae13aa 100644 --- a/ko/tutorials/partner-nodes/runway/video-generation.mdx +++ b/ko/tutorials/partner-nodes/runway/video-generation.mdx @@ -37,9 +37,7 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen3a_turbo_image_to_video/runway_image_to_video_gen3a_turbo.mp4" > - -

Json 형식 워크플로우 파일 다운로드

- + 아래 이미지를 입력 이미지로 다운로드하세요. @@ -67,9 +65,7 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen4_turbo_image_to_video/runway_gen4_turo_image_to_video.mp4" > - -

Json 형식 워크플로우 파일 다운로드

- + 아래 이미지를 입력 이미지로 다운로드하세요. @@ -99,9 +95,7 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/first_last_frame_to_video/runway_first_last_frame.mp4" > - -

Json 형식 워크플로우 파일 다운로드

- + 아래 이미지를 입력 이미지로 다운로드하세요. diff --git a/ko/tutorials/partner-nodes/tripo/model-generation.mdx b/ko/tutorials/partner-nodes/tripo/model-generation.mdx index f51ee9ef9..ad2ec6454 100644 --- a/ko/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ko/tutorials/partner-nodes/tripo/model-generation.mdx @@ -35,9 +35,11 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + 텍스트 기반 3D 모델 생성 워크플로우 + + ### 2. 워크플로우를 단계별로 완료하세요 @@ -58,9 +60,11 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + 이미지 기반 3D 모델 생성 워크플로우 + + 아래 이미지를 입력 이미지로 다운로드하세요. @@ -86,9 +90,11 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + 멀티뷰 모델 생성 워크플로우 + + 아래 이미지를 입력 이미지로 다운로드하세요. diff --git a/ko/tutorials/utility/depth-anything-3.mdx b/ko/tutorials/utility/depth-anything-3.mdx index 6dddf7005..ebc59b59b 100644 --- a/ko/tutorials/utility/depth-anything-3.mdx +++ b/ko/tutorials/utility/depth-anything-3.mdx @@ -39,10 +39,10 @@ ComfyUI는 이제 Depth Anything 3 노드를 기본 지원합니다. 시작하 Depth Anything 3 체크포인트를 다운로드하여 해당 ComfyUI 폴더에 저장합니다: -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — 가볍고 빠른 추론 -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — 균형 잡힌 성능 -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 단일 뷰 깊이에 최적 (하늘 감지 포함) -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 미터 단위의 물리적 깊이 (하늘 감지 포함) +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — 가볍고 빠른 추론 +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — 균형 잡힌 성능 +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 단일 뷰 깊이에 최적 (하늘 감지 포함) +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 미터 단위의 물리적 깊이 (하늘 감지 포함) ``` ComfyUI/ diff --git a/ko/tutorials/utility/face-detection/mediapipe.mdx b/ko/tutorials/utility/face-detection/mediapipe.mdx index 539561b3c..e80bdec03 100644 --- a/ko/tutorials/utility/face-detection/mediapipe.mdx +++ b/ko/tutorials/utility/face-detection/mediapipe.mdx @@ -55,7 +55,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` → `템 MediaPipe 얼굴 감지 모델은 [Comfy-Org MediaPipe 모델 저장소](https://huggingface.co/Comfy-Org/mediapipe)에 호스팅되어 있습니다. -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) 다음과 같은 디렉토리 구조에 배치하세요: diff --git a/ko/tutorials/utility/moge.mdx b/ko/tutorials/utility/moge.mdx index af99d09df..9867dd52b 100644 --- a/ko/tutorials/utility/moge.mdx +++ b/ko/tutorials/utility/moge.mdx @@ -50,8 +50,8 @@ ComfyUI는 이제 MoGe 노드를 기본적으로 지원합니다. 시작하기 MoGe 체크포인트를 다운로드하고 해당 ComfyUI 폴더에 저장하세요: -- **MoGe-2 (권장)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1 (기준)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2 (권장)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1 (기준)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/utility/pose-detection-sdpose.mdx b/ko/tutorials/utility/pose-detection-sdpose.mdx index e4f0b2dbb..8ad8fdd07 100644 --- a/ko/tutorials/utility/pose-detection-sdpose.mdx +++ b/ko/tutorials/utility/pose-detection-sdpose.mdx @@ -87,11 +87,11 @@ ComfyUI를 최신 버전으로 업데이트한 다음, `Workflow` -> `Browse Tem SDPose 및 RT-DETRv4 모델 체크포인트는 [Comfy-Org SDPose 모델 저장소](https://huggingface.co/Comfy-Org/SDPose)에 호스팅되어 있습니다. **체크포인트** (SDPose 모델): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models** (RT-DETRv4 감지기): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (권장) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (전체 정밀도, 용량이 큼) +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (권장) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (전체 정밀도, 용량이 큼) 다음과 같은 디렉토리 구조에 배치하세요: diff --git a/ko/tutorials/utility/remove-background-birefnet.mdx b/ko/tutorials/utility/remove-background-birefnet.mdx index b23e07d8b..1e4acf988 100644 --- a/ko/tutorials/utility/remove-background-birefnet.mdx +++ b/ko/tutorials/utility/remove-background-birefnet.mdx @@ -48,7 +48,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 BiRefNet 모델은 [Comfy-Org BiRefNet 모델 저장소](https://huggingface.co/Comfy-Org/BiRefNet)에 호스팅되어 있습니다. -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) 다음과 같은 디렉토리 구조에 배치하세요: diff --git a/ko/tutorials/utility/video-segment-sam3.mdx b/ko/tutorials/utility/video-segment-sam3.mdx index 8dfa7da68..8efcee351 100644 --- a/ko/tutorials/utility/video-segment-sam3.mdx +++ b/ko/tutorials/utility/video-segment-sam3.mdx @@ -68,7 +68,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 SAM 3.1 모델은 [Comfy-Org SAM 3.1 모델 저장소](https://huggingface.co/Comfy-Org/sam3.1)에 호스팅되어 있습니다. -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) 다음과 같은 디렉터리 구조에 배치하세요: diff --git a/ko/tutorials/utility/void-video-inpainting.mdx b/ko/tutorials/utility/void-video-inpainting.mdx index e8b63bf01..f26c7f60d 100644 --- a/ko/tutorials/utility/void-video-inpainting.mdx +++ b/ko/tutorials/utility/void-video-inpainting.mdx @@ -65,24 +65,24 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 **확산 모델** — 핵심 2회 통과 인페인팅 모델: -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 정밀화 통과, 더 나은 시간적 안정성 -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 기본 통과 +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — 정밀화 통과, 더 나은 시간적 안정성 +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — 기본 통과 **VAE:** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) **광학 흐름:** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 체크포인트** — 세그먼테이션용: -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **텍스트 인코더:** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ diff --git a/ko/tutorials/video/bytedance/bernini-r.mdx b/ko/tutorials/video/bytedance/bernini-r.mdx index ba520d810..82597a548 100644 --- a/ko/tutorials/video/bytedance/bernini-r.mdx +++ b/ko/tutorials/video/bytedance/bernini-r.mdx @@ -49,16 +49,16 @@ ComfyUI는 이제 Bernini-R 노드를 기본 지원합니다. 시작하기 전 필요한 모델 가중치를 다운로드하여 해당 ComfyUI 폴더에 저장합니다: **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index ce6479416..0d25fe5dd 100644 --- a/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -15,17 +15,27 @@ Cosmos-Predict2는 텍스트 기반 이미지 생성(Text2Image)과 비디오- 산업 시뮬레이션, 자율주행, 도시 계획, 과학 연구 등 여러 분야에서 널리 사용되고 있습니다. 이는 지능형 비전과 물리 세계의 깊은 통합을 촉진하는 핵심 기반 도구입니다. -GitHub: [Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) -huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) + + + Cosmos-Predict2 소스 코드 및 문서 + + + Cosmos-Predict2 모음집 + + 이 가이드는 ComfyUI에서 **Video2World** 생성을 완료하는 과정을 안내합니다. 텍스트 기반 이미지 생성 섹션은 다음을 참고하세요: - - Cosmos-Predict2를 이용한 텍스트 기반 이미지 생성 - -{/* + + + Cosmos-Predict2를 이용한 텍스트 기반 이미지 생성 + + + 강력한 GPU로 Comfy Cloud에서 Cosmos-Predict2 워크플로우 실행 + + ## Cosmos Predict2 Video2World 워크플로우 @@ -42,9 +52,14 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - -

Json 형식 워크플로우 파일 다운로드

- + + + JSON 형식 워크플로우 파일 다운로드 + + + Comfy Cloud에서 이 워크플로우 실행 (모델 사전 설치됨) + + 다음 이미지를 입력으로 다운로드하세요: @@ -56,17 +71,23 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **확산 모델** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) + + cosmos_predict2_2B_video2world_480p_16fps.safetensors + 기타 가중치는 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged)에서 다운로드하세요. **텍스트 인코더** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) + + oldt5_xxl_fp8_e4m3fn_scaled.safetensors + **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + wan_2.1_vae.safetensors + 파일 저장 위치 ``` @@ -93,4 +114,4 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- 6. (선택사항) `ClipTextEncode` 노드에서 프롬프트를 수정할 수 있습니다. 7. (선택사항) `CosmosPredict2ImageToVideoLatent` 노드에서 크기와 프레임 수를 조정합니다. 8. `Run` 버튼을 클릭하거나 `Ctrl(cmd) + Enter` 단축키를 눌러 워크플로우를 실행합니다. -9. 생성이 완료되면 비디오가 자동으로 `ComfyUI/output/` 디렉토리에 저장되며, `save video` 노드에서 미리볼 수 있습니다. */} \ No newline at end of file +9. 생성이 완료되면 비디오가 자동으로 `ComfyUI/output/` 디렉토리에 저장되며, `save video` 노드에서 미리볼 수 있습니다. \ No newline at end of file diff --git a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 414f7b0d2..8dd2bba11 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -29,17 +29,17 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 모델 링크 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **디퓨전 모델** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **VAE** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) 모델 저장 위치 diff --git a/ko/tutorials/video/hunyuan/hunyuan-video.mdx b/ko/tutorials/video/hunyuan/hunyuan-video.mdx index 16902e496..cb53c37e7 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video.mdx @@ -46,9 +46,9 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델들은 텍스트-투-비디오와 이미지-투-비디오 워크플로우 모두에 사용됩니다. 아래 모델들을 다운로드하여 지정된 디렉토리에 저장해주세요: -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors?download=true) +- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) 저장 위치: @@ -73,7 +73,7 @@ ComfyUI/ ### 2. 수동 모델 설치 -[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models` 폴더에 저장하세요. +[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models` 폴더에 저장하세요. 다음 모델 파일들이 올바른 위치에 있는지 확인하세요: @@ -126,7 +126,7 @@ ComfyUI/ ### v1과 v2 버전 공통 모델 다음 파일을 다운로드해 `ComfyUI/models/clip_vision` 디렉토리에 저장하세요: -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) +- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) ### v1 "concat" 이미지-투-비디오 워크플로우 @@ -140,7 +140,7 @@ ComfyUI/ #### 2. 관련 모델 수동 설치 -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) 다음 모델 파일들이 올바른 위치에 있는지 확인하세요: @@ -185,7 +185,7 @@ v2 워크플로우는 기본적으로 v1 워크플로우와 동일합니다. 다 #### 2. 관련 모델 수동 설치 -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) 다음 모델 파일들이 올바른 위치에 있는지 확인하세요: diff --git a/ko/tutorials/video/kandinsky/kandinsky-5.mdx b/ko/tutorials/video/kandinsky/kandinsky-5.mdx index f75218eb3..4730e9e55 100644 --- a/ko/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/ko/tutorials/video/kandinsky/kandinsky-5.mdx @@ -51,21 +51,39 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ComfyUI를 최신 버전으로 업데이트해 주시고, 메뉴 `워크플로우` -> `템플릿 탐색` -> `비디오`를 통해 "칸딘스키 5.0 T2V"를 찾아 워크플로우를 로드해 주세요. - -

JSON 워크플로우 파일 다운로드

- + + + T2V 워크플로우를 다운로드하여 로컬에서 사용 + + + Comfy Cloud에서 열기 + + ### 2. 모델 수동 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B 텍스트 인코더 (FP8) + + + CLIP-L 텍스트 인코더 + + **확산 모델** -- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) + + + Kandinsky 5.0 T2V Lite SFT 확산 모델 (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ @@ -85,21 +103,39 @@ ComfyUI/ ComfyUI를 최신 버전으로 업데이트해 주시고, 메뉴 `워크플로우` -> `템플릿 탐색` -> `비디오`를 통해 "칸딘스키 5.0 I2V"를 찾아 워크플로우를 로드해 주세요. - -

JSON 워크플로우 파일 다운로드

- + + + I2V 워크플로우를 다운로드하여 로컬에서 사용 + + + Comfy Cloud에서 열기 + + ### 2. 모델 수동 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B 텍스트 인코더 (FP8) + + + CLIP-L 텍스트 인코더 + + **확산 모델** -- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) + + + Kandinsky 5.0 I2V Lite 확산 모델 (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ diff --git a/ko/tutorials/video/ltxv.mdx b/ko/tutorials/video/ltxv.mdx index 67608d2eb..6f361c5d4 100644 --- a/ko/tutorials/video/ltxv.mdx +++ b/ko/tutorials/video/ltxv.mdx @@ -26,9 +26,17 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 첫 번째 [프레임 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)를 통해 비디오를 제어할 수 있습니다. - -

Comfy Cloud에서 실행하기

- + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "LTX-Video" 검색 + + + 이 워크플로우의 예제 입력 이미지 가져오기 + + LTX-Video 이미지에서 비디오로 @@ -38,6 +46,15 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 텍스트에서 비디오로 + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "LTX-Video" 검색 + + + LTX-Video 텍스트에서 비디오로 @@ -48,8 +65,13 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델을 다운로드하여 아래 지정된 위치에 배치하세요: -- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true) + + 다운로드하여 ComfyUI/models/checkpoints/에 배치 + + + + 다운로드하여 ComfyUI/models/text_encoders/에 배치 + ``` ├── checkpoints/ diff --git a/ko/tutorials/video/wan/fun-camera.mdx b/ko/tutorials/video/wan/fun-camera.mdx index d4be5c366..fa8d854f9 100644 --- a/ko/tutorials/video/wan/fun-camera.mdx +++ b/ko/tutorials/video/wan/fun-camera.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Performance Reference": 32425486 --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Wan2.1 Fun Camera 소개 @@ -36,21 +35,49 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 다음 모델들은 모두 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged)에서 찾을 수 있습니다. -**디퓨전 모델**은 1.3B 또는 14B 중 하나를 선택하세요: -- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) -- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) +### 디퓨전 모델 + +1.3B 또는 14B 중 하나를 선택하세요: + + + + Wan2.1 Fun Camera 1.3B 디퓨전 모델 + + + Wan2.1 Fun Camera 14B 디퓨전 모델 + + 이전에 Wan2.1 관련 모델을 사용한 적이 있다면 이미 다음 모델들이 있을 것입니다. 아직 없다면 다운로드해 주세요: -**텍스트 인코더**는 하나를 선택하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +### 텍스트 인코더 + +하나를 선택하세요: + + + + 전체 정밀도 텍스트 인코더 + + + FP8 양자화 텍스트 인코더 (낮은 VRAM 권장) + + + +### VAE -**VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + + Wan2.1 VAE 모델 + + -**CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +### CLIP 비전 + + + + CLIP 비전 인코더 + + 파일 저장 위치: @@ -70,9 +97,16 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 1.3B 기본 워크플로우 예시 -### 1. 워크플로우 관련 파일 다운로드 +### 1. 워크플로우 다운로드 -#### 1.1 워크플로우 파일 + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan 2.1 Fun Camera 1.3B" 검색 + + 아래 동영상을 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요: @@ -82,21 +116,19 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > - -

Json 워크플로우 파일 다운로드

- - 14B 버전을 사용하고 싶다면 모델 파일을 14B 버전으로 교체하면 되지만, VRAM 요구 사항에 유의해 주세요. -#### 1.2 입력 이미지 다운로드 - -아래 이미지를 다운로드해 시작 프레임으로 사용하세요: +### 2. 입력 자료 다운로드 -![입력 참조 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) + + + 아래 이미지를 다운로드하여 1.3B 워크플로우의 시작 프레임으로 사용하세요 + + -### 2. 워크플로우 단계별 완료하기 +### 3. 워크플로우 단계별 완료하기 ![Wan2.1 Fun Camera 워크플로우 단계](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -113,18 +145,30 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 14B 워크플로우 및 입력 이미지 +### 1. 워크플로우 다운로드 + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan 2.1 Fun Camera 14B" 검색 + + + - -

Json 워크플로우 파일 다운로드

- +### 2. 입력 자료 다운로드 -**입력 이미지** -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) + + + 아래 이미지를 다운로드하여 14B 워크플로우의 시작 프레임으로 사용하세요 + + ## 성능 참고사항 diff --git a/ko/tutorials/video/wan/fun-control.mdx b/ko/tutorials/video/wan/fun-control.mdx index 9cb0cc3d9..a2467675e 100644 --- a/ko/tutorials/video/wan/fun-control.mdx +++ b/ko/tutorials/video/wan/fun-control.mdx @@ -52,18 +52,18 @@ ComfyUI는 현재 Wan2.1 Fun Control 모델을 **네이티브로 지원**합니 해당 링크를 클릭해 다운로드하세요. 이전에 Wan 관련 워크플로우를 사용한 적이 있다면 **Diffusion 모델**만 다운로드하면 됩니다. **Diffusion 모델** - 1.3B 또는 14B 중 선택하세요. 14B 버전은 파일 크기가 더 크고 (32GB), VRAM 요구 사항도 높습니다: -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-Control.safetensors`로 이름 변경 **텍스트 인코더** - 다음 모델 중 하나를 선택하세요 (fp16 정밀도는 파일 크기가 더 크고 성능 요구 사항도 높습니다): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/video/wan/fun-inp.mdx b/ko/tutorials/video/wan/fun-inp.mdx index dc9d11849..6bfd4ff40 100644 --- a/ko/tutorials/video/wan/fun-inp.mdx +++ b/ko/tutorials/video/wan/fun-inp.mdx @@ -53,18 +53,18 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델들은 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 및 [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334)에서 찾을 수 있습니다. **디퓨전 모델** - 1.3B 또는 14B를 선택하세요. 14B 버전은 파일 크기가 더 크고(32GB), VRAM 요구 사항도 높습니다: -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-InP.safetensors`로 이름 변경 +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-InP.safetensors`로 이름 변경 **텍스트 인코더** - 다음 모델 중 하나를 선택하세요 (fp16 정밀도는 크기가 더 크고 성능 요구 사항도 높습니다): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/video/wan/vace.mdx b/ko/tutorials/video/wan/vace.mdx index 7069223b3..4b8f0aed2 100644 --- a/ko/tutorials/video/wan/vace.mdx +++ b/ko/tutorials/video/wan/vace.mdx @@ -59,18 +59,18 @@ VACE 14B는 알리바바 통이 완샹팀이 출시한 오픈소스 통합 비 ### 모델 다운로드 **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 이전에 Wan Video 관련 워크플로를 사용하셨다면 이미 다음 모델 파일을 다운로드하셨을 것입니다. **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **텍스트 인코더** 중 하나를 선택해 다운로드하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/wan/wan-ati.mdx b/ko/tutorials/video/wan/wan-ati.mdx index 345de9b22..c8d1aee32 100644 --- a/ko/tutorials/video/wan/wan-ati.mdx +++ b/ko/tutorials/video/wan/wan-ati.mdx @@ -47,18 +47,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 워크플로우에서 모델 파일을 성공적으로 다운로드하지 못했다면, 아래 링크를 이용해 수동으로 다운로드해 보세요. **디퓨전 모델** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **텍스트 인코더** 다음 모델 중 하나를 선택하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/wan/wan-causal-forcing.mdx b/ko/tutorials/video/wan/wan-causal-forcing.mdx index 7f90de227..c6231a10a 100644 --- a/ko/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ko/tutorials/video/wan/wan-causal-forcing.mdx @@ -83,10 +83,10 @@ Wan2.1 I2V 모델과 필요한 파일을 다운로드합니다. 해당 `models/` ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B 체크포인트 - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B 체크포인트 (최소 8GB VRAM) @@ -94,10 +94,10 @@ Wan2.1 I2V 모델과 필요한 파일을 다운로드합니다. 해당 `models/` ### CLIP 및 VAE - + google-bert/bert-base-uncased — CLIP 텍스트 인코더 - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/ko/tutorials/video/wan/wan-dancer.mdx b/ko/tutorials/video/wan/wan-dancer.mdx index 87022fb2d..2f90e471a 100644 --- a/ko/tutorials/video/wan/wan-dancer.mdx +++ b/ko/tutorials/video/wan/wan-dancer.mdx @@ -61,20 +61,20 @@ ComfyUI를 최신 버전으로 업데이트한 후 워크플로 파일을 다운 ### 3. 모델 수동 다운로드 **확산 모델** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **텍스트 인코더** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan-flf.mdx b/ko/tutorials/video/wan/wan-flf.mdx index 825284aa2..97b624eea 100644 --- a/ko/tutorials/video/wan/wan-flf.mdx +++ b/ko/tutorials/video/wan/wan-flf.mdx @@ -58,7 +58,7 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 이 가이드에 포함된 모든 모델은 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)에서 확인할 수 있습니다. **diffusion_models** 하드웨어 환경에 따라 버전을 선택하세요. -- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8: [wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -66,14 +66,14 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 **Text encoders**에서 버전을 하나 선택해 다운로드하세요. -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치 diff --git a/ko/tutorials/video/wan/wan-move.mdx b/ko/tutorials/video/wan/wan-move.mdx index b767247e3..061f43ccb 100644 --- a/ko/tutorials/video/wan/wan-move.mdx +++ b/ko/tutorials/video/wan/wan-move.mdx @@ -28,37 +28,37 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Wan-Move 이미지에서 비디오 워크플로우 - -

JSON 워크플로우 파일 다운로드

- + + 워크플로우 다운로드 + - -

ComfyUI 클라우드에서 실행하기

- + + 클라우드에서 열기 + ## 모델 링크 -**텍스트 인코더** + + **텍스트 인코더** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors + -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + **클립 비전** -- clip_vision_h.safetensors + -**클립 비전** + + **로라** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors + -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) + + **디퓨전 모델** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors + -**로라** - -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) - -**디퓨전 모델** - -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) - -**VAE** - -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + **VAE** -- wan_2.1_vae.safetensors + **모델 저장 위치** diff --git a/ko/tutorials/video/wan/wan-video.mdx b/ko/tutorials/video/wan/wan-video.mdx index 657814f1b..13d23d236 100644 --- a/ko/tutorials/video/wan/wan-video.mdx +++ b/ko/tutorials/video/wan/wan-video.mdx @@ -41,14 +41,14 @@ Wan2.1 Video 시리즈는 2025년 2월 알리바바가 [Apache 2.0 라이선스] 이 가이드에서 언급된 모든 모델은 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)에서 확인할 수 있습니다. 아래는 이 가이드의 예시에 필요한 일반적인 모델들로, 미리 다운로드해 두시면 됩니다: **텍스트 인코더**에서 하나의 버전을 선택해 다운로드하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` @@ -70,7 +70,7 @@ ComfyUI/ ## Wan2.1 텍스트 투 비디오 워크플로우 -워크플로우를 시작하기 전에 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +워크플로우를 시작하기 전에 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. > 다른 t2v 정밀도 버전이 필요하시다면 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models)를 방문해 다운로드해 주세요. @@ -108,7 +108,7 @@ ComfyUI/ #### 2. 모델 다운로드 -[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. #### 3. 워크플로우 단계별 완료 @@ -136,7 +136,7 @@ ComfyUI/ #### 2. 모델 다운로드 -[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. #### 3. 워크플로우 단계별 완료 diff --git a/ko/tutorials/video/wan/wan2-2-animate.mdx b/ko/tutorials/video/wan/wan2-2-animate.mdx index 9155a436e..3b7f9a838 100644 --- a/ko/tutorials/video/wan/wan2-2-animate.mdx +++ b/ko/tutorials/video/wan/wan2-2-animate.mdx @@ -55,13 +55,14 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 다음 워크플로우 파일을 다운로드해 ComfyUI로 끌어다 놓으면 워크플로우가 로드됩니다. - -

JSON 워크플로우 다운로드

- - - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 실행 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Animate" 검색 + + 아래 자료를 입력으로 다운로드하세요: @@ -77,21 +78,40 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 ### 2. 모델 링크 -**diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) Kijai의 리포지토리에서 가져온 모델입니다. -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 원본 모델 가중치 +**Diffusion Models** + + + + Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: Kijai의 리포지토리에서 제공하는 스케일링 FP8 모델 + + + wan2.2_animate_14B_bf16.safetensors: 원본 bf16 모델 가중치 + + + +**CLIP Vision** + + + clip_vision_h.safetensors: CLIP Vision 인코더 + + +**LoRAs** + + + lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4단계 가속 LoRA + -**clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +**VAE** -**loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 4단계 가속 Lora입니다. + + wan_2.1_vae.safetensors: 인코딩 및 디코딩용 Wan2.1 VAE + -**vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +**Text Encoders** -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: 스케일링 FP8 텍스트 인코더 + ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-fun-camera.mdx b/ko/tutorials/video/wan/wan2-2-fun-camera.mdx index ae9e27c58..7645aa374 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -46,31 +46,60 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

JSON 워크플로우 다운로드

- + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Camera" 검색 + + 아래 이미지를 다운로드해 입력으로 사용하세요. -![입력 시작 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) + + + 동영상 생성을 위한 시작 프레임입니다. 이 이미지를 다운로드하여 사용하거나, 자신의 이미지로 교체하세요. + + ### 2. 모델 링크 다음 모델들은 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인할 수 있습니다: **디퓨전 모델** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) + + + + Wan2.2 Fun Camera 고노이즈 디퓨전 모델 + + + Wan2.2 Fun Camera 저노이즈 디퓨전 모델 + + **Wan2.2-Lightning LoRA (선택사항, 가속화용)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + 고노이즈 모델용 4단계 가속 LoRA + + + 저노이즈 모델용 4단계 가속 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + 인코딩/디코딩용 Wan2.1 VAE + + +**텍스트 인코더** + + + FP8 스케일링 텍스트 인코더 + 파일 저장 위치 @@ -112,4 +141,4 @@ ComfyUI/ - **너비/높이**: 동영상 해상도 설정 - **길이**: 동영상 프레임 수 설정(기본값은 81프레임) - **속도**: 동영상 속도 설정(기본값은 1.0) -8. `Run` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 동영상 생성을 실행하세요. \ No newline at end of file +8. `Run` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 동영상 생성을 실행하세요. diff --git a/ko/tutorials/video/wan/wan2-2-fun-control.mdx b/ko/tutorials/video/wan/wan2-2-fun-control.mdx index b60dbd1cc..416a9b56e 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-control.mdx @@ -55,27 +55,27 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ### 1. 워크플로우 및 자료 다운로드 -아래 비디오 또는 JSON 파일을 다운로드해 ComfyUI로 드래그하여 워크플로우를 로드하세요. +ComfyUI를 최신 버전으로 업데이트한 후 워크플로우 파일을 다운로드하여 ComfyUI로 드래그하거나, 템플릿 라이브러리의 `Workflow` → `Browse Templates` → `Video`에서 "Wan2.2 Fun Control"을 찾아보세요. - - - -

JSON 워크플로우 다운로드

- + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Control" 검색 + + 다음 이미지와 비디오를 입력 자료로 다운로드해 주세요. -![입력 시작 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) - - + + + 비디오 생성을 위한 시작 프레임입니다. 이 이미지를 다운로드하여 사용하거나, 자신의 이미지로 교체하세요. + + + 전처리된 포즈 제어 비디오입니다. 이 비디오를 다운로드하여 사용하거나, 자신의 비디오로 교체하세요. + + > 여기서는 사전 처리된 비디오를 사용합니다. @@ -83,19 +83,39 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 아래 모델들은 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인하실 수 있습니다. -**디퓨전 모델** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) +**Diffusion Models** + + + + wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors：고노이즈 확산 모델 + + + wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors：저노이즈 확산 모델 + + **Wan2.2-Lightning LoRA (선택사항, 가속화용)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors：고노이즈 4단계 가속 LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors：저노이즈 4단계 가속 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors：인코딩/디코딩용 Wan2.1 VAE + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors：스케일링 FP8 텍스트 인코더 + ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx index 6a076123b..492c6ef28 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -61,13 +61,14 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> 또는 ComfyUI를 최신 버전으로 업데이트한 후 아래 워크플로우를 다운로드해 ComfyUI에 드래그하여 로드하세요. - -

JSON 워크플로우 다운로드

- - - -

Comfy Cloud에서 실행

- + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Inp" 검색 + + + Comfy Cloud에서 열기 + + 다음 자료를 시작 및 끝 프레임으로 사용하세요. @@ -77,18 +78,38 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> ### 2. 모델 **디퓨전 모델** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) + + + + wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: 시작-끝 프레임 인페인팅용 고노이즈 확산 모델 + + + wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: 시작-끝 프레임 인페인팅용 저노이즈 확산 모델 + + **Lightning LoRA (선택사항, 가속화용)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 고노이즈 모델용 4단계 가속 LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 저노이즈 모델용 4단계 가속 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors: 인코딩 및 디코딩용 Wan2.1 VAE + + +**텍스트 인코더** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: 스케일링된 FP8 텍스트 인코더 + ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-s2v.mdx b/ko/tutorials/video/wan/wan2-2-s2v.mdx index bd4747e65..41cb67f32 100644 --- a/ko/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ko/tutorials/video/wan/wan2-2-s2v.mdx @@ -35,38 +35,58 @@ Wan2.2 S2V 모델: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

JSON 워크플로우 다운로드

- - - -

Comfy Cloud에서 실행

- + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 S2V" 검색 + + 다음 이미지와 오디오를 입력으로 다운로드하세요: -![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - - -

입력 오디오 다운로드

- + + + 기본 입력 이미지를 다운로드하거나, 자신의 이미지를 사용하세요. + + + 기본 입력 오디오를 다운로드하거나, 자신의 오디오를 사용하세요. + + ### 2. 모델 링크 모델들은 [우리 리포지토리](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인하실 수 있습니다. -**diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +**diffusion_models** + + + + FP8 scaled diffusion model. ComfyUI/models/diffusion_models/에 배치 + + + BF16 diffusion model. ComfyUI/models/diffusion_models/에 배치 + + **audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) + + + Audio encoder model. ComfyUI/models/audio_encoders/에 배치 + **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + Wan2.1 VAE model. ComfyUI/models/vae/에 배치 + + +**text_encoders** + + + FP8 scaled text encoder. ComfyUI/models/text_encoders/에 배치 + ``` @@ -94,8 +114,14 @@ ComfyUI/ 두 모델 모두 [여기](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models)에서 확인하실 수 있습니다: -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) + + + FP8 scaled diffusion model + + + BF16 diffusion model + + 이 템플릿에서는 `wan2.2_s2v_14B_fp8_scaled.safetensors`를 사용하며, 이 모델은 더 적은 VRAM을 필요로 합니다. 하지만 품질 저하를 줄이기 위해 `wan2.2_s2v_14B_bf16.safetensors`를 시도해볼 수도 있습니다. diff --git a/ko/tutorials/video/wan/wan2_2.mdx b/ko/tutorials/video/wan/wan2_2.mdx index 1303255f4..ea1a1d460 100644 --- a/ko/tutorials/video/wan/wan2_2.mdx +++ b/ko/tutorials/video/wan/wan2_2.mdx @@ -101,24 +101,25 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - -

JSON 워크플로우 파일 다운로드

- - - -

Comfy Cloud에서 실행

- + + + JSON 워크플로우 파일을 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 5B" 검색 + + + Comfy Cloud에서 열기 + + ### 2. 모델 수동 다운로드 **디퓨전 모델** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) +- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) +- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan2.2_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -156,25 +157,26 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - -

JSON 워크플로우 파일 다운로드

- - - -

Comfy Cloud에서 실행

- + + + JSON 워크플로우 파일을 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 14B T2V" 검색 + + + Comfy Cloud에서 열기 + + ### 2. 모델 수동 다운로드 **디퓨전 모델** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -213,13 +215,14 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - -

JSON 워크플로 파일 다운로드

- - - -

Comfy Cloud에서 실행

- + + + JSON 워크플로우 파일을 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 14B I2V" 검색 + + + Comfy Cloud에서 열기 + + 다음 이미지를 입력으로 사용할 수 있습니다: ![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -227,14 +230,14 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` ### 2. 모델 수동 다운로드 **디퓨전 모델** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) +- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -272,13 +275,14 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - -

JSON 워크플로 다운로드

- - - -

Comfy Cloud에서 실행

- + + + JSON 워크플로우를 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 14B FLF2V" 검색 + + + Comfy Cloud에서 열기 + + 다음 이미지를 입력 자료로 다운로드하세요: diff --git a/ko/tutorials/video/zai/scail2.mdx b/ko/tutorials/video/zai/scail2.mdx index 61c6af2ff..3edc8d8b7 100644 --- a/ko/tutorials/video/zai/scail2.mdx +++ b/ko/tutorials/video/zai/scail2.mdx @@ -105,23 +105,23 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ### 필수 모델 **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders** (하나 선택) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### 파일 저장 위치 diff --git a/tutorials/3d/hunyuan3D-2.mdx b/tutorials/3d/hunyuan3D-2.mdx index ccf839ffb..aca45d0fd 100644 --- a/tutorials/3d/hunyuan3D-2.mdx +++ b/tutorials/3d/hunyuan3D-2.mdx @@ -47,9 +47,14 @@ In the Hunyuan3D-2mv workflow, we'll use multi-view images to generate a 3D mode - -

Run on Comfy Cloud

- + + + Run this workflow instantly on Comfy Cloud + + + Download the workflow JSON file + + ### 1. Workflow @@ -71,7 +76,7 @@ In this example, the input images have already been preprocessed to remove exces Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv.safetensors` +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv.safetensors` ``` ComfyUI/ @@ -94,9 +99,14 @@ If you need to add more views, make sure to load other view images in the `Hunyu In the Hunyuan3D-2mv-turbo workflow, we'll use the Hunyuan3D-2mv-turbo model to generate 3D models. This model is a step distillation version of Hunyuan3D-2mv, allowing for faster 3D model generation. In this version of the workflow, we set `cfg` to 1.0 and add a `flux guidance` node to control the `distilled cfg` generation. - -

Run on Comfy Cloud

- + + + Run this workflow instantly on Comfy Cloud + + + Download the workflow JSON file + + ### 1. Workflow @@ -115,7 +125,7 @@ Download the images below we will use them as input images. Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv-turbo.safetensors` +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv-turbo.safetensors` ``` ComfyUI/ @@ -136,9 +146,14 @@ ComfyUI/ In the Hunyuan3D-2 workflow, we'll use the Hunyuan3D-2 model to generate 3D models. This model is not a multi-view model. In this workflow, we use the `Hunyuan3Dv2Conditioning` node instead of the `Hunyuan3Dv2ConditioningMultiView` node. - -

Run on Comfy Cloud

- + + + Run this workflow instantly on Comfy Cloud + + + Download the workflow JSON file + + ### 1. Workflow @@ -153,7 +168,7 @@ Download the image below we will use it as input image. Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2.safetensors` +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2.safetensors` ``` ComfyUI/ diff --git a/tutorials/flux/flux-1-controlnet.mdx b/tutorials/flux/flux-1-controlnet.mdx index 5d234636f..19d176506 100644 --- a/tutorials/flux/flux-1-controlnet.mdx +++ b/tutorials/flux/flux-1-controlnet.mdx @@ -40,9 +40,14 @@ For image preprocessors, you can use the following custom nodes to complete imag ## FLUX.1-Canny-dev Complete Version Workflow - -

Run on Comfy Cloud

- + + + Download JSON or search "Flux.1 Canny" in Template Library + + + Open in Comfy Cloud + + ### 1. Workflow and Asset @@ -106,9 +111,14 @@ Or use the following custom nodes to complete image preprocessing: ## FLUX.1-Depth-dev-lora Workflow - -

Run on Comfy Cloud

- + + + Download JSON or search "Flux.1 Depth LoRA" in Template Library + + + Open in Comfy Cloud + + The LoRA version workflow builds on the complete version by adding the LoRA model. Compared to the [complete version of the Flux workflow](/tutorials/flux/flux-1-text-to-image), it adds nodes for loading and using the corresponding LoRA model. diff --git a/tutorials/flux/flux-1-fill-dev.mdx b/tutorials/flux/flux-1-fill-dev.mdx index 48c005526..fb4ed5dc9 100644 --- a/tutorials/flux/flux-1-fill-dev.mdx +++ b/tutorials/flux/flux-1-fill-dev.mdx @@ -31,10 +31,10 @@ However, since downloading the corresponding model requires agreeing to the corr ![Flux Agreement](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) Complete model list: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) +|- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) +|- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) +|- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors) +|- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors) File storage location: ``` @@ -53,13 +53,14 @@ ComfyUI/ ### 1. Inpainting workflow and asset - -

Download Workflow Image

- - - -

Run on Comfy Cloud

- + + + Download JSON or search "flux_fill_inpaint" in Template Library + + + Open in Comfy Cloud + + Please download the image below and drag it into ComfyUI to load the corresponding workflow ![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -88,6 +89,15 @@ If you want to draw your own mask, please [click here](https://raw.githubusercon ### 1. Outpainting workflow and asset + + + Download JSON or search "flux_fill_outpaint" in Template Library + + + Open in Comfy Cloud + + + Please download the image below and drag it into ComfyUI to load the corresponding workflow ![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/tutorials/flux/flux-1-kontext-dev.mdx b/tutorials/flux/flux-1-kontext-dev.mdx index 4ca60d5b9..52c562850 100644 --- a/tutorials/flux/flux-1-kontext-dev.mdx +++ b/tutorials/flux/flux-1-kontext-dev.mdx @@ -29,9 +29,9 @@ While the previously released API version offers the highest fidelity and speed, ### Version Information -- **[FLUX.1 Kontext [pro]** - Commercial version, focused on rapid iterative editing -- **FLUX.1 Kontext [max]** - Experimental version with stronger prompt adherence -- **FLUX.1 Kontext [dev]** - Open source version (used in this tutorial), 12B parameters, mainly for research +- **[FLUX.1 Kontext [pro]** — Commercial version, focused on rapid iterative editing +- **FLUX.1 Kontext [max]** — Experimental version with stronger prompt adherence +- **FLUX.1 Kontext [dev]** — Open source version (used in this tutorial), 12B parameters, mainly for research Currently in ComfyUI, you can use all these versions, where [Pro and Max versions](/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) can be called through Partner Nodes, while the Dev open source version please refer to the instructions in this guide. @@ -43,7 +43,7 @@ To run the workflows in this guide successfully, you first need to download the **Diffusion Model** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) If you want to use the original weights, you can visit Black Forest Labs' related repository to obtain and use the original model weights. @@ -54,7 +54,7 @@ If you want to use the original weights, you can visit Black Forest Labs' relate **Text Encoder** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) or [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) or [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) Model save location @@ -72,9 +72,14 @@ Model save location ## Flux.1 Kontext Dev Workflow - -

Run on Comfy Cloud

- + + + Download JSON or search "Flux Kontext Dev" in Template Library + + + Open in Comfy Cloud + + This workflow uses the `Load Image(from output)` node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing. diff --git a/tutorials/flux/flux-1-text-to-image.mdx b/tutorials/flux/flux-1-text-to-image.mdx index 25706e6c0..1d0a0bde0 100644 --- a/tutorials/flux/flux-1-text-to-image.mdx +++ b/tutorials/flux/flux-1-text-to-image.mdx @@ -43,25 +43,30 @@ If you can't download models from [black-forest-labs/FLUX.1-dev](https://hugging #### 1. Workflow File + + + Run this workflow on Comfy Cloud + + + Download JSON or search "Flux.1 Dev" in Template Library + + + Please download the image below and drag it into ComfyUI to load the workflow. ![Flux Dev Original Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - -

Run on Comfy Cloud

- - #### 2. Manual Model Installation - The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) agreement before downloading via browser. -- If your VRAM is low, you can try using [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) to replace the `t5xxl_fp16.safetensors` file. +- If your VRAM is low, you can try using [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) to replace the `t5xxl_fp16.safetensors` file. Please download the following model files: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) Storage location: ``` @@ -96,14 +101,19 @@ Thanks to Flux's excellent prompt following capability, we don't need any negati #### 1. Workflow File + + + Run this workflow on Comfy Cloud + + + Download JSON or search "Flux.1 Schnell" in Template Library + + + Please download the image below and drag it into ComfyUI to load the workflow. ![Flux Schnell Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) - -

Run on Comfy Cloud

- - #### 2. Manual Models Installation @@ -113,10 +123,10 @@ In this workflow, only two model files are different from the Flux1 Dev version Complete model file list: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) File storage location: ``` @@ -150,24 +160,38 @@ but it also requires less VRAM, and you only need to install one model file to t ### Flux.1 Dev + + + Run this workflow on Comfy Cloud + + + Download JSON or search "Flux.1 Dev FP8" in Template Library + + + Please download the image below and drag it into ComfyUI to load the workflow. ![Flux Dev fp8 Checkpoint Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) - -

Run on Comfy Cloud

- - -Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. +Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. Ensure that the corresponding `Load Checkpoint` node loads `flux1-dev-fp8.safetensors`, and you can try to run the workflow. ### Flux.1 Schnell + + + Run this workflow on Comfy Cloud + + + Download JSON or search "Flux.1 Schnell FP8" in Template Library + + + Please download the image below and drag it into ComfyUI to load the workflow. ![Flux Schnell fp8 Checkpoint Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. +Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. Ensure that the corresponding `Load Checkpoint` node loads `flux1-schnell-fp8.safetensors`, and you can try to run the workflow. \ No newline at end of file diff --git a/tutorials/flux/flux-1-uso.mdx b/tutorials/flux/flux-1-uso.mdx index 55cb0dde6..8bcd8201a 100644 --- a/tutorials/flux/flux-1-uso.mdx +++ b/tutorials/flux/flux-1-uso.mdx @@ -29,18 +29,14 @@ Download the image below and drag it into ComfyUI to load the corresponding work ![Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + Download the workflow JSON and drag it into ComfyUI + + + Run this workflow on Comfy Cloud + + Use the image below as an input image. @@ -51,18 +47,18 @@ Use the image below as an input image. **checkpoints** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) **loras** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **model_patches** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **clip_visions** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) Please download all models and place them in the following directories: diff --git a/tutorials/flux/flux1-krea-dev.mdx b/tutorials/flux/flux1-krea-dev.mdx index 6a434e27b..924f47547 100644 --- a/tutorials/flux/flux1-krea-dev.mdx +++ b/tutorials/flux/flux1-krea-dev.mdx @@ -29,13 +29,14 @@ This model is released under the [flux-1-dev-non-commercial-license](https://hug Download the image or JSON below and drag it into ComfyUI to load the corresponding workflow ![Flux Krea Dev Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + Run this workflow on Comfy Cloud + + + Download JSON or search "Flux.1 Krea Dev" in Template Library + + #### 2. Manual Model Installation @@ -46,7 +47,7 @@ Please download the following model files: If you want to pursue higher quality and have enough VRAM, you can try the original model weights -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) agreement before downloading via browser. @@ -55,12 +56,12 @@ The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLU If you have used Flux related workflows before, the following models are the same and don't need to be downloaded again **Text encoders** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) File save location: ``` diff --git a/tutorials/image/hidream/hidream-e1.mdx b/tutorials/image/hidream/hidream-e1.mdx index 7102c23b1..1c6f04768 100644 --- a/tutorials/image/hidream/hidream-e1.mdx +++ b/tutorials/image/hidream/hidream-e1.mdx @@ -32,8 +32,8 @@ This model requires a large amount of VRAM to run. Please refer to the relevant **Diffusion Model** You do not need to download both models. Since E1.1 is an iterative version based on E1, our tests show that its quality and performance are significantly improved compared to E1. -- [hidream_e1_1_bf16.safetensors (Recommended)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors (Recommended)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **Text Encoder**: @@ -66,6 +66,15 @@ Model Save Location ## HiDream E1.1 ComfyUI Native Workflow Example + + + Open in Comfy Cloud + + + Download JSON or search "HiDream E1.1" in Template Library + + + E1.1 is an updated version released on July 16, 2025. This version supports dynamic 1-megapixel resolution, and the workflow uses the `Scale Image to Total Pixels` node to dynamically adjust the input image to 1 million pixels. @@ -109,11 +118,16 @@ Follow these steps to run the workflow: ## HiDream E1 ComfyUI Native Workflow Example - -

Run on Comfy Cloud

- + + + Open in Comfy Cloud + + + Download JSON or search "HiDream E1 Full" in Template Library + + -E1 is a model released on April 28, 2025. This model only supports 768*768 resolution. +E1 is a model released on April 28, 2025. For reference, this workflow takes about 500s for the first run and 370s for the second run with 28 sampling steps on Google Colab L4 with 22.5GB VRAM. diff --git a/tutorials/image/hidream/hidream-i1.mdx b/tutorials/image/hidream/hidream-i1.mdx index 5f6f17afc..685fdaf21 100644 --- a/tutorials/image/hidream/hidream-i1.mdx +++ b/tutorials/image/hidream/hidream-i1.mdx @@ -94,16 +94,21 @@ Model file save location ``` ### HiDream-I1 Full Version Workflow - -

Run on Comfy Cloud

- + + + Run this workflow on Comfy Cloud with zero setup + + + Download the workflow JSON file + + #### 1. Model File Download Please select the appropriate version based on your hardware. Click the link and download the corresponding model file to save it to the `ComfyUI/models/diffusion_models/` folder. -- FP8 version: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) requires more than 27GB of VRAM +- FP8 version: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) requires more than 27GB of VRAM #### 2. Workflow File Download @@ -132,15 +137,20 @@ Complete the workflow execution step by step ### HiDream-I1 Dev Version Workflow - -

Run on Comfy Cloud

- + + + Run this workflow on Comfy Cloud with zero setup + + + Download the workflow JSON file + + #### 1. Model File Download Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -- FP8 version: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) requires more than 27GB of VRAM +- FP8 version: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) requires more than 27GB of VRAM #### 2. Workflow File Download Please download the image below and drag it into ComfyUI to load the corresponding workflow @@ -168,15 +178,20 @@ Complete the workflow execution step by step ### HiDream-I1 Fast Version Workflow - -

Run on Comfy Cloud

- + + + Run this workflow on Comfy Cloud with zero setup + + + Download the workflow JSON file + + #### 1. Model File Download Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM +- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM #### 2. Workflow File Download Please download the image below and drag it into ComfyUI to load the corresponding workflow diff --git a/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 0485471fc..ea45a64bf 100644 --- a/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -22,13 +22,14 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## NewBie-image text-to-image workflow - -

Download JSON Workflow File

- - - -

Run on ComfyUI Cloud

- + + + Download JSON or search "NewBie-image" in Template Library + + + Open in cloud + + diff --git a/tutorials/image/omnigen/omnigen2.mdx b/tutorials/image/omnigen/omnigen2.mdx index 433c0bf33..3a3f8d10c 100644 --- a/tutorials/image/omnigen/omnigen2.mdx +++ b/tutorials/image/omnigen/omnigen2.mdx @@ -57,9 +57,11 @@ File save location: ### 1. Download Workflow File - -

Run on Comfy Cloud

- + + + Open and run this workflow directly in Comfy Cloud. + + ![Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -85,9 +87,11 @@ OmniGen2 has rich image editing capabilities and supports adding text to images ### 1. Download Workflow File - -

Run on Comfy Cloud

- + + + Open and run this workflow directly in Comfy Cloud. + + ![Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/tutorials/image/ovis/ovis-image.mdx b/tutorials/image/ovis/ovis-image.mdx index 779df9892..b65093002 100644 --- a/tutorials/image/ovis/ovis-image.mdx +++ b/tutorials/image/ovis/ovis-image.mdx @@ -20,13 +20,14 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Ovis-Image text-to-image workflow - -

Download JSON Workflow File

- - - -

Run on ComfyUI Cloud

- + + + Open in Comfy Cloud + + + Download JSON or search "Ovis image" in Template Library + + diff --git a/tutorials/image/qwen/qwen-image-2512.mdx b/tutorials/image/qwen/qwen-image-2512.mdx index 86da13449..3fd6f2940 100644 --- a/tutorials/image/qwen/qwen-image-2512.mdx +++ b/tutorials/image/qwen/qwen-image-2512.mdx @@ -36,9 +36,14 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - - Run on Comfy Cloud - + + + Open in Comfy Cloud + + + Download JSON or search "Qwen-Image-2512" in Template Library + + ### 1. Workflow file @@ -48,28 +53,24 @@ The workflow includes two subgraphs: - **Text to Image (Qwen-Image 2512)**: Standard 50-step generation - **Text to Image (Qwen-Image 2512 4steps)**: Accelerated 4-step generation using Lightning LoRA - -

Download JSON Workflow

- - ### 2. Model download **Text Encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (Optional - for 4-step Lightning acceleration)** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **Diffusion Models** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (Recommended for most users) -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (If you have enough VRAM and want better quality) +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (Recommended for most users) +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (If you have enough VRAM and want better quality) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **Model Storage Location** diff --git a/tutorials/image/qwen/qwen-image-edit-2511.mdx b/tutorials/image/qwen/qwen-image-edit-2511.mdx index 730e8b280..906df8a3c 100644 --- a/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -31,31 +31,32 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow below into ComfyUI to load it. - -

Download JSON Workflow

- - - -

Run on ComfyUI Cloud

- + + + Open in Comfy Cloud + + + Download JSON or search "Qwen-Image-Edit-2511" in Template Library + + ### 2. Model download **Text Encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (Optional - for 4-step Lightning acceleration)** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **Diffusion Models** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **Model Storage Location** diff --git a/tutorials/image/qwen/qwen-image-edit.mdx b/tutorials/image/qwen/qwen-image-edit.mdx index 2d9adc583..6cbc89c1f 100644 --- a/tutorials/image/qwen/qwen-image-edit.mdx +++ b/tutorials/image/qwen/qwen-image-edit.mdx @@ -40,13 +40,14 @@ Features include: After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow below into ComfyUI to load it. ![Qwen-image Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

Download JSON Workflow

- - - -

Run on ComfyUI Cloud

- + + + Download JSON or search "image_qwen_image_edit" in Template Library + + + Run this workflow on Cloud GPUs with zero setup + + Download the image below as input ![Qwen-image Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -57,19 +58,19 @@ All models can be found at [Comfy-Org/Qwen-Image_ComfyUI](https://huggingface.co **Diffusion model** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) Model Storage Location diff --git a/tutorials/image/qwen/qwen-image-layered.mdx b/tutorials/image/qwen/qwen-image-layered.mdx index 05275ee64..945c965da 100644 --- a/tutorials/image/qwen/qwen-image-layered.mdx +++ b/tutorials/image/qwen/qwen-image-layered.mdx @@ -21,13 +21,15 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Qwen-Image-Layered workflow - -

Download JSON Workflow File

- - - -

Run on ComfyUI Cloud

- +| +| +| Download the JSON workflow file +| +| +| +| Run ComfyUI online with zero setup +| +| @@ -35,15 +37,15 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +|- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **Model Storage Location** @@ -78,4 +80,4 @@ For input size, 640px is recommended. Use 1024px for high-resolution output. ### Prompt (optional) -The text prompt is intended to describe the overall content of the input image—including elements that may be partially occluded (e.g., you may specify the text hidden behind a foreground object). It is not designed to control the semantic content of individual layers explicitly. +The text prompt is intended to describe the overall content of the input image, including elements that may be partially occluded (e.g., you may specify the text hidden behind a foreground object). It is not designed to control the semantic content of individual layers explicitly. diff --git a/tutorials/image/qwen/qwen-image.mdx b/tutorials/image/qwen/qwen-image.mdx index 636fe2ba4..82e081193 100644 --- a/tutorials/image/qwen/qwen-image.mdx +++ b/tutorials/image/qwen/qwen-image.mdx @@ -48,9 +48,12 @@ Currently Qwen-Image has multiple ControlNet support options available: - - Run on Comfy Cloud - + + + + + + There are three different models used in the workflow attached to this document: 1. Qwen-Image original model fp8_e4m3fn @@ -72,14 +75,9 @@ GPU: RTX4090D 24GB After updating ComfyUI, you can find the workflow file in the templates, or drag the workflow below into ComfyUI to load it. ![Qwen-image Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

Download Workflow for Qwen-Image Official Model

- - Distilled version - -

Download Workflow for Distilled Model

- + + ### 2. Model Download @@ -93,12 +91,12 @@ All models are available at [Huggingface](https://huggingface.co/Comfy-Org/Qwen- **Diffusion model** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - The original author of the distilled version recommends using 15 steps with cfg 1.0. @@ -107,15 +105,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **Model Storage Location** @@ -154,26 +152,25 @@ Qwen_image_distill This is a ControlNet model, so you can use it as normal ControlNet. - - Run on Comfy Cloud - + + + + + + ### 1. Workflow and Input Images Download the image below and drag it into ComfyUI to load the workflow ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

Download JSON Format Workflow

- - Download the image below as input ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) ### 2. Model Links 1. InstantX Controlnet -Download [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) and save it to the `ComfyUI/models/controlnet/` folder +Download [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) and save it to the `ComfyUI/models/controlnet/` folder 2. **Lotus Depth model** @@ -181,11 +178,11 @@ We will use this model to generate the depth map of the image. The following two **Diffusion Model** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAE Model** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) or any SD1.5 VAE +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) or any SD1.5 VAE ``` ComfyUI/ @@ -209,9 +206,12 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow - - Run on Comfy Cloud - + + + + + + This model is actually not a ControlNet, but a Model patch that supports three different control modes: canny, depth, and inpaint. @@ -224,10 +224,6 @@ Comfy Org rehost address: [Qwen-Image-DiffSynth-ControlNets/model_patches](https Download the image below and drag it into ComfyUI to load the corresponding workflow ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

Download JSON Format Workflow

- - Download the image below as input: ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/input.png) @@ -236,9 +232,9 @@ Download the image below as input: Other models are the same as the Qwen-Image basic workflow. You only need to download the models below and save them to the `ComfyUI/models/model_patches` folder -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. Workflow Usage Instructions @@ -280,9 +276,12 @@ For the Inpaint model, it requires using the [Mask Editor](/interface/maskeditor ## Qwen Image Union ControlNet LoRA Workflow - - Run on Comfy Cloud - + + + + + + Original model address: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org rehost address: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): Image structure control LoRA supporting canny, depth, pose, lineart, softedge, normal, openpose @@ -291,9 +290,6 @@ Comfy Org rehost address: [qwen_image_union_diffsynth_lora.safetensors](https:// Download the image below and drag it into ComfyUI to load the workflow ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

Download JSON Format Workflow

- Download the image below as input diff --git a/tutorials/partner-nodes/google/gemini.mdx b/tutorials/partner-nodes/google/gemini.mdx index 292eaf1ad..f404b5245 100644 --- a/tutorials/partner-nodes/google/gemini.mdx +++ b/tutorials/partner-nodes/google/gemini.mdx @@ -20,13 +20,11 @@ In this guide, we will walk you through completing the corresponding conversatio Please download the Json file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- + + + Download Json Format Workflow File + + ### 2. Complete the Workflow Execution Step by Step diff --git a/tutorials/partner-nodes/openai/chat.mdx b/tutorials/partner-nodes/openai/chat.mdx index 074b1fe09..e8e3a0eea 100644 --- a/tutorials/partner-nodes/openai/chat.mdx +++ b/tutorials/partner-nodes/openai/chat.mdx @@ -20,13 +20,9 @@ In this guide, we will walk you through completing the corresponding conversatio Please download the Json file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- +| +| Download the JSON format workflow file. +| ### 2. Complete the Workflow Execution Step by Step diff --git a/tutorials/partner-nodes/rodin/model-generation.mdx b/tutorials/partner-nodes/rodin/model-generation.mdx index 978629e89..0863d781a 100644 --- a/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/tutorials/partner-nodes/rodin/model-generation.mdx @@ -24,13 +24,9 @@ Currently, ComfyUI's Partner nodes support the following Rodin model generation Download the file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- + + Single-view Model Generation (Json Format) + Download the image below as input image @@ -58,13 +54,9 @@ You can modify the single-view workflow to a multi-view workflow, or directly do Download the file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- + + Multi-view Model Generation (Json Format) + Download the images below as input images diff --git a/tutorials/video/bytedance/bernini-r.mdx b/tutorials/video/bytedance/bernini-r.mdx index 2a2c8c699..66a3c5e5b 100644 --- a/tutorials/video/bytedance/bernini-r.mdx +++ b/tutorials/video/bytedance/bernini-r.mdx @@ -1,6 +1,6 @@ --- title: "ComfyUI Bernini-R Examples" -description: "Learn how to use Bernini-R in ComfyUI for image and video editing with in-context conditioning — relighting, restyling, subject insertion, and more." +description: "Learn how to use Bernini-R in ComfyUI for image and video editing with in-context conditioning: relighting, restyling, subject insertion, and more." sidebarTitle: "Bernini-R" --- @@ -8,13 +8,13 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' # ComfyUI Bernini-R Introduction -[Bernini-R](https://github.com/bytedance/Bernini) is ByteDance's **renderer-only** Wan 2.2 model for in-context image and video conditioning. It uses a set of conditioning streams (source video, reference images, reference video) to guide generation — no LoRA training or fine-tuning required. +[Bernini-R](https://github.com/bytedance/Bernini) is ByteDance's **renderer-only** Wan 2.2 model for in-context image and video conditioning. It uses a set of conditioning streams (source video, reference images, reference video) to guide generation. No LoRA training or fine-tuning required. Key capabilities: - **Multiple task types in one model**: image/video generation, editing, relighting, restyling, subject insertion - **In-context conditioning**: reference images/videos act as visual prompts, injected as tokens -- **Lightweight**: renderer-only model — no diffusion-based text-to-video backbone +- **Lightweight**: renderer-only model. No diffusion-based text-to-video backbone. - **Flexible input support**: single or multi-image references, video-to-video, reference-guided editing Bernini-R supports these task types: @@ -39,16 +39,16 @@ ComfyUI now natively supports Bernini-R nodes. Make sure you have updated to the Download the required model weights and save them to the corresponding ComfyUI folders: **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ @@ -87,16 +87,18 @@ ComfyUI/ ### Steps to Run -1. **Select Task Type** — choose your task (Image Editing, Subject to Image, etc.) -2. **Connect Inputs** — load source image and optional reference images -3. **Write Prompt** — describe the desired edit -4. **Run** — click Queue or use `Cmd+Enter` +1. **Select Task Type**: choose your task (Image Editing, Subject to Image, etc.) +2. **Connect Inputs**: load source image and optional reference images +3. **Write Prompt**: describe the desired edit +4. **Run**: click Queue or use `Cmd+Enter` -**Reference Image input:** Use for **Subject to Image** when you need one or more reference images (subject, outfit, scene, props). In the prompt, use `image0`, `image1`, … to reference each image. Not needed for **Image Editing** — that task uses `source_image` instead. +**Reference Image input:** Use for **Subject to Image** when you need one or more reference images (subject, outfit, scene, props). In the prompt, use `image0`, `image1`, … to reference each image. Not needed for **Image Editing**: that task uses `source_image` instead. + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. + --- @@ -117,22 +119,24 @@ ComfyUI/ ### Steps to Run -1. **Load Source Video** — connect your input video -2. **(Optional) Load References** — reference image(s) or reference video -3. **Select Task Type** — v2v, rv2v, r2v, or ads2v -4. **Write Prompt** — describe the desired edit -5. **Run** — click Queue or use `Cmd+Enter` +1. **Load Source Video**: connect your input video +2. **(Optional) Load References**: reference image(s) or reference video +3. **Select Task Type**: v2v, rv2v, r2v, or ads2v +4. **Write Prompt**: describe the desired edit +5. **Run**: click Queue or use `Cmd+Enter` **Reference Image input:** Use when a task needs one or more reference images (rv2v, r2v, multi-piece outfits). Each batched image becomes its own in-context token. Mention `image0`, `image1`, … in the prompt if references play different roles. + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. + --- ## Community Resources -- [Bernini GitHub (bytedance/Bernini)](https://github.com/bytedance/Bernini) — Research paper and task documentation -- [Comfy-Org/Bernini-R](https://huggingface.co/Comfy-Org/Bernini-R) — Official ComfyUI model weights -- [Bernini: Latent Semantic Planning for Video Diffusion](https://arxiv.org/abs/2605.22344) — Research paper +- [Bernini GitHub (bytedance/Bernini)](https://github.com/bytedance/Bernini): Research paper and task documentation +- [Comfy-Org/Bernini-R](https://huggingface.co/Comfy-Org/Bernini-R): Official ComfyUI model weights +- [Bernini: Latent Semantic Planning for Video Diffusion](https://arxiv.org/abs/2605.22344): Research paper diff --git a/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 17b988455..44733df9f 100644 --- a/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -13,17 +13,28 @@ Cosmos-Predict2 supports various generation methods including Text-to-Image (Tex and is widely used in industrial simulation, autonomous driving, urban planning, scientific research, and other fields. It serves as a crucial foundational tool for promoting deep integration of intelligent vision and the physical world. -GitHub:[Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) -huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) + + + Cosmos-Predict2 source code and documentation + + + Cosmos-Predict2 model collection + + This guide will walk you through completing **Video2World** generation in ComfyUI. For the text-to-image section, please refer to the following part: - - Using Cosmos-Predict2 for text-to-image generation - -{/* + + + Using Cosmos-Predict2 for text-to-image generation + + + Run Cosmos-Predict2 workflows on Comfy Cloud with powerful GPUs + + + ## Cosmos Predict2 Video2World Workflow @@ -40,9 +51,14 @@ Please download the video below and drag it into ComfyUI to load the workflow. T src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - -

Download Json Format Workflow File

- + + + Download the JSON format workflow file + + + Run this workflow on Comfy Cloud with pre-installed models + + Please download the following image as input: @@ -50,21 +66,27 @@ Please download the following image as input: ### 2. Manual Model Installation -If the model download wasn't successful, you can try to download them manually by yourself in this section. +If the model download was not successful, you can try to download them manually by yourself in this section. **Diffusion model** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) + + cosmos_predict2_2B_video2world_480p_16fps.safetensors + -For other weights, please visit [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) to download +For other weights, please visit [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) to download. **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) + + oldt5_xxl_fp8_e4m3fn_scaled.safetensors + **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + wan_2.1_vae.safetensors + File Storage Location ``` @@ -91,4 +113,4 @@ Please follow the steps in the image to run the workflow: 6. (Optional) You can modify the prompts in the `ClipTextEncode` node 7. (Optional) Modify the size and frame count in the `CosmosPredict2ImageToVideoLatent` node 8. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -9. Once generation is complete, the video will automatically save to the `ComfyUI/output/` directory, you can also preview it in the `save video` node */} +9. Once generation is complete, the video will automatically save to the `ComfyUI/output/` directory, you can also preview it in the `save video` node diff --git a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 870efd313..a5a2d5802 100644 --- a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -18,39 +18,161 @@ import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; - **Cinematic quality**: Native 720p output (upscalable to 1080p) with professional aesthetics. - **Rich features**: Supports diverse styles (realistic, anime, 3D) and in-video text rendering (Chinese/English). -## Workflow templates +## Common models for all workflows -[video_hunyuan_video_1.5_720p_i2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_i2v.json) +The following models are used in both Text-to-Video and Image-to-Video workflows. Download and save them to the specified directories: -[video_hunyuan_video_1.5_720p_t2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_t2v.json) + + qwen_2.5_vl_7b_fp8_scaled.safetensors — text encoder shared across all workflows + -## Model links -**text_encoders** + + byt5_small_glyphxl_fp16.safetensors — text encoder shared across all workflows + -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) + + hunyuanvideo15_vae_fp16.safetensors — VAE shared across all workflows + -**diffusion_models** +Storage location: -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +``` +ComfyUI/ +├── models/ +│ ├── text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors +│ │ └── byt5_small_glyphxl_fp16.safetensors +│ └── vae/ +│ └── hunyuanvideo15_vae_fp16.safetensors +``` + +## Hunyuan Video 1.5 Text-to-Video Workflow + +HunyuanVideo 1.5 Text-to-Video generates 5-10 second videos from natural language descriptions with enhanced quality and reduced VRAM requirements. + +### 1. Workflow + + + + Open in Comfy Cloud + + + Download JSON or search "Hunyuan Video 1.5 T2V" in Template Library + + + +Download the image below and drag it into ComfyUI to load the workflow: +![ComfyUI Workflow - Hunyuan Video 1.5 T2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_t2v-1.webp) + +### 2. Manual models installation + + + hunyuanvideo1.5_720p_t2v_fp16.safetensors — save to ComfyUI/models/diffusion_models + + +Ensure you have all these model files in the correct locations: + +``` +ComfyUI/ +├── models/ +│ ├── text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // Shared model +│ │ └── byt5_small_glyphxl_fp16.safetensors // Shared model +│ ├── vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors // Shared model +│ └── diffusion_models/ +│ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V model +``` + +### 3. Steps to run the workflow + +1. Ensure the `DualCLIPLoader` node has loaded these models: + - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` + - clip_name2: `byt5_small_glyphxl_fp16.safetensors` +2. Ensure the `Load Diffusion Model` node has loaded `hunyuanvideo1.5_720p_t2v_fp16.safetensors` +3. Ensure the `Load VAE` node has loaded `hunyuanvideo15_vae_fp16.safetensors` +4. Click the `Queue` button or use the shortcut `Ctrl(Cmd) + Enter` to run the workflow + + +The workflow includes a super-resolution upscaler node. When enabled, it uses `hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors` to upscale the output to 1080p. + -**vae** +## Hunyuan Video 1.5 Image-to-Video Workflow -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +HunyuanVideo 1.5 Image-to-Video transforms static images into smooth, high-quality videos with improved consistency and motion dynamics. +### 1. Workflow -Model Storage Location + + + Open in Comfy Cloud + + + Download JSON or search "Hunyuan Video 1.5 I2V" in Template Library + + +Download the image below and drag it into ComfyUI to load the workflow: +![ComfyUI Workflow - Hunyuan Video 1.5 I2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_i2v-1.webp) + +### 2. Related models manual installation + + + sigclip_vision_patch14_384.safetensors — save to ComfyUI/models/clip_vision + + + + hunyuanvideo1.5_720p_i2v_fp16.safetensors — save to ComfyUI/models/diffusion_models + + +Ensure you have all these model files in the correct locations: + +``` +ComfyUI/ +├── models/ +│ ├── clip_vision/ +│ │ └── sigclip_vision_patch14_384.safetensors // I2V vision encoder +│ ├── text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // Shared model +│ │ └── byt5_small_glyphxl_fp16.safetensors // Shared model +│ ├── vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors // Shared model +│ └── diffusion_models/ +│ └── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V model +``` + +### 3. Steps to run the workflow + +1. Ensure the `DualCLIPLoader` node has loaded these models: + - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` + - clip_name2: `byt5_small_glyphxl_fp16.safetensors` +2. Ensure the `CLIPVisionLoader` node has loaded `sigclip_vision_patch14_384.safetensors` +3. Ensure the `Load Diffusion Model` node has loaded `hunyuanvideo1.5_720p_i2v_fp16.safetensors` +4. Ensure the `Load VAE` node has loaded `hunyuanvideo15_vae_fp16.safetensors` +5. Click the `Queue` button or use the shortcut `Ctrl(Cmd) + Enter` to run the workflow + +## Super-resolution upscaler + +Both workflows include a super-resolution node that can upscale the output video from 720p to 1080p. This optional upscaler uses a distilled model for efficient high-resolution output. + + + hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors — save to ComfyUI/models/diffusion_models + + +All shared and workflow-specific models combined: + +``` +ComfyUI/ +├── models/ +│ ├── clip_vision/ +│ │ └── sigclip_vision_patch14_384.safetensors +│ ├── text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors +│ │ └── byt5_small_glyphxl_fp16.safetensors +│ ├── vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors +│ └── diffusion_models/ +│ ├── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V model +│ ├── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V model +│ └── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors // Upscaler ``` -:open_file_folder: ComfyUI/ -├── :open_file_folder: models/ -│ ├── :open_file_folder: text_encoders/ -│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors -│ │ └── byt5_small_glyphxl_fp16.safetensors -│ ├── :open_file_folder: diffusion_models/ -│ │ ├── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors -│ │ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors -│ └── :open_file_folder: vae/ -│ └── hunyuanvideo15_vae_fp16.safetensors -``` \ No newline at end of file diff --git a/tutorials/video/hunyuan/hunyuan-video.mdx b/tutorials/video/hunyuan/hunyuan-video.mdx index 918082b1d..ccefbc727 100644 --- a/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/tutorials/video/hunyuan/hunyuan-video.mdx @@ -35,11 +35,19 @@ Alternatively, this guide provides direct model links if automatic downloads fai ## Common Models for All Workflows -The following models are used in both Text-to-Video and Image-to-Video workflows. Please download and save them to the specified directories: +The following models are used in both Text-to-Video and Image-to-Video workflows. Download and save them to the specified directories: -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) + + clip_l.safetensors — text encoder shared across all workflows + + + + llava_llama3_fp8_scaled.safetensors — text encoder shared across all workflows + + + + hunyuan_video_vae_bf16.safetensors — VAE shared across all workflows + Storage location: @@ -59,12 +67,23 @@ Hunyuan Text-to-Video was open-sourced in December 2024, supporting 5-second sho ### 1. Workflow + + + Open in Comfy Cloud + + + Download JSON or search "Hunyuan Video" in Template Library + + + Download the image below and drag it into ComfyUI to load the workflow: ![ComfyUI Workflow - Hunyuan Text-to-Video](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/t2v/kitchen.webp) ### 2. Manual Models Installation -Download [hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) and save it to the `ComfyUI/models/diffusion_models` folder. + + hunyuan_video_t2v_720p_bf16.safetensors — save to ComfyUI/models/diffusion_models + Ensure you have all these model files in the correct locations: @@ -97,7 +116,7 @@ When the `length` parameter in the `EmptyHunyuanLatentVideo` node is set to 1, t ## Hunyuan Image-to-Video Workflow -Hunyuan Image-to-Video model was open-sourced on March 6, 2025, based on the HunyuanVideo framework. It transforms static images into smooth, high-quality videos and also provides LoRA training code to customize special video effects like hair growth, object transformation, etc. +Hunyuan Image-to-Video model was open-sourced on March 6, 2025, based on the HunyuanVideo framework. It transforms static images into smooth, high-quality videos and also provides LoRA training code to customize special video effects like hair growth, object transformation, and more. Currently, the Hunyuan Image-to-Video model has two versions: - v1 "concat": Better motion fluidity but less adherence to the image guidance @@ -116,8 +135,9 @@ Currently, the Hunyuan Image-to-Video model has two versions: ### Shared Model for v1 and v2 Versions -Download the following file and save it to the `ComfyUI/models/clip_vision` directory: -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) + + llava_llama3_vision.safetensors — save to ComfyUI/models/clip_vision + ### V1 "concat" Image-to-Video Workflow @@ -126,12 +146,14 @@ Download the following file and save it to the `ComfyUI/models/clip_vision` dire Download the workflow image below and drag it into ComfyUI to load the workflow: ![ComfyUI Workflow - Hunyuan Image-to-Video v1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/i2v/v1_robot.webp) -Download the image below, which we'll use as the starting frame for the image-to-video generation: +Download the image below, which we will use as the starting frame for the image-to-video generation: ![Starting Frame](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/hunyuan-video/i2v/robot-ballet.png) -#### 2. Related models manual installation +#### 2. Related Models Manual Installation -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) + + hunyuan_video_image_to_video_720p_bf16.safetensors — save to ComfyUI/models/diffusion_models + Ensure you have all these model files in the correct locations: @@ -171,12 +193,14 @@ The v2 workflow is essentially the same as the v1 workflow. You just need to dow Download the workflow image below and drag it into ComfyUI to load the workflow: ![ComfyUI Workflow - Hunyuan Image-to-Video v2](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/i2v/v2_fennec_gril.webp) -Download the image below, which we'll use as the starting frame for the image-to-video generation: +Download the image below, which we will use as the starting frame for the image-to-video generation: ![Starting Frame](https://comfyanonymous.github.io/ComfyUI_examples/flux/flux_dev_example.png) -#### 2. Related models manual installation +#### 2. Related Models Manual Installation -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) + + hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors — save to ComfyUI/models/diffusion_models + Ensure you have all these model files in the correct locations: @@ -207,9 +231,9 @@ ComfyUI/ 5. Ensure the `Load Diffusion Model` node has loaded `hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors` 6. Click the `Queue` button or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -## Try it yourself +## Try It Yourself -Here are some images and prompts we provide. Based on that content or make an adjustment to create your own video. +Here are some images and prompts we provide. Use them as-is or adjust to create your own video. ![example](/images/tutorial/advanced/hunyuanvideo/humanoid_android_dressed_in_a_flowing.png) @@ -240,4 +264,4 @@ flying car fastly moving and flying through the city ``` cyberpunk car race in night city, dynamic, super fast, fast shot -``` \ No newline at end of file +``` diff --git a/tutorials/video/kandinsky/kandinsky-5.mdx b/tutorials/video/kandinsky/kandinsky-5.mdx index 06615ffa5..c6a91682b 100644 --- a/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/tutorials/video/kandinsky/kandinsky-5.mdx @@ -41,21 +41,39 @@ All models are available in 5-second and 10-second video generation versions. Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 T2V" to load the workflow. - -

Download JSON Workflow File

- + + + Download the T2V workflow to use locally + + + Open in Comfy Cloud + + ### 2. Manually download models **Text Encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B text encoder (FP8) + + + CLIP-L text encoder + + **Diffusion Model** -- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) + + + Kandinsky 5.0 T2V Lite SFT diffusion model (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ @@ -75,21 +93,39 @@ ComfyUI/ Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 I2V" to load the workflow. - -

Download JSON Workflow File

- + + + Download the I2V workflow to use locally + + + Open in Comfy Cloud + + ### 2. Manually download models **Text Encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B text encoder (FP8) + + + CLIP-L text encoder + + **Diffusion Model** -- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) + + + Kandinsky 5.0 I2V Lite diffusion model (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ diff --git a/tutorials/video/ltx/ltx-2.mdx b/tutorials/video/ltx/ltx-2.mdx index 5aad54e81..05778e3c8 100644 --- a/tutorials/video/ltx/ltx-2.mdx +++ b/tutorials/video/ltx/ltx-2.mdx @@ -45,16 +45,17 @@ LTX-2 is natively supported in ComfyUI. To get started: Generate videos from text prompts. - + + Download workflow - - + Open in cloud + **Distilled version** (faster, 8 steps): - + Download workflow @@ -62,16 +63,17 @@ Generate videos from text prompts. Generate videos from an input image. - + + Download workflow - - + Open in cloud + **Distilled version** (faster, 8 steps): - + Download workflow @@ -80,31 +82,34 @@ Generate videos from an input image. Generate videos with structural control using IC-LoRAs. **Depth control:** - + + Download workflow - - + Open in cloud + **Canny control:** - + + Download workflow - - + Open in cloud + **Pose control:** - + + Download workflow - - + Open in cloud + ## Prompting tips diff --git a/tutorials/video/ltxv.mdx b/tutorials/video/ltxv.mdx index 0bc4af1a1..7d8f75de8 100644 --- a/tutorials/video/ltxv.mdx +++ b/tutorials/video/ltxv.mdx @@ -24,9 +24,17 @@ Drag the video directly into ComfyUI to run the workflow. Allows you to control the video with a first [frame image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png). - -

Run on Comfy Cloud

- + + + Open in Comfy Cloud + + + Download JSON or search "LTX-Video" in Template Library + + + Get the example input image for this workflow + + LTX-Video Image to Video @@ -36,6 +44,15 @@ Drag the video directly into ComfyUI to run the workflow. ## Text to Video + + + Open in Comfy Cloud + + + Download JSON or search "LTX-Video" in Template Library + + + LTX-Video Text to Video @@ -46,8 +63,13 @@ Drag the video directly into ComfyUI to run the workflow. Download the following models and place them in the locations specified below: -- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true) + + Download and place in ComfyUI/models/checkpoints/ + + + + Download and place in ComfyUI/models/text_encoders/ + ``` ├── checkpoints/ diff --git a/tutorials/video/wan/fun-camera.mdx b/tutorials/video/wan/fun-camera.mdx index a99670e54..3ff0af386 100644 --- a/tutorials/video/wan/fun-camera.mdx +++ b/tutorials/video/wan/fun-camera.mdx @@ -26,21 +26,49 @@ These models only need to be installed once. Additionally, model download inform All of the following models can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) -**Diffusion Models** choose either 1.3B or 14B: -- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) -- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) +### Diffusion Models + +Choose either 1.3B or 14B: + + + + Wan2.1 Fun Camera 1.3B diffusion model + + + Wan2.1 Fun Camera 14B diffusion model + + If you've used Wan2.1 related models before, you should already have the following models. If not, please download them: -**Text Encoders** choose one: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +### Text Encoders + +Choose one of the following: + + + + Full precision text encoder + + + FP8 quantized text encoder (recommended for lower VRAM) + + + +### VAE -**VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + + Wan2.1 VAE model + + -**CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +### CLIP Vision + + + + CLIP vision encoder + + File Storage Location: @@ -58,11 +86,18 @@ File Storage Location: │ └── clip_vision_h.safetensors ``` -## ComfyUI Wan2.1 Fun Camera 1.3B Native Workflow Example +## ComfyUI Wan2.1 Fun Camera 1.3B Workflow -### 1. Workflow Related Files Download +### 1. Download Workflow -#### 1.1 Workflow File + + + Open in Comfy Cloud + + + Download JSON or search "Wan 2.1 Fun Camera 1.3B" in Template Library + + Download the video below and drag it into ComfyUI to load the corresponding workflow: @@ -72,21 +107,19 @@ Download the video below and drag it into ComfyUI to load the corresponding work src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > - -

Download Json Workflow File

- - If you want to use the 14B version, simply replace the model file with the 14B version, but please be aware of the VRAM requirements. -#### 1.2 Input Image Download - -Please download the image below, which we will use as the starting frame: +### 2. Download Input Material -![Input Reference Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) + + + Download the image below to use as the starting frame for the 1.3B workflow + + -### 2. Complete the Workflow Step by Step +### 3. Complete the Workflow Step by Step ![Wan2.1 Fun Camera Workflow Steps](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -101,7 +134,18 @@ Please download the image below, which we will use as the starting frame: 7. Set camera motion in the `WanCameraEmbedding` node 8. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to execute generation -## ComfyUI Wan2.1 Fun Camera 14B Workflow and Input Image +## ComfyUI Wan2.1 Fun Camera 14B Workflow + +### 1. Download Workflow + + + + Open in Comfy Cloud + + + Download JSON or search "Wan 2.1 Fun Camera 14B" in Template Library + + - -

Download Json Workflow File

- +### 2. Download Input Material -**Input Image** -![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) + + + Download the image below to use as the starting frame for the 14B workflow + + ## Performance Reference **1.3B Version**: -- 512×512 resolution on RTX 4090 takes about 72 seconds to generate 81 frames +- 512x512 resolution on RTX 4090 takes about 72 seconds to generate 81 frames **14B Version**: -- RTX4090 24GB VRAM may experience insufficient memory when generating 512×512 resolution, and memory issues have also occurred on A100 when using larger sizes +- RTX4090 24GB VRAM may experience insufficient memory when generating 512x512 resolution, and memory issues have also occurred on A100 when using larger sizes diff --git a/tutorials/video/wan/fun-control.mdx b/tutorials/video/wan/fun-control.mdx index 5fbea31f4..f6dbbc9fb 100644 --- a/tutorials/video/wan/fun-control.mdx +++ b/tutorials/video/wan/fun-control.mdx @@ -25,13 +25,13 @@ Here are the relevant code repositories: ComfyUI now **natively supports** the Wan2.1 Fun Control model. Before starting this tutorial, please update your ComfyUI to ensure you're using a version after [this commit](https://github.com/Comfy-Org/ComfyUI/commit/3661c833bcc41b788a7c9f0e7bc48524f8ee5f82). -In this guide, we'll provide two workflows: +In this guide, we will provide two workflows: 1. A workflow using only native Comfy Core nodes 2. A workflow using custom nodes Due to current limitations in native nodes for video support, the native-only workflow ensures users can complete the process without installing custom nodes. -However, we've found that providing a good user experience for video generation is challenging without custom nodes, so we're providing both workflow versions in this guide. +However, we have found that providing a good user experience for video generation is challenging without custom nodes, so we are providing both workflow versions in this guide. @@ -42,21 +42,41 @@ You only need to install these models once. The workflow images also contain mod The following models can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) and [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) -Click the corresponding links to download. If you've used Wan-related workflows before, you only need to download the **Diffusion models**. +Click the corresponding links to download. If you have used Wan-related workflows before, you only need to download the **Diffusion models**. -**Diffusion models** - choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true): Rename to `Wan2.1-Fun-14B-Control.safetensors` after downloading +**Diffusion models** -- choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: -**Text encoders** - choose one of the following models (fp16 precision has a larger size and higher performance requirements): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) + + + Wan2.1 Fun Control 1.3B diffusion model -- lightweight, lower VRAM requirements + + + Wan2.1 Fun Control 14B diffusion model. Rename to Wan2.1-Fun-14B-Control.safetensors after downloading. 32GB+ file size, higher VRAM requirements + + + +**Text encoders** -- choose one of the following models (fp16 precision has a larger size and higher performance requirements): + + + + UMT5 XXL text encoder -- fp16 precision, larger size, higher performance + + + UMT5 XXL text encoder -- fp8 precision, smaller size, lower performance requirements + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) -**CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) + + Wan2.1 VAE model + + +**CLIP Vision** + + + CLIP Vision model for processing reference images + File storage location: ``` @@ -74,26 +94,32 @@ File storage location: ## ComfyUI Native Workflow -In this workflow, we use videos converted to **WebP format** since the `Load Image` node doesn't currently support mp4 format. We also use **Canny Edge** to preprocess the original video. +In this workflow, we use videos converted to **WebP format** since the `Load Image` node does not currently support mp4 format. We also use **Canny Edge** to preprocess the original video. Because many users encounter installation failures and environment issues when installing custom nodes, this version of the workflow uses only native nodes to ensure a smoother experience. Thanks to our powerful ComfyUI authors who provide feature-rich nodes. If you want to directly check the related version, see [Workflow Using Custom Nodes](#workflow-using-custom-nodes). ### 1. Workflow File Download -#### 1.1 Workflow File - -Download the image below and drag it into ComfyUI to load the workflow: - -![Wan2.1 Fun Control Native Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/wan2.1_fun_control_native.webp) - -#### 1.2 Input Images and Videos Download - -Please download the following image and video for input: - -![Input Reference Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/input/01-portrait_remix.png) - -![Input Reference Video](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/input/01-portrait_video.webp) + + + Open in Comfy Cloud + + + Download the workflow image and drag it into ComfyUI to load the workflow + + + +#### Input Materials + + + + Input reference image for the native workflow. Download and use this image, or replace with your own. + + + Input control video in WebP format for the native workflow. The native workflow requires WebP format since Load Image does not support mp4. + + ### 2. Complete the Workflow Step by Step @@ -104,21 +130,21 @@ Please download the following image and video for input: 3. Ensure the `Load VAE` node has loaded `wan_2.1_vae.safetensors` 4. Ensure the `Load CLIP Vision` node has loaded `clip_vision_h.safetensors` 5. Upload the starting frame to the `Load Image` node (renamed to `Start_image`) -6. Upload the control video to the second `Load Image` node. Note: This node currently doesn't support mp4, only WebP videos +6. Upload the control video to the second `Load Image` node. Note: This node currently does not support mp4, only WebP videos 7. (Optional) Modify the prompt (both English and Chinese are supported) 8. (Optional) Adjust the video size in `WanFunControlToVideo`, avoiding overly large dimensions 9. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to execute video generation ### 3. Usage Notes -- Since we need to input the same number of frames as the control video into the `WanFunControlToVideo` node, if the specified frame count exceeds the actual control video frames, the excess frames may display scenes not conforming to control conditions. We'll address this issue in the [Workflow Using Custom Nodes](#workflow-using-custom-nodes) +- Since we need to input the same number of frames as the control video into the `WanFunControlToVideo` node, if the specified frame count exceeds the actual control video frames, the excess frames may display scenes not conforming to control conditions. We will address this issue in the [Workflow Using Custom Nodes](#workflow-using-custom-nodes) - Avoid setting overly large dimensions, as this can make the sampling process very time-consuming. Try generating smaller images first, then upscale - Use your imagination to build upon this workflow by adding text-to-image or other types of workflows to achieve direct text-to-video generation or style transfer - Use tools like [ComfyUI-comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux) for richer control options ## Workflow Using Custom Nodes -We'll need to install the following two custom nodes: +We will need to install the following two custom nodes: - [ComfyUI-VideoHelperSuite](https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite) - [ComfyUI-comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux) @@ -127,31 +153,31 @@ You can use [ComfyUI Manager](https://github.com/Comfy-Org/ComfyUI-Manager) to i ### 1. Workflow File Download -#### 1.1 Workflow File - -Download the image below and drag it into ComfyUI to load the workflow: - -![Workflow File](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/wan2.1_fun_control_use_custom_nodes.webp) - - -Due to the large size of video files, you can also click [here](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/wan2.1_fun_control_use_custom_nodes.json) to download the workflow file in JSON format. - - -#### 1.2 Input Images and Videos Download -Please download the following image and video for input: -![Input Reference Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/input/02-robot's_eye.png) - - + + + Open in Comfy Cloud + + + Download JSON or search "Wan 2.1 ControlNet" in Template Library + + + +#### Input Materials + + + + Input reference image for the custom nodes workflow. Download and use this image, or replace with your own. + + + Input control video in MP4 format for the custom nodes workflow. The custom Load Video node supports mp4 format. + + ### 2. Complete the Workflow Step by Step ![Wan2.1 Fun Control Workflow Using Custom Nodes Steps](/images/tutorial/video/wan/fun_control_using_custom_nodes_flow_diagram.png) -> The model part is essentially the same. If you've already experienced the native-only workflow, you can directly upload the corresponding images and run it. +> The model part is essentially the same. If you have already experienced the native-only workflow, you can directly upload the corresponding images and run it. 1. Ensure the `Load Diffusion Model` node has loaded `wan2.1_fun_control_1.3B_bf16.safetensors` 2. Ensure the `Load CLIP` node has loaded `umt5_xxl_fp8_e4m3fn_scaled.safetensors` @@ -178,7 +204,7 @@ Thanks to the ComfyUI community authors for their custom node packages: - A useful tip is that you can combine multiple image preprocessing techniques and then use the `Image Blend` node to achieve the goal of applying multiple control methods simultaneously. - You can use the `Video Combine` node from `ComfyUI-VideoHelperSuite` to save videos in mp4 format -- We use `SaveAnimatedWEBP` because we currently don't support embedding workflow into **mp4** and some other custom nodes may not support embedding workflow too. To preserve the workflow in the video, we choose `SaveAnimatedWEBP` node. +- We use `SaveAnimatedWEBP` because we currently do not support embedding workflow into **mp4** and some other custom nodes may not support embedding workflow too. To preserve the workflow in the video, we choose the `SaveAnimatedWEBP` node. - In the `WanFunControlToVideo` node, `control_video` is not mandatory, so sometimes you can skip using a control video, first generate a very small video size like 320x320, and then use them as control video input to achieve consistent results. - [ComfyUI-WanVideoWrapper](https://github.com/kijai/ComfyUI-WanVideoWrapper) diff --git a/tutorials/video/wan/fun-inp.mdx b/tutorials/video/wan/fun-inp.mdx index 3d7d22cc2..655087099 100644 --- a/tutorials/video/wan/fun-inp.mdx +++ b/tutorials/video/wan/fun-inp.mdx @@ -12,7 +12,7 @@ import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; **Key features**: - **First and last frame control**: Supports inputting both first and last frame images to generate transitional video between them, enhancing video coherence and creative freedom. Compared to earlier community versions, Alibaba's official model produces more stable and significantly higher quality results. -- **Multi-resolution support**: Supports generating videos at 512×512, 768×768, 1024×1024 and other resolutions to accommodate different scenario requirements. +- **Multi-resolution support**: Supports generating videos at 512x512, 768x768, 1024x1024 and other resolutions to accommodate different scenario requirements. **Model versions**: - **1.3B** Lightweight: Suitable for local deployment and quick inference with **lower VRAM requirements** @@ -32,31 +32,69 @@ Currently, ComfyUI natively supports the Wan2.1 Fun InP model. Before starting t ## Wan2.1 Fun InP Workflow +### 1. Download Workflow + + + + Open in Comfy Cloud + + + Download JSON or search "Wan 2.1 Inpainting" in Template Library + + + Download the image below and drag it into ComfyUI to load the workflow: ![Workflow File](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_inp/wan2.1_fun_inp.webp) -### 1. Workflow File Download + + + Starting frame for the video generation. Download and use this image, or replace with your own. + + + Ending frame for the video generation. Download and use this image, or replace with your own. + + ### 2. Manual Model Installation If automatic model downloading is ineffective, please download the models manually and save them to the corresponding folders. -The following models can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) and [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) +All models involved in this guide can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) and [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334). + +**Diffusion Models** — Choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: -**Diffusion models** - choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): Rename to `Wan2.1-Fun-14B-InP.safetensors` after downloading + + + Wan2.1 Fun InP 1.3B diffusion model — lower VRAM requirements + + + Wan2.1 Fun InP 14B diffusion model — rename to Wan2.1-Fun-14B-InP.safetensors after downloading + + -**Text encoders** - choose one of the following models (fp16 precision has a larger size and higher performance requirements): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +**Text Encoders** — Choose one of the following models (fp16 precision has a larger size and higher performance requirements): + + + + Full precision text encoder + + + FP8 quantized text encoder (recommended for lower VRAM) + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) -**CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) + + Wan2.1 VAE for encoding/decoding + + +**CLIP Vision** + + + CLIP vision encoder + File storage location: ``` @@ -89,7 +127,7 @@ File storage location: ### 4. Workflow Notes -Please make sure to use the correct model, as `wan2.1_fun_inp_1.3B_bf16.safetensors` and `wan2.1_fun_control_1.3B_bf16.safetensors` are stored in the same folder and have very similar names. Ensure you're using the right model. +Please make sure to use the correct model, as `wan2.1_fun_inp_1.3B_bf16.safetensors` and `wan2.1_fun_control_1.3B_bf16.safetensors` are stored in the same folder and have very similar names. Ensure you are using the right model. - When using Wan Fun InP, you may need to frequently modify prompts to ensure the accuracy of the corresponding scene transitions. @@ -98,4 +136,4 @@ Please make sure to use the correct model, as `wan2.1_fun_inp_1.3B_bf16.safetens - [ComfyUI-VideoHelperSuite](https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite) - [ComfyUI-WanVideoWrapper](https://github.com/kijai/ComfyUI-WanVideoWrapper) -- [ComfyUI-KJNodes](https://github.com/kijai/ComfyUI-KJNodes) \ No newline at end of file +- [ComfyUI-KJNodes](https://github.com/kijai/ComfyUI-KJNodes) diff --git a/tutorials/video/wan/vace.mdx b/tutorials/video/wan/vace.mdx index 659bdb0b1..6f74256f6 100644 --- a/tutorials/video/wan/vace.mdx +++ b/tutorials/video/wan/vace.mdx @@ -8,7 +8,7 @@ import CancelBypass from '/snippets/interface/cancel-bypass.mdx' import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; -As we have made adjustments to the template and added related usage and instructions for CausVid LoRA, this document needs to be updated and requires some preparation time. Until then, please refer to the notes in the template for usage +As we have made adjustments to the template and added related usage and instructions for CausVid LoRA, this document needs to be updated and requires some preparation time. Until then, please refer to the notes in the template for usage. ## About VACE @@ -19,14 +19,14 @@ The model is open-sourced under the [Apache-2.0](https://github.com/ali-vilab/VA Here is a comprehensive analysis of its core features and technical highlights: -- Multi-modal input: Supports multiple input forms including text, images, video, masks, and control signals -- Unified architecture: Single model supports multiple tasks with freely combinable functions -- Motion transfer: Generates coherent actions based on reference videos -- Local replacement: Replaces specific areas in videos through masks -- Video extension: Completes actions or extends backgrounds -- Background replacement: Preserves subjects while changing environmental backgrounds +- Multi-modal input: supports multiple input forms including text, images, video, masks, and control signals +- Unified architecture: single model supports multiple tasks with freely combinable functions +- Motion transfer: generates coherent actions based on reference videos +- Local replacement: replaces specific areas in videos through masks +- Video extension: completes actions or extends backgrounds +- Background replacement: preserves subjects while changing environmental backgrounds -Currently VACE has released two versions - 1.3B and 14B. Compared to the 1.3B version, the 14B version supports 720P resolution output with better image details and stability. +Currently VACE has released two versions: 1.3B and 14B. Compared to the 1.3B version, the 14B version supports 720P resolution output with better image details and stability. | Model | 480P | 720P | | ----------------------------------------------------------- | ---- | ---- | @@ -35,34 +35,66 @@ Currently VACE has released two versions - 1.3B and 14B. Compared to the 1.3B ve Related model weights and code repositories: -- [VACE-1.3B](https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B) -- [VACE-14B](https://huggingface.co/Wan-AI/Wan2.1-VACE-14B) -- [Github](https://github.com/ali-vilab/VACE) -- [VACE Project Homepage](https://ali-vilab.github.io/VACE-Page/) + + + 1.3B model weights on Hugging Face + + + 14B model weights on Hugging Face + + + VACE source code and documentation + + + Official project homepage with examples + + ## Model Download and Loading in Workflows -Since the workflows covered in this document all use the same workflow template, we can first complete the model download and loading information introduction, then enable/disable different inputs through Bypassing different nodes to achieve different workflows. +Since the workflows covered in this document all use the same workflow template, we can first complete the model download and loading information introduction, then enable or disable different inputs through bypassing different nodes to achieve different workflows. The model download information is already embedded in the workflow information in specific examples, so you can also complete the model download when downloading specific example workflows. ### Model Download -**diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) + +**Diffusion Models** + + + + VACE 14B diffusion model (recommended, ~32GB) + + + VACE 1.3B diffusion model (lighter, ~7GB) + + + If you have used Wan Video related workflows before, you have already downloaded the following model files. **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) -Choose one version from **Text encoders** to download -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) + + + Wan VAE model + + + +**Text Encoders** (choose one version) + + + + Full precision text encoder (higher quality, larger size) + + + FP8 text encoder (lower VRAM usage) + + + +File save location: -File save location ``` 📂 ComfyUI/ ├── 📂 models/ @@ -76,7 +108,7 @@ File save location ### Model Loading -Since the models used in the workflows covered in this document are consistent, the workflows are also the same, and only the nodes are bypassed to enable/disable different inputs, please refer to the following image to ensure that the corresponding models are correctly loaded in different workflows. +Since the models used in the workflows covered in this document are consistent, the workflows are also the same, and only the nodes are bypassed to enable or disable different inputs. Please refer to the following image to ensure that the corresponding models are correctly loaded in different workflows. ![Wan2.1 VACE Model Loading](/images/tutorial/video/wan/wan-vace-model-loading.jpg) @@ -86,16 +118,20 @@ Since the models used in the workflows covered in this document are consistent, -## VACE Text-to-Video Workflow +--- - -If you cannot load the workflow from mp4 file, please ensure that your ComfyUI front-end version is up to date version in [requirements.txt](https://github.com/Comfy-Org/ComfyUI/blob/master/requirements.txt) , make sure you can load the workflow from mp4 file. +## 1. VACE Text-to-Video -Currently 1.19.9 is the latest ComfyUI front-end version in the requirements.txt file. - +**What it does:** Transform text descriptions into high-quality videos using the Wan2.1 VACE 14B model. Supports both 480p and 720p resolution. -### 1. Workflow Download -Download the video below and drag it into ComfyUI to load the corresponding workflow + + + Download JSON or search "Wan2.1 VACE Text to Video" in Template Library + + + Open in Comfy Cloud + + -### 2. Complete the Workflow Step by Step + +If you cannot load the workflow from mp4 file, please ensure that your ComfyUI front-end version is up to date version in [requirements.txt](https://github.com/Comfy-Org/ComfyUI/blob/master/requirements.txt), make sure you can load the workflow from mp4 file. + +Currently 1.19.9 is the latest ComfyUI front-end version in the requirements.txt file. + + +### Steps to Run ![image](/images/tutorial/video/wan/wan-vace-t2v-step-guide.jpg) -Please follow the numbered steps in the image to ensure smooth workflow execution +Please follow the numbered steps in the image to ensure smooth workflow execution: 1. Enter positive prompts in the `CLIP Text Encode (Positive Prompt)` node 2. Enter negative prompts in the `CLIP Text Encode (Negative Prompt)` node @@ -123,12 +165,22 @@ During testing with a 4090 GPU: However, 720P video quality is better. -## VACE Image-to-Video Workflow +--- + +## 2. VACE Image-to-Video -You can continue using the workflow above, just unbypass the `Load image` node in **Load reference image** and input your image. You can also use the image below - in this file we've already set up the corresponding parameters. +**What it does:** Generate videos that match the style and content of a reference image. Perfect for style-consistent video generation. -### 1. Workflow Download -Download the video below and drag it into ComfyUI to load the corresponding workflow +You can continue using the workflow above, just unbypass the `Load image` node in **Load reference image** and input your image. You can also use the image below. In this file we have already set up the corresponding parameters. + + + + Download JSON or search "Wan2.1 VACE Reference to Video" in Template Library + + + Open in Comfy Cloud + + -Please download the image below as input +Please download the image below as input: ![vace-i2v-input](https://github.com/Comfy-Org/example_workflows/raw/refs/heads/main/video/wan/vace/i2v/input.jpg) -### 2. Complete the Workflow Step by Step +### Steps to Run ![Workflow Steps](/images/tutorial/video/wan/wan-vace-i2v-step-guide.jpg) -Please follow the numbered steps in the image to ensure smooth workflow execution +Please follow the numbered steps in the image to ensure smooth workflow execution: 1. Input the corresponding image in the `Load image` node 2. You can modify and edit prompts like in the text-to-video workflow @@ -156,16 +208,25 @@ Please follow the numbered steps in the image to ensure smooth workflow executio You may want to use nodes like getting image dimensions to set the resolution, but due to width and height step requirements of the corresponding nodes, you may get error messages if your image dimensions are not divisible by 16. -### 3. Additional Workflow Notes +### Additional Notes -VACE also supports inputting multiple reference images in a single image to generate corresponding videos. You can see related examples on the VACE project [page](https://ali-vilab.github.io/VACE-Page/) - +VACE also supports inputting multiple reference images in a single image to generate corresponding videos. You can see related examples on the VACE project [page](https://ali-vilab.github.io/VACE-Page/). -## VACE Video-to-Video Workflow +--- -### 1. Workflow Download +## 3. VACE Video-to-Video + +**What it does:** Generate videos by controlling input videos and reference images using Wan VACE. Control video style and motion through reference materials. + + + + Download JSON or search "Wan2.1 VACE Control Video" in Template Library + + + Open in Comfy Cloud + + -Download the video below and drag it into ComfyUI to load the corresponding workflow -3. The video below is the original video. You can download these materials and use preprocessing nodes like [comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux) to preprocess the images +3. The video below is the original video. You can download these materials and use preprocessing nodes like [comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux) to preprocess the images: - -### 2. Complete the Workflow Step by Step +### Steps to Run ![Workflow Steps](/images/tutorial/video/wan/wan-vace-v2v-step-guide.jpg) -Please follow the numbered steps in the image to ensure smooth workflow execution +Please follow the numbered steps in the image to ensure smooth workflow execution: -1. Input the reference image in the `Load Image` node under `Load reference image` -2. Input the control video in the `Load Video` node under `Load control video`. Since the provided video is preprocessed, no additional processing is needed -3. If you need to preprocess the original video yourself, you can modify the `Image preprocessing` group or use `comfyui_controlnet_aux` nodes to complete the preprocessing +1. Input the reference image in the `Load Image` node under **Load reference image** +2. Input the control video in the `Load Video` node under **Load control video**. Since the provided video is preprocessed, no additional processing is needed +3. If you need to preprocess the original video yourself, you can modify the **Image preprocessing** group or use `comfyui_controlnet_aux` nodes to complete the preprocessing 4. Modify prompts 5. Set the image dimensions (640x640 resolution recommended for first run) and frame count (video duration) in `WanVaceToVideo` 6. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to execute video generation 7. Once generated, the video will automatically save to `ComfyUI/output/video` directory (subfolder location depends on `save video` node settings) -## VACE Video Outpainting Workflow +--- + +## 4. VACE Inpainting + +**What it does:** Edit specific regions in videos while preserving surrounding content. Great for object removal or replacement. + + + + Download JSON or search "Wan2.1 VACE Inpainting" in Template Library + + + Open in Comfy Cloud + + [To be updated] +--- + +## 5. VACE Video Outpainting + +**What it does:** Generate extended videos by expanding video size using Wan VACE outpainting. -## VACE First-Last Frame Video Generation + + + Download JSON or search "Wan2.1 VACE Outpainting" in Template Library + + + Open in Comfy Cloud + + [To be updated] -To ensure that the first and last frames are effective, the video `length` setting must satisfy that `length-1` is divisible by 4. +--- + +## 6. VACE First-Last Frame Video Generation -The corresponding `Batch_size` setting must satisfy `Batch_size = length - 2` +**What it does:** Generate smooth video transitions by defining start and end frames. Supports custom keyframe sequences. + + + Download JSON or search "Wan2.1 VACE First-Last Frame" in Template Library + + + Open in Comfy Cloud + + +To ensure that the first and last frames are effective, the video `length` setting must satisfy that `length-1` is divisible by 4. + +The corresponding `Batch_size` setting must satisfy `Batch_size = length - 2`. + +[To be updated] diff --git a/tutorials/video/wan/wan-alpha.mdx b/tutorials/video/wan/wan-alpha.mdx index 7d39b19c6..0a2d6c9b7 100644 --- a/tutorials/video/wan/wan-alpha.mdx +++ b/tutorials/video/wan/wan-alpha.mdx @@ -4,6 +4,8 @@ description: "Learn how to generate videos with alpha channel transparency using sidebarTitle: "Wan-Alpha" --- +import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; + Wan-Alpha is a specialized text-to-video model that generates high-quality videos with alpha channel transparency. Built on the Wan2.1-14B-T2V base model, it creates videos with transparent backgrounds and semi-transparent objects, perfect for compositing workflows. The model excels at generating transparent backgrounds, semi-transparent objects (bubbles, glass, water), glowing effects, and fine details with proper alpha channels (hair, smoke, particles). @@ -20,11 +22,101 @@ The model excels at generating transparent backgrounds, semi-transparent objects allowfullscreen > -[Download workflow](https://github.com/Comfy-Org/workflows/blob/main/tutorial_workflows/Get_Comfy_With_Comfy_Wan_Alpha.json) - ## Resources - [Wan-Alpha GitHub](https://github.com/WeChatCV/Wan-Alpha) - [Hugging Face Model](https://huggingface.co/htdong/Wan-Alpha) - [ComfyUI Version](https://huggingface.co/htdong/Wan-Alpha_ComfyUI) - [Research Paper](https://arxiv.org/pdf/2509.24979) + + + +## Wan-Alpha Text-to-Video Workflow (14B) + +### 1. Download the Workflow + +Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan 2.1 Alpha T2V 14B" in the Template Library under `Workflow` → `Browse Templates` → `Video`. + + + + Open in Comfy Cloud + + + Download JSON or search "Wan 2.1 Alpha T2V 14B" in Template Library + + + +### 2. Install Models + +All models mentioned can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged). + +**Diffusion Models** — Choose one version: + + + + FP8 scaled diffusion model. Place in ComfyUI/models/diffusion_models/ + + + BF16 diffusion model. Place in ComfyUI/models/diffusion_models/ + + + +**Text Encoder** + + + + FP8 text encoder. Place in ComfyUI/models/text_encoders/ + + + +**VAE Models** + + + + RGB channel VAE. Place in ComfyUI/models/vae/ + + + Alpha channel VAE. Place in ComfyUI/models/vae/ + + + +**LoRA Models** + + + + Alpha generation LoRA. Place in ComfyUI/models/loras/ + + + Lightning LoRA for fast inference. Place in ComfyUI/models/loras/ + + + +File storage locations: + +``` +ComfyUI/ +├── models/ +│ ├── diffusion_models/ +│ │ ├── wan2.1_t2v_14B_fp8_scaled.safetensors +│ │ └── wan2.1_t2v_14B_bf16.safetensors +│ ├── text_encoders/ +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── vae/ +│ │ ├── wan_alpha_2.1_vae_rgb_channel.safetensors +│ │ └── wan_alpha_2.1_vae_alpha_channel.safetensors +│ └── loras/ +│ ├── wan_alpha_2.1_rgba_lora.safetensors +│ └── lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors +``` + +### 3. Complete the Workflow Step by Step + +1. Ensure the correct diffusion model is loaded in the `Load Diffusion Model` node +2. Ensure the `Load CLIP` node has loaded `umt5_xxl_fp8_e4m3fn_scaled.safetensors` +3. Ensure the `Load VAE` node for RGB has loaded `wan_alpha_2.1_vae_rgb_channel.safetensors` +4. Ensure the second `Load VAE` node for Alpha has loaded `wan_alpha_2.1_vae_alpha_channel.safetensors` +5. Ensure the `LoRA Loader` node has loaded `wan_alpha_2.1_rgba_lora.safetensors` +6. (Optional) Load `lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors` in the Lightning LoRA node for faster generation +7. Set your positive prompt in the `CLIP Text Encode (Positive Prompt)` node +8. (Optional) Modify video dimensions in the `EmptyHunyuanLatentVideo` node +9. Click the `Run` button or use `Ctrl(Cmd) + Enter` to execute generation diff --git a/tutorials/video/wan/wan-ati.mdx b/tutorials/video/wan/wan-ati.mdx index 6606cc456..0e0628782 100644 --- a/tutorials/video/wan/wan-ati.mdx +++ b/tutorials/video/wan/wan-ati.mdx @@ -6,10 +6,16 @@ sidebarTitle: "WAN2l1 ATI" import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - **ATI (Any Trajectory Instruction)** is a controllable video generation framework proposed by the ByteDance team. ATI is implemented based on Wan2.1 and supports unified control of objects, local regions, and camera motion in videos through arbitrary trajectory instructions. -Project URL: [https://github.com/bytedance/ATI](https://github.com/bytedance/ATI) + + + ATI source code and documentation on GitHub + + + Visual tool to draw and edit motion trajectories on images + + ## Key Features @@ -18,14 +24,23 @@ Project URL: [https://github.com/bytedance/ATI](https://github.com/bytedance/ATI - **Wan2.1 Compatible**: Based on the official Wan2.1 implementation, compatible with environments and model structures. - **Rich Visualization Tools**: Supports visualization of input trajectories, output videos, and trajectory overlays. - ## WAN ATI Trajectory Control Workflow Example -### 1. Workflow Download +### 1. Workflow Download + +Download the video below and drag it into ComfyUI to load the corresponding workflow. + + + + Open in Comfy Cloud + + + Download the workflow video and drag it into ComfyUI to load the workflow + + -Download the video below and drag it into ComfyUI to load the corresponding workflow + ### 2. Model Download -If you haven't successfully downloaded the model files from the workflow, you can try downloading them manually using the links below +If you haven't successfully downloaded the model files from the workflow, you can try downloading them manually using the links below. + +#### Diffusion Model + + + Wan2.1 I2V ATI 14B diffusion model (fp8 precision) + + +#### VAE + + + Wan2.1 VAE model + + +#### Text Encoders + +Choose one of the following models: -**Diffusion Model** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) + + + Full precision text encoder (higher quality, larger size) + + + FP8 text encoder (lower VRAM usage) + + -**VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +#### CLIP Vision -**Text encoders** Chose one of following model -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) + + CLIP Vision model for processing reference images + -**clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +#### File Save Location -File save location ``` -ComfyUI/ -├───📂 models/ -│ ├───📂 diffusion_models/ -│ │ └───Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors -│ ├───📂 text_encoders/ +📂 ComfyUI/ +├── 📂 models/ +│ ├── 📂 diffusion_models/ +│ │ └── Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors +│ ├── 📂 text_encoders/ │ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors # or other version -│ ├───📂 clip_vision/ +│ ├── 📂 clip_vision/ │ │ └─── clip_vision_h.safetensors -│ └───📂 vae/ +│ └── 📂 vae/ │ └── wan_2.1_vae.safetensors ``` -### 3. Complete the workflow execution step by step +### 3. Complete the Workflow Execution Step by Step ![Workflow step diagram](/images/tutorial/video/wan/wan_ati_guide.jpg) -Please follow the numbered steps in the image to ensure smooth execution of the corresponding workflow +Please follow the numbered steps in the image to ensure smooth execution of the corresponding workflow. 1. Ensure the `Load Diffusion Model` node has loaded the `Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors` model 2. Ensure the `Load CLIP` node has loaded the `umt5_xxl_fp8_e4m3fn_scaled.safetensors` model 3. Ensure the `Load VAE` node has loaded the `wan_2.1_vae.safetensors` model 4. Ensure the `Load CLIP Vision` node has loaded the `clip_vision_h.safetensors` model 5. Upload the provided input image in the `Load Image` node -6. Trajectory editing: Currently there is no corresponding trajectory editor in ComfyUI yet, you can use the following link to complete trajectory editing +6. Trajectory editing: Currently there is no corresponding trajectory editor in ComfyUI yet. You can use the following link to complete trajectory editing. - [Online Trajectory Editing Tool](https://comfyui-wiki.github.io/Trajectory-Annotation-Tool/) 7. If you need to modify the prompts (positive and negative), please make changes in the `CLIP Text Encoder` node numbered `5` 8. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute video generation diff --git a/tutorials/video/wan/wan-causal-forcing.mdx b/tutorials/video/wan/wan-causal-forcing.mdx index da05d019f..cf3d71ce7 100644 --- a/tutorials/video/wan/wan-causal-forcing.mdx +++ b/tutorials/video/wan/wan-causal-forcing.mdx @@ -1,6 +1,6 @@ --- title: "Causal Forcing I2V ComfyUI Workflow Example" -description: "Generate videos from images using Causal Forcing or Causal Forcing++ with Wan2.1 — achieving smooth, temporally consistent video in as few as 1 inference step." +description: "Generate videos from images using Causal Forcing or Causal Forcing++ with Wan2.1. Achieves smooth, temporally consistent video in as few as 1 inference step." sidebarTitle: "Causal Forcing I2V" --- @@ -14,7 +14,10 @@ This workflow uses **Wan2.1** and supports both **Causal Forcing** (standard) an - + + + Open in Comfy Cloud + Download JSON or search "Causal Forcing" in Template Library @@ -29,7 +32,7 @@ Unlike standard video generation which processes all frames in parallel, Causal 3. Each new frame becomes the input for the next prediction 4. This repeats for the desired number of frames -This recurrent approach creates **strong temporal consistency** — each frame naturally follows from the one before it — and can produce high-quality results with very few inference steps (1 to 4). +This recurrent approach creates **strong temporal consistency**: each frame naturally follows from the one before it. This can produce high-quality results with very few inference steps (1 to 4). This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. @@ -59,22 +62,22 @@ The workflow accepts a single input image (the first frame) and a text prompt de These are exposed as controls on the blueprint subgraph node: | Parameter | Default | Description | -|---|---|---| -| `unet_name` | — | The Wan2.1 I2V model checkpoint to use | -| `clip_name` | — | The CLIP / text encoder model for the prompt | -| `vae_name` | — | The VAE model for encoding/decoding | -| `width` | — | Output video width | -| `height` | — | Output video height | -| `noise_seed` | — | Seed for reproducibility | +||---|---|---| +| `unet_name` | - | The Wan2.1 I2V model checkpoint to use | +| `clip_name` | - | The CLIP / text encoder model for the prompt | +| `vae_name` | - | The VAE model for encoding/decoding | +| `width` | - | Output video width | +| `height` | - | Output video height | +| `noise_seed` | - | Seed for reproducibility | ## Steps to run -1. **Load an image** — use the **LoadImage** node to load your starting frame -2. **Write a prompt** (optional) — describe the desired video content -3. **Set duration** — how many frames to generate -4. **Select models** — choose Wan2.1 I2V checkpoint, CLIP, and VAE -5. **Choose mode** — Causal Forcing or Causal Forcing++ (set via the subgraph's internal configuration or a Causal Forcing-specific input if available) -6. **Run** — frames will be generated sequentially and saved to `ComfyUI/output/` +1. **Load an image**: use the **LoadImage** node to load your starting frame +2. **Write a prompt** (optional): describe the desired video content +3. **Set duration**: how many frames to generate +4. **Select models**: choose Wan2.1 I2V checkpoint, CLIP, and VAE +5. **Choose mode**: Causal Forcing or Causal Forcing++ (set via the subgraph's internal configuration or a Causal Forcing-specific input if available) +6. **Run**: frames will be generated sequentially and saved to `ComfyUI/output/` ## Model downloads @@ -83,22 +86,22 @@ Download the Wan2.1 I2V model and required files. Place them in the correspondin ### Wan2.1 I2V - - wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B checkpoint + + wan2.1_i2v_480p_14B_fp16.safetensors - Wan2.1 I2V 14B checkpoint - - wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B checkpoint (8GB VRAM minimum) + + wan2.1_t2v_1.3B_fp16.safetensors - Wan2.1 1.3B checkpoint (8GB VRAM minimum) ### CLIP and VAE - - google-bert/bert-base-uncased — CLIP text encoder + + google-bert/bert-base-uncased - CLIP text encoder - - Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE + + Wan2.1_VAE_bf16.safetensors - Wan2.1 VAE diff --git a/tutorials/video/wan/wan-flf.mdx b/tutorials/video/wan/wan-flf.mdx index 0d368ffa1..af781a447 100644 --- a/tutorials/video/wan/wan-flf.mdx +++ b/tutorials/video/wan/wan-flf.mdx @@ -14,7 +14,7 @@ Users only need to provide two images as the starting and ending frames, and the 1. **Precise First-Last Frame Control**: The matching rate of first and last frames reaches 98%, defining video boundaries through starting and ending scenes, intelligently filling intermediate dynamic changes to achieve scene transitions and object morphing effects. 2. **Stable and Smooth Video Generation**: Using CLIP semantic features and cross-attention mechanisms, the video jitter rate is reduced by 37% compared to similar models, ensuring natural and smooth transitions. 3. **Multi-functional Creative Capabilities**: Supports dynamic embedding of Chinese and English subtitles, generation of anime/realistic/fantasy and other styles, adapting to different creative needs. -4. **720p HD Output**: Directly generates 1280×720 resolution videos without post-processing, suitable for social media and commercial applications. +4. **720p HD Output**: Directly generates 1280x720 resolution videos without post-processing, suitable for social media and commercial applications. 5. **Open-source Ecosystem Support**: Model weights, code, and training framework are fully open-sourced, supporting deployment on mainstream AI platforms. **Technical Principles and Architecture** @@ -39,40 +39,67 @@ Since this model is trained on high-resolution images, using smaller sizes may n If needed, please adjust the video generation size for testing. A small generation size may not produce good output with this model, please notice that. -Please download the WebP file below, and drag it into ComfyUI to load the corresponding workflow. The workflow has embedded the corresponding model download file information. - -![Wan2.1 FLF2V 720P f16 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1_flf2v/wan2.1_flf2v_720_f16.webp) - - -Please download the two images below, which we will use as the starting and ending frames of the video - -![start_image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1_flf2v/input/start_image.png) -![end_image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1_flf2v/input/end_image.png) +Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan2.1 FLF2V 720P" in the Template Library under `Workflow` → `Browse Templates` → `Video`. + + + + Open in Comfy Cloud + + + Download JSON or search "Wan2.1 FLF2V" in Template Library + + + + + + Starting frame for the video generation. Download and use this image, or replace with your own. + + + Ending frame for the video generation. Download and use this image, or replace with your own. + + ### 2. Manual Model Installation -If corresponding - All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files). -**diffusion_models** Choose one version based on your hardware conditions -- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) -- FP8:[wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) +**Diffusion Models** — Choose one version based on your hardware + + + + wan2.1_flf2v_720p_14B_fp16.safetensors — Full precision, requires more VRAM + + + wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors — Quantized version, lower VRAM usage + + If you have previously tried Wan Video related workflows, you may already have the following files. -Choose one version from **Text encoders** for download, -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +**Text Encoders** — Choose one version + + + + umt5_xxl_fp16.safetensors — Full precision text encoder + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors — Quantized text encoder + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) + + + wan_2.1_vae.safetensors — Wan2.1 VAE for encoding/decoding + **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) + + clip_vision_h.safetensors — CLIP Vision encoder + File Storage Location ``` @@ -101,4 +128,3 @@ ComfyUI/ 7. (Optional) Modify the positive and negative prompts, both Chinese and English are supported 8. (**Important**) In `WanFirstLastFrameToVideo` we use 720*1280 as default size.because it's a 720P model, so using a small size will not yield good output. Please use size around 720*1280 for good generation. 9. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute video generation - diff --git a/tutorials/video/wan/wan-move.mdx b/tutorials/video/wan/wan-move.mdx index d8c001d52..1db485be7 100644 --- a/tutorials/video/wan/wan-move.mdx +++ b/tutorials/video/wan/wan-move.mdx @@ -21,37 +21,37 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Wan-Move image-to-video workflow - -

Download JSON Workflow File

- + + Download workflow + - -

Run on ComfyUI Cloud

- + + Open in cloud + ## Model links -**text_encoders** + + **text_encoders** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors + -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + **clip_vision** -- clip_vision_h.safetensors + -**clip_vision** + + **loras** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors + -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) + + **diffusion_models** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors + -**loras** - -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) - -**diffusion_models** - -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) - -**vae** - -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + **vae** -- wan_2.1_vae.safetensors + **Model Storage Location** diff --git a/tutorials/video/wan/wan-video.mdx b/tutorials/video/wan/wan-video.mdx index 694610891..6514a89ac 100644 --- a/tutorials/video/wan/wan-video.mdx +++ b/tutorials/video/wan/wan-video.mdx @@ -1,9 +1,10 @@ --- -title: ComfyUI Wan2.1 Video Examples +title: "ComfyUI Wan2.1 Video Examples" description: "This guide demonstrates how to generate videos with first and last frames using Wan2.1 Video in ComfyUI" -sidebarTitle: Wan2.1 +sidebarTitle: "Wan2.1" --- +import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; Wan2.1 Video series is a video generation model open-sourced by Alibaba in February 2025 under the [Apache 2.0 license](https://github.com/Wan-Video/Wan2.1?tab=Apache-2.0-1-ov-file). It offers two versions: @@ -32,46 +33,65 @@ Please update ComfyUI to the latest version before starting the examples to make All models mentioned in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files). Below are the common models you'll need for the examples in this guide, which you can download in advance: Choose one version from **Text encoders** to download: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) + + + + FP16 precision text encoder. Place in ComfyUI/models/text_encoders/ + + + FP8 scaled text encoder. Place in ComfyUI/models/text_encoders/ + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) + + + Wan2.1 VAE model. Place in ComfyUI/models/vae/ + **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) + + + CLIP Vision model for image conditioning. Place in ComfyUI/models/clip_vision/ + File storage locations: ``` ComfyUI/ ├── models/ │ ├── diffusion_models/ -│ ├── ... # Let's download the models in the corresponding workflow +│ │ └── ... (download per workflow) │ ├── text_encoders/ -│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors -│ └── vae/ -│ │ └── wan_2.1_vae.safetensors +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── vae/ +│ │ └── wan_2.1_vae.safetensors │ └── clip_vision/ -│ └── clip_vision_h.safetensors +│ └── clip_vision_h.safetensors ``` For diffusion models, we'll use the fp16 precision models in this guide because we've found that they perform better than the bf16 versions. If you need other precision versions, please visit [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) to download them. -## Wan2.1 Text-to-Video Workflow +## Wan2.1 Text-to-Video Workflow (1.3B) -Before starting the workflow, please download [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) and save it to the `ComfyUI/models/diffusion_models/` directory. + + + Download the workflow image and drag it into ComfyUI to load the workflow + + -> If you need other t2v precision versions, please visit [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) to download them. +### Model Downloads -### 1. Workflow File Download + + Diffusion model for Wan2.1 Text-to-Video. Place in ComfyUI/models/diffusion_models/ + -Download the file below and drag it into ComfyUI to load the corresponding workflow: +> If you need other t2v precision versions, please visit [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) to download them. ![Wan2.1 Text-to-Video Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_t2v_1.3b.webp) -### 2. Complete the Workflow Step by Step +### Steps to Run ![ComfyUI Wan2.1 Workflow Steps](/images/tutorial/video/wan/wan2.1_t2v_1.3b_flow_diagram.jpg) @@ -82,25 +102,36 @@ Download the file below and drag it into ComfyUI to load the corresponding workf 5. (Optional) If you need to modify the prompts (positive and negative), make changes in the `CLIP Text Encoder` node at number `5` 6. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to execute the video generation -## Wan2.1 Image-to-Video Workflow +## Wan2.1 Image-to-Video Workflow (14B) **Since Wan Video separates the 480P and 720P models**, we'll need to provide examples for both resolutions in this guide. In addition to using different models, they also have slight parameter differences. ### 480P Version -#### 1. Workflow and Input Image + + + Open in Comfy Cloud + + + Download JSON or search "Wan 2.1 Image to Video" in Template Library + + -Download the image below and drag it into ComfyUI to load the corresponding workflow: -![Wan2.1 Image-to-Video Workflow 14B 480P Workflow Example Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_480P.webp) + + + Download the default input image, or use your own image. + + -We'll use the following image as input: +![Wan2.1 Image-to-Video Workflow 14B 480P](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_480P.webp) -![Wan2.1 Image-to-Video Workflow 14B 480P Workflow Example Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) +#### Model Downloads -#### 2. Model Download -Please download [wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) and save it to the `ComfyUI/models/diffusion_models/` directory. + + Diffusion model for Wan2.1 I2V 480P. Place in ComfyUI/models/diffusion_models/ + -#### 3. Complete the Workflow Step by Step +#### Steps to Run ![ComfyUI Wan2.1 Workflow Steps](/images/tutorial/video/wan/wan2.1_i2v_14b_480p_flow_diagram.jpg) @@ -115,20 +146,27 @@ Please download [wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Co ### 720P Version -#### 1. Workflow and Input Image - -Download the image below and drag it into ComfyUI to load the corresponding workflow: -![Wan2.1 Image-to-Video Workflow 14B 720P Workflow Example Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_720P.webp) + + + Download the workflow image and drag it into ComfyUI to load the workflow + + -We'll use the following image as input: + + + Download the default input image, or use your own image. + + -![Wan2.1 Image-to-Video Workflow 14B 720P Workflow Example Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/magician.png) +![Wan2.1 Image-to-Video Workflow 14B 720P](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_720P.webp) -#### 2. Model Download +#### Model Downloads -Please download [wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) and save it to the `ComfyUI/models/diffusion_models/` directory. + + Diffusion model for Wan2.1 I2V 720P. Place in ComfyUI/models/diffusion_models/ + -#### 3. Complete the Workflow Step by Step +#### Steps to Run ![ComfyUI Wan2.1 Workflow Steps](/images/tutorial/video/wan/wan2.1_i2v_14b_720p_flow_diagram.jpg) diff --git a/tutorials/video/wan/wan2-2-animate.mdx b/tutorials/video/wan/wan2-2-animate.mdx index ca94cc4a2..7ac1a29d1 100644 --- a/tutorials/video/wan/wan2-2-animate.mdx +++ b/tutorials/video/wan/wan2-2-animate.mdx @@ -35,24 +35,18 @@ It can also replace characters in a video with animated characters, preserving t ## About Wan2.2 Animate workflow -In this docs, we will provide two workflow: +This workflow supports two modes: Mix (replace a character in a video with a character from a reference image) and Move (animate a character using motion from an input video). -1. Workflow that only uses core nodes (It is incomplete; you need to preprocess the image by yourself first) -2. Workflow that includes some custom nodes (It is complete; you can use it directly, but some new user might not know how to install the custom nodes) +Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan2.2 Animate" in the Template Library under `Workflow` → `Browse Templates` → `Video`. -## Wan2.2 Anmate ComfyUI native workflow(without custom nodes) - -### 1. Download Workflow File - -Download the following workflow file and drag it into ComfyUI to load the workflow. - - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + Open in Comfy Cloud + + + Download JSON or search "Wan2.2 Animate" in Template Library + + Download materials below as input: @@ -68,21 +62,42 @@ Download materials below as input: ### 2. Model links -**diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) This is the model that from Kijai's repo -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) original model weight +All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files). + +**Diffusion Models** + + + + Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: Scaled FP8 version from Kijai's repo + + + wan2.2_animate_14B_bf16.safetensors: Original bf16 model weight + + + +**CLIP Vision** + + + clip_vision_h.safetensors: CLIP Vision encoder + + +**LoRAs** + + + lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4-step acceleration LoRA + -**clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +**VAE** -**loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 这是一个 4 步的加速 lora + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding + -**vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +**Text Encoders** -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + ``` ComfyUI/ diff --git a/tutorials/video/wan/wan2-2-fun-camera.mdx b/tutorials/video/wan/wan2-2-fun-camera.mdx index fbbfbf8f3..69163f42a 100644 --- a/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -44,31 +44,60 @@ Download the video or JSON file below and drag it into ComfyUI to load the corre src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

Download JSON Workflow

- + + + Open in Comfy Cloud + + + Download JSON or search "Wan2.2 Fun Camera" in Template Library + + Please download the image below, which we will use as input. -![Input Starting Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) + + + Starting frame for the video generation. Download and use this image, or replace with your own. + + ### 2. Model Links The following models can be found in [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged): -**Diffusion Model** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) +**Diffusion Models** + + + + High noise diffusion model for Wan2.2 Fun Camera + + + Low noise diffusion model for Wan2.2 Fun Camera + + **Wan2.2-Lightning LoRA (Optional, for acceleration)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + 4-step acceleration LoRA for high noise model + + + 4-step acceleration LoRA for low noise model + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + Wan2.1 VAE for encoding/decoding + + +**Text Encoder** + + + FP8 scaled text encoder + File save location @@ -82,7 +111,7 @@ ComfyUI/ │ │ ├─── wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors │ │ └─── wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors │ ├───📂 text_encoders/ -│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors │ └───📂 vae/ │ └── wan_2.1_vae.safetensors ``` @@ -110,4 +139,4 @@ ComfyUI/ - **Width/Height**: Set video resolution - **Length**: Set the number of video frames (default is 81 frames) - **Speed**: Set video speed (default is 1.0) -8. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to execute video generation \ No newline at end of file +8. Click the `Run` button or use the shortcut `Ctrl(cmd) + Enter` to execute video generation diff --git a/tutorials/video/wan/wan2-2-fun-control.mdx b/tutorials/video/wan/wan2-2-fun-control.mdx index 6fd348867..024ebec5b 100644 --- a/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/tutorials/video/wan/wan2-2-fun-control.mdx @@ -32,7 +32,6 @@ Below are the relevant model weights and code repositories: - This workflow provides two versions: 1. A version using [Wan2.2-Lightning](https://huggingface.co/lightx2v/Wan2.2-Lightning) 4-step LoRA from lightx2v: may cause some loss in video dynamics but offers faster speed 2. A fp8_scaled version without acceleration LoRA @@ -48,47 +47,67 @@ Since using the 4-step LoRA provides a better experience for first-time workflow ### 1. Download Workflow and Materials -Download the video below or JSON file and drag it into ComfyUI to load the workflow - - +Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan2.2 Fun Control" in the Template Library under `Workflow` → `Browse Templates` → `Video`. - -

Download JSON Workflow

- + + + Open in Comfy Cloud + + + Download JSON or search "Wan2.2 Fun Control" in Template Library + + Please download the following images and videos as input materials. -![Input start image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) + + + Start frame for video generation. Download and use this image, or replace with your own. + + + Preprocessed pose control video. Download and use this video, or replace with your own. + + - - -> We use a preprocessed video here. +> We use a preprocessed video here. ### 2. Models -You can find the models below at [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) +All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files). + +**Diffusion Models** -**Diffusion Model** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) + + + wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors: High noise diffusion model + + + wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors: Low noise diffusion model + + **Wan2.2-Lightning LoRA (Optional, for acceleration)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: High noise 4-step acceleration LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: Low noise 4-step acceleration LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding/decoding + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + ``` ComfyUI/ @@ -111,7 +130,7 @@ ComfyUI/ ![Wan2.2 Fun Control Workflow Steps](/images/tutorial/video/wan/wan2_2/wan_2.2_14b_fun_control.jpg) - This workflow uses LoRA. Please ensure the corresponding Diffusion model and LoRA are matched - high noise and low noise models and LoRAs need to be used correspondingly. + This workflow uses LoRA. Please ensure the corresponding Diffusion model and LoRA are matched: high noise and low noise models and LoRAs need to be used correspondingly. 1. **High noise** model and **LoRA** loading @@ -125,10 +144,10 @@ ComfyUI/ 5. Upload the start frame in the `Load Image` node 6. In the second `Load video` node, load the pose control video. The provided video has been preprocessed and can be used directly 7. Since we provide a preprocessed pose video, the corresponding video image preprocessing node needs to be disabled. You can select it and use `Ctrl + B` to disable it -8. Modify the Prompt - you can use both Chinese and English +8. Modify the Prompt: you can use both Chinese and English 9. In `Wan22FunControlToVideo`, modify the video dimensions. The default is set to 640×640 resolution to avoid excessive processing time for users with low VRAM 10. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute video generation ### Additional Notes -Since ComfyUI's built-in nodes only include Canny preprocessor, you can use tools like [ComfyUI-comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux) to implement other types of image preprocessing \ No newline at end of file +Since ComfyUI's built-in nodes only include Canny preprocessor, you can use tools like [ComfyUI-comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux) to implement other types of image preprocessing diff --git a/tutorials/video/wan/wan2-2-fun-inp.mdx b/tutorials/video/wan/wan2-2-fun-inp.mdx index 981863515..a744c578a 100644 --- a/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -11,7 +11,7 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' **Key Features**: - **Start-End Frame Control**: Supports inputting start and end frame images to generate intermediate transition videos, enhancing video coherence and creative freedom - **High-Quality Video Generation**: Based on the Wan2.2 architecture, outputs film-level quality videos -- **Multi-Resolution Support**: Supports generating videos at 512×512, 768×768, 1024×1024 and other resolutions to suit different scenarios +- **Multi-Resolution Support**: Supports generating videos at 512x512, 768x768, 1024x1024 and other resolutions to suit different scenarios **Model Version**: - **14B High-Performance Version**: Model size exceeds 32GB, with better results but requires high VRAM @@ -39,14 +39,14 @@ This workflow provides two versions: 1. A version using [Wan2.2-Lightning](https://huggingface.co/lightx2v/Wan2.2-Lightning) 4-step LoRA from lightx2v for accelerated video generation 2. A fp8_scaled version without acceleration LoRA -Below are the test results using an RTX4090D 24GB VRAM GPU at 640×640 resolution with 81 frames +Below are the test results using an RTX4090D 24GB VRAM GPU at 640x640 resolution with 81 frames | Model Type | VRAM Usage | First Generation Time | Second Generation Time | | ------------------------ | ---------- | -------------------- | --------------------- | | fp8_scaled | 83% | ≈ 524s | ≈ 520s | | fp8_scaled + 4-step LoRA | 89% | ≈ 138s | ≈ 79s | -Since the acceleration with LoRA is significant but the video dynamic is lost, the provided workflows enable the accelerated LoRA version by default. If you want to enable the other workflow, select it and use **Ctrl+B** to activate. +Since the acceleration with LoRA is significant but the video dynamic is lost, the provided workflows enable the accelerated LoRA version by default. If you want to enable the other workflow, select it and use **Ctrl+B** to activate. ### 1. Download Workflow File @@ -54,13 +54,14 @@ Please update your ComfyUI to the latest version, and find "**Wan2.2 Fun Inp**" Or, after updating ComfyUI to the latest version, download the workflow below and drag it into ComfyUI to load. - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + Download JSON or search "Wan2.2 Fun Inp" in Template Library + + + Open in Comfy Cloud + + Use the following materials as the start and end frames @@ -69,19 +70,41 @@ Use the following materials as the start and end frames ### 2. Models -**Diffusion Model** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) +All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files). + +**Diffusion Models** + + + + wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: High noise diffusion model for start-end frame inpainting + + + wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: Low noise diffusion model for start-end frame inpainting + + **Lightning LoRA (Optional, for acceleration)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 4-step acceleration LoRA for high noise model + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 4-step acceleration LoRA for low noise model + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + ``` ComfyUI/ diff --git a/tutorials/video/wan/wan2-2-s2v.mdx b/tutorials/video/wan/wan2-2-s2v.mdx index 36a587e73..7363fa519 100644 --- a/tutorials/video/wan/wan2-2-s2v.mdx +++ b/tutorials/video/wan/wan2-2-s2v.mdx @@ -33,38 +33,58 @@ Download the following workflow file and drag it into ComfyUI to load the workfl src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + Open in Comfy Cloud + + + Download JSON or search "Wan2.2 S2V" in Template Library + + Download the following image and audio as input: -![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - - -

Download Input Audio

- + + + Download the default input image, or use your own image. + + + Download the default input audio, or use your own audio. + + ### 2. Model Links You can find the models in [our repo](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) -**diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +**diffusion_models** + + + + FP8 scaled diffusion model. Place in ComfyUI/models/diffusion_models/ + + + BF16 diffusion model. Place in ComfyUI/models/diffusion_models/ + + **audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) + + + Audio encoder model. Place in ComfyUI/models/audio_encoders/ + **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + Wan2.1 VAE model. Place in ComfyUI/models/vae/ + + +**text_encoders** + + + FP8 scaled text encoder. Place in ComfyUI/models/text_encoders/ + ``` @@ -92,8 +112,14 @@ ComfyUI/ You can find both models [here](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models): -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) + + + FP8 scaled diffusion model + + + BF16 diffusion model + + This template uses `wan2.2_s2v_14B_fp8_scaled.safetensors`, which requires less VRAM. But you can try `wan2.2_s2v_14B_bf16.safetensors` to reduce quality degradation. diff --git a/tutorials/video/wan/wan2_2.mdx b/tutorials/video/wan/wan2_2.mdx index 62190cd74..702ac5bbf 100644 --- a/tutorials/video/wan/wan2_2.mdx +++ b/tutorials/video/wan/wan2_2.mdx @@ -89,24 +89,42 @@ Please update your ComfyUI to the latest version, and through the menu `Workflow src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - -

Download JSON Workflow File

- - - -

Run on Comfy Cloud

- + + + Download JSON or search "Wan2.2 5B" in Template Library + + + Open in Comfy Cloud + + ### 2. Manually Download Models +All models mentioned can be found at [Comfy-Org/Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged). + **Diffusion Model** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) + + + + 5B hybrid diffusion model. Place in ComfyUI/models/diffusion_models/ + + **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + + Wan2.2 VAE. Place in ComfyUI/models/vae/ + + + +**Text Encoder** + + + + FP8 text encoder. Place in ComfyUI/models/text_encoders/ + + ``` ComfyUI/ @@ -144,26 +162,45 @@ Or update your ComfyUI to the latest version, then download the following video src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - -

Download JSON Workflow File

- - - -

Run on Comfy Cloud

- + + + Download JSON or search "Wan2.2 14B T2V" in Template Library + + + Open in Comfy Cloud + + ### 2. Manually Download Models -**Diffusion Model** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +All models mentioned can be found at [Comfy-Org/Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged). + +**Diffusion Models** + + + + High noise diffusion model. Place in ComfyUI/models/diffusion_models/ + + + Low noise diffusion model. Place in ComfyUI/models/diffusion_models/ + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + + Wan2.1 VAE (compatible with Wan2.2). Place in ComfyUI/models/vae/ + + +**Text Encoder** + + + + FP8 text encoder. Place in ComfyUI/models/text_encoders/ + + ``` ComfyUI/ @@ -201,28 +238,48 @@ Or update your ComfyUI to the latest version, then download the following video src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - -

Download JSON Workflow File

- - - -

Run on Comfy Cloud

- + + + Download JSON or search "Wan2.2 14B I2V" in Template Library + + + Open in Comfy Cloud + + You can use the following image as input: ![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) ### 2. Manually Download Models -**Diffusion Model** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +All models mentioned can be found at [Comfy-Org/Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged). + +**Diffusion Models** + + + + High noise I2V diffusion model. Place in ComfyUI/models/diffusion_models/ + + + Low noise I2V diffusion model. Place in ComfyUI/models/diffusion_models/ + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + + Wan2.1 VAE (compatible with Wan2.2). Place in ComfyUI/models/vae/ + + + +**Text Encoder** + + + + FP8 text encoder. Place in ComfyUI/models/text_encoders/ + + ``` ComfyUI/ @@ -260,13 +317,14 @@ Download the video or the JSON workflow below and open it in ComfyUI. src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + Download JSON or search "Wan2.2 14B FLF2V" in Template Library + + + Open in Comfy Cloud + + Download the following images as input materials: @@ -306,4 +364,4 @@ Download the following images as input materials: [Kijai/WanVideo_comfy/Lightx2v](https://huggingface.co/Kijai/WanVideo_comfy/tree/main/Lightx2v) **Lightx2v 4steps LoRA** -- [Wan2.2-T2V-A14B-4steps-lora-rank64-V1](https://huggingface.co/lightx2v/Wan2.2-Lightning/tree/main/Wan2.2-T2V-A14B-4steps-lora-rank64-V1) \ No newline at end of file +- [Wan2.2-T2V-A14B-4steps-lora-rank64-V1](https://huggingface.co/lightx2v/Wan2.2-Lightning/tree/main/Wan2.2-T2V-A14B-4steps-lora-rank64-V1) diff --git a/tutorials/video/zai/scail2.mdx b/tutorials/video/zai/scail2.mdx index b72de7e83..e969fef1c 100644 --- a/tutorials/video/zai/scail2.mdx +++ b/tutorials/video/zai/scail2.mdx @@ -12,7 +12,7 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - **End-to-End Character Animation**: Drive a still character image with motion from a driving video - **Two Modes**: Animation Mode (character performs the motion) and Replacement Mode (swap tracked person with reference character) - **Long Video Support**: Chunk-based extended generation with frame overlap between segments -- **Built-in ComfyUI Nodes**: Uses native `WanSCAILToVideo`, `SCAIL2ColoredMask`, and `SAM3` tracking — no custom nodes required beyond standard model downloads +- **Built-in ComfyUI Nodes**: Uses native `WanSCAILToVideo`, `SCAIL2ColoredMask`, and `SAM3` tracking. No custom nodes required beyond standard model downloads. **Related Links**: - [Hugging Face: SCAIL-2](https://huggingface.co/zai-org/SCAIL-2) @@ -33,7 +33,7 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## How the Workflow Works -This workflow uses two subgraph nodes — a **Base** subgraph (first segment) and an **Extend** subgraph (subsequent segments) — to support character animation for both short and long videos. +This workflow uses two subgraph nodes: a **Base** subgraph (first segment) and an **Extend** subgraph (subsequent segments). This supports character animation for both short and long videos. 1. **Load** a driving video (`pose_video`) and a reference character image 2. **Base subgraph** processes the first segment (81 frames by default) @@ -48,7 +48,7 @@ This workflow uses two subgraph nodes — a **Base** subgraph (first segment) an For longer videos, calculate the number of segments: `ceil(total_frames / 76)`. Each segment except the first uses the Extend subgraph. Duplicate the Extend node for more segments, chain the `previous_frames` output, and increment `segment_index`. -> **Note:** `WanSCAILToVideo` cannot queue all segments automatically — run each segment manually. +> **Note:** `WanSCAILToVideo` cannot queue all segments automatically. Run each segment manually. ## Two Modes @@ -77,7 +77,7 @@ Set the `replace_mode` parameter on both subgraph nodes. ### SAM3 Tracking (two inputs) -The `sam3_video_object` and `sam3_image_object` inputs control the SAM3 mask tracking — **not** the SCAIL-2 output prompt. These determine which objects are tracked for the colored masks: +The `sam3_video_object` and `sam3_image_object` inputs control the SAM3 mask tracking, **not** the SCAIL-2 output prompt. These determine which objects are tracked for the colored masks: | Input | Target | Output | |-------|--------|--------| diff --git a/zh/tutorials/3d/hunyuan3D-2.mdx b/zh/tutorials/3d/hunyuan3D-2.mdx index 8e581a080..ac0d5f809 100644 --- a/zh/tutorials/3d/hunyuan3D-2.mdx +++ b/zh/tutorials/3d/hunyuan3D-2.mdx @@ -58,9 +58,14 @@ Hunyuan3D-2mv 工作流中，我们将使用多视角的图片来生成3D模型 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 上立即运行此工作流 + + + 下载工作流 JSON 文件 + + ### 1. 工作流 @@ -83,7 +88,7 @@ Hunyuan3D-2mv 工作流中，我们将使用多视角的图片来生成3D模型 下载下面的模型，并保存到对应的 ComfyUI 文件夹 -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv.safetensors` +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv.safetensors` ``` ComfyUI/ @@ -106,9 +111,14 @@ ComfyUI/ Hunyuan3D-2mv-turbo 工作流中，我们将使用 Hunyuan3D-2mv-turbo 模型来生成3D模型，这个模型是 Hunyuan3D-2mv 的分步蒸馏（Step Distillation）版本，可以更快地生成3D模型，在这个版本的工作流中我们设置 `cfg` 为 1.0 并添加 `flux guidance` 节点来控制 `distilled cfg` 的生成。 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 上立即运行此工作流 + + + 下载工作流 JSON 文件 + + ### 1. 工作流 @@ -125,7 +135,7 @@ Hunyuan3D-2mv-turbo 工作流中，我们将使用 Hunyuan3D-2mv-turbo 模型来 下载下面的模型，并保存到对应的 ComfyUI 文件夹 -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv-turbo.safetensors` +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv-turbo.safetensors` ``` ComfyUI/ @@ -146,9 +156,14 @@ ComfyUI/ Hunyuan3D-2 工作流中，我们将使用 Hunyuan3D-2 模型来生成3D模型，这个模型不是一个多视角的模型，在这个工作流中，我们使用`Hunyuan3Dv2Conditioning` 节点替换掉 `Hunyuan3Dv2ConditioningMultiView` 节点。 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 上立即运行此工作流 + + + 下载工作流 JSON 文件 + + ### 1. 工作流 @@ -164,7 +179,7 @@ Hunyuan3D-2 工作流中，我们将使用 Hunyuan3D-2 模型来生成3D模型 下载下面的模型，并保存到对应的 ComfyUI 文件夹 -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2.safetensors` +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2.safetensors` ``` ComfyUI/ diff --git a/zh/tutorials/3d/triposplat.mdx b/zh/tutorials/3d/triposplat.mdx index 634c07066..254287c1b 100644 --- a/zh/tutorials/3d/triposplat.mdx +++ b/zh/tutorials/3d/triposplat.mdx @@ -106,23 +106,23 @@ TripoSplat 使用 **前馈架构**，接收单张 RGB 图像并直接预测一 下载 TripoSplat 模型及所需文件。放入对应的 `models/` 子目录。 - + triposplat_fp16.safetensors — TripoSplat 扩散模型检查点 - + triposplat_vae_decoder_fp16.safetensors — VAE 解码器 - + flux2-vae.safetensors — Flux.2 VAE，用于潜空间编码 - + dino_v3_vit_h.safetensors — CLIP 视觉编码器（DINOv2） - + birefnet.safetensors — 用于预处理的背景去除模型 diff --git a/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx b/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx index 266b852e4..0b2ef80ea 100644 --- a/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -45,7 +45,7 @@ AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和 ### AIO 模型下载 - + 一体化 checkpoint 文件（推荐大多数用户使用）。 @@ -74,19 +74,19 @@ AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和 ### 分离模型下载 - + 扩散模型。 - + 文本编码器 (0.6B)。 - + 文本编码器 (1.7B)。 - + VAE 模型。 diff --git a/zh/tutorials/audio/ace-step/ace-step-v1.mdx b/zh/tutorials/audio/ace-step/ace-step-v1.mdx index 8e87ef6a6..4844eee60 100644 --- a/zh/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/zh/tutorials/audio/ace-step/ace-step-v1.mdx @@ -33,9 +33,15 @@ ACE-Step 作为一个强大的音乐生成基座，提供了丰富的扩展能 点击下面的按钮下载对应的工作流文件，拖入 ComfyUI 中即可加载对应的工作流信息，对应工作流已包含模型下载信息。 - -

下载 Json 格式工作流文件

- + + + 下载 Json 格式工作流文件 + + 你也可以手动下载[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors) 后保存到 `ComfyUI/models/checkpoints` 文件夹下 @@ -60,16 +66,28 @@ ACE-Step 作为一个强大的音乐生成基座，提供了丰富的扩展能 点击下面的按钮下载对应的工作流文件，拖入 ComfyUI 中即可加载对应的工作流信息 - -

下载 Json 格式工作流文件

- + + + 下载 Json 格式工作流文件 + + 下载下面的音频作为输入音频 - -

下载示例音频文件用于输入

- + + + 下载示例音频文件用于输入 + + ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/audio/stable-audio/stable-audio-1.mdx b/zh/tutorials/audio/stable-audio/stable-audio-1.mdx index 29e124b64..e1487cdfe 100644 --- a/zh/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/zh/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -41,7 +41,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" ### 检查点 - + 2.3GB。放入 models/checkpoints/ @@ -56,7 +56,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" ### 文本编码器 - + 提示词处理的文本编码器。放入 models/text_encoders/ diff --git a/zh/tutorials/audio/stable-audio/stable-audio-3.mdx b/zh/tutorials/audio/stable-audio/stable-audio-3.mdx index 114dc7b44..03969d679 100644 --- a/zh/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/zh/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -77,11 +77,11 @@ Stable Audio 3 提供三个变体： ### 检查点 - + 用于 Medium 工作流。放入 models/checkpoints/ - + 用于 Medium Base 工作流。放入 models/checkpoints/ @@ -97,11 +97,11 @@ Stable Audio 3 提供三个变体： ### 文本编码器 - + 所有 Stable Audio 3 工作流都需要。放入 models/text_encoders/ - + Medium 工作流需要（Qwen 重新提示）。放入 models/text_encoders/ diff --git a/zh/tutorials/basic/inpaint.mdx b/zh/tutorials/basic/inpaint.mdx index f109e95f1..18fd9fa29 100644 --- a/zh/tutorials/basic/inpaint.mdx +++ b/zh/tutorials/basic/inpaint.mdx @@ -35,7 +35,7 @@ translationBlockHashes: #### 1. 模型安装 下载下面的模型文件，并保存到`ComfyUI/models/checkpoints`目录下 -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) #### 2. 局部重绘素材 @@ -64,7 +64,7 @@ translationBlockHashes: ![ComfyUI 局部重绘工作流 - SD1.5](/images/tutorial/basic/inpaint/inpaint_sd1.5_pruned_emaonly.png) -你会发现 [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) 模型生成的结果局部重绘的效果更好过渡更自然。 +你会发现 [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) 模型生成的结果局部重绘的效果更好过渡更自然。 这因为这个模型是专为 inpainting 设计的模型，它可以帮助我们更好地控制生成区域，从而获得更好的局部重绘效果。 记得我们一直用的比喻吗？不同的模型就像能力不同的画家一样，但每个画家都有自己能力的上限，选择合适的模型可以让你的生成效果更好。 diff --git a/zh/tutorials/basic/outpaint.mdx b/zh/tutorials/basic/outpaint.mdx index 1cff771a3..34733f562 100644 --- a/zh/tutorials/basic/outpaint.mdx +++ b/zh/tutorials/basic/outpaint.mdx @@ -38,7 +38,7 @@ import InstallationModels from '/snippets/zh/tutorials/basic/installation-models #### 1. 模型安装 -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) #### 2. 输入图片 diff --git a/zh/tutorials/controlnet/controlnet.mdx b/zh/tutorials/controlnet/controlnet.mdx index 753c45c72..813295c8d 100644 --- a/zh/tutorials/controlnet/controlnet.mdx +++ b/zh/tutorials/controlnet/controlnet.mdx @@ -74,8 +74,8 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [dreamCreationVirtual3DECommerce_v10.safetensors](https://civitai.com/api/download/models/731340?type=Model&format=SafeTensor&size=full&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/depth-controlnet.mdx b/zh/tutorials/controlnet/depth-controlnet.mdx index df103196b..92d6456c0 100644 --- a/zh/tutorials/controlnet/depth-controlnet.mdx +++ b/zh/tutorials/controlnet/depth-controlnet.mdx @@ -55,7 +55,7 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [architecturerealmix_v11.safetensors](https://civitai.com/api/download/models/431755?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) +- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/depth-t2i-adapter.mdx b/zh/tutorials/controlnet/depth-t2i-adapter.mdx index 6c774a563..ebc2603e7 100644 --- a/zh/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/zh/tutorials/controlnet/depth-t2i-adapter.mdx @@ -76,7 +76,7 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [interiordesignsuperm_v2.safetensors](https://civitai.com/api/download/models/93152?type=Model&format=SafeTensor&size=full&fp=fp16) -- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/resolve/main/models/t2iadapter_depth_sd15v2.pth?download=true) +- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_depth_sd15v2.pth?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/mixing-controlnets.mdx b/zh/tutorials/controlnet/mixing-controlnets.mdx index 05e0f3c74..6530fbe94 100644 --- a/zh/tutorials/controlnet/mixing-controlnets.mdx +++ b/zh/tutorials/controlnet/mixing-controlnets.mdx @@ -54,9 +54,9 @@ translationBlockHashes: - [awpainting_v14.safetensors](https://civitai.com/api/download/models/624939?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx b/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx index 63c1d973f..88713f882 100644 --- a/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -49,10 +49,10 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 如果你网络无法顺利完成对应模型的自动下载，请尝试手动下载下面的模型，并放置到指定目录中 -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) - [majicmixRealistic_v7.safetensors](https://civitai.com/api/download/models/176425?type=Model&format=SafeTensor&size=pruned&fp=fp16) - [japaneseStyleRealistic_v20.safetensors](https://civitai.com/api/download/models/85426?type=Model&format=SafeTensor&size=pruned&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/flux/flux-1-controlnet.mdx b/zh/tutorials/flux/flux-1-controlnet.mdx index 018ff43cc..2937a115a 100644 --- a/zh/tutorials/flux/flux-1-controlnet.mdx +++ b/zh/tutorials/flux/flux-1-controlnet.mdx @@ -44,9 +44,14 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 ## FLUX.1-Canny-dev 完整版工作流 - -

Run on Comfy Cloud

- + + + 下载 JSON 或在模板库中搜索 "Flux.1 Canny" + + + 在 Comfy Cloud 中打开 + + ### 1. 工作流及相关素材 请下载下面的工作流图片,并拖入 ComfyUI 以加载工作流 @@ -66,10 +71,10 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 完整模型列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true) （请确保你已经同意了对应 repo 的协议） +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true) （请确保你已经同意了对应 repo 的协议） 文件保存位置： ``` @@ -109,9 +114,14 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora 工作流 - -

Run on Comfy Cloud

- + + + 下载 JSON 或在模板库中搜索 "Flux.1 Depth LoRA" + + + 在 Comfy Cloud 中打开 + + LoRA 版本的工作流是在完整版本的基础上，添加了 LoRA 模型,相对于[完整版本的 Flux 工作流](/zh/tutorials/flux/flux-1-text-to-image),增加了对应 LoRA 模型的加载使用节点。 @@ -132,11 +142,11 @@ LoRA 版本的工作流是在完整版本的基础上，添加了 LoRA 模型, 完整模型列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) 文件保存位置： ``` diff --git a/zh/tutorials/flux/flux-1-fill-dev.mdx b/zh/tutorials/flux/flux-1-fill-dev.mdx index 6e924c295..ba66cd3f0 100644 --- a/zh/tutorials/flux/flux-1-fill-dev.mdx +++ b/zh/tutorials/flux/flux-1-fill-dev.mdx @@ -39,10 +39,10 @@ Flux.1 fill dev 的核心特点： ![Flux Agreement](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 完整模型列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) 文件保存位置： ``` @@ -61,13 +61,14 @@ ComfyUI/ ### 1. Inpainting 工作流及相关素材 - -

下载工作流图片

- - - -

在 Comfy Cloud 上运行

- + + + Download JSON or search "flux_fill_inpaint" in Template Library + + + Open in Comfy Cloud + + 请下载下面的图片，并拖入 ComfyUI 以加载对应的工作流 ![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -93,7 +94,16 @@ ComfyUI/ ## Flux.1 Fill dev Outpainting 工作流 -### 1. Outpainting 工作流 +### 1. Outpainting 工作流及相关素材 + + + + Download JSON or search "flux_fill_outpaint" in Template Library + + + Open in Comfy Cloud + + 请下载下面的图片，并拖入 ComfyUI 以加载对应的工作流 ![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/zh/tutorials/flux/flux-1-kontext-dev.mdx b/zh/tutorials/flux/flux-1-kontext-dev.mdx index 7934f644d..93e2dd03f 100644 --- a/zh/tutorials/flux/flux-1-kontext-dev.mdx +++ b/zh/tutorials/flux/flux-1-kontext-dev.mdx @@ -37,9 +37,9 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 ### 版本说明 -- **[FLUX.1 Kontext [pro]** - 商业版本，专注快速迭代编辑 -- **FLUX.1 Kontext [max]** - 实验版本，更强的提示遵循能力 -- **FLUX.1 Kontext [dev]** - 开源版本（本教程使用），12B参数，主要用于研究 +- **[FLUX.1 Kontext [pro]** — 商业版本，专注快速迭代编辑 +- **FLUX.1 Kontext [max]** — 实验版本，更强的提示遵循能力 +- **FLUX.1 Kontext [dev]** — 开源版本（本教程使用），12B参数，主要用于研究 目前在 ComfyUI 中，你可以使用所有的这些版本，其中 [Pro 及 Max 版本](/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) 可以通过 API 节点来进行调用，而 Dev 版本开源版本请参考本篇指南中的说明。 @@ -52,7 +52,7 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 **Diffusion Model** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) 如果你想要使用原始权重，可以访问 Black Forest Labs 的相关仓库获取原始模型权重进行使用。 @@ -63,7 +63,7 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 **Text Encoder** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) 或 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) 或 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) 模型保存位置 @@ -81,9 +81,14 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 ## Flux.1 Kontext Dev 工作流 - -

Run on Comfy Cloud

- + + + 下载 JSON 或在模板库中搜索 "Flux Kontext Dev" + + + 在 Comfy Cloud 中打开 + + 这个工作流使用了 `Load Image(from output)` 节点来加载需要编辑的图像，可以让你更方便地获取到编辑后的图像，从而进行多轮次编辑 diff --git a/zh/tutorials/flux/flux-1-text-to-image.mdx b/zh/tutorials/flux/flux-1-text-to-image.mdx index 9cf090d5a..d315f0305 100644 --- a/zh/tutorials/flux/flux-1-text-to-image.mdx +++ b/zh/tutorials/flux/flux-1-text-to-image.mdx @@ -50,25 +50,30 @@ Flux 以其卓越的画面质量和灵活性而闻名，能够生成高质量、 #### 1. 工作流文件 + + + 在 Comfy Cloud 上运行此工作流 + + + 下载 JSON 或在模板库中搜索"Flux.1 Dev" + + + 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Dev 原始版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - -

在 Comfy Cloud 上运行

- - #### 2. 手动安装模型 - `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) 的协议后才能使用浏览器进行下载。 -- 如果你的显存较低，可以尝试使用 [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) 来替换 `t5xxl_fp16.safetensors` 文件。 +- 如果你的显存较低，可以尝试使用 [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) 来替换 `t5xxl_fp16.safetensors` 文件。 请下载下面的模型文件： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) 文件保存位置： @@ -104,14 +109,19 @@ ComfyUI/ #### 1. 工作流文件 + + + 在 Comfy Cloud 上运行此工作流 + + + 下载 JSON 或在模板库中搜索"Flux.1 Schnell" + + + 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Schnell 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) - -

在 Comfy Cloud 上运行

- - #### 2. 手动安装模型 @@ -122,10 +132,10 @@ ComfyUI/ 完整模型文件列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) 文件保存位置： ``` @@ -157,24 +167,38 @@ fp8 版本是对 flux1 原版 fp16 版本的量化版本，在一定程度上这 ### Flux.1 Dev fp8 Checkpoint 版工作流 + + + 在 Comfy Cloud 上运行此工作流 + + + 下载 JSON 或在模板库中搜索"Flux.1 Dev FP8" + + + 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Dev fp8 Checkpoint 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) - -

在 Comfy Cloud 上运行

- - -请下载 [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 +请下载 [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 确保对应的 `Load Checkpoint` 节点加载了 `flux1-dev-fp8.safetensors`，即可测试运行。 ### Flux.1 Schnell fp8 Checkpoint 版工作流 + + + 在 Comfy Cloud 上运行此工作流 + + + 下载 JSON 或在模板库中搜索"Flux.1 Schnell FP8" + + + 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Schnell fp8 Checkpoint 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -请下载[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 +请下载[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 确保对应的 `Load Checkpoint` 节点加载了 `flux1-schnell-fp8.safetensors`，即可测试运行。 diff --git a/zh/tutorials/flux/flux-1-uso.mdx b/zh/tutorials/flux/flux-1-uso.mdx index 599402de6..fad515cc9 100644 --- a/zh/tutorials/flux/flux-1-uso.mdx +++ b/zh/tutorials/flux/flux-1-uso.mdx @@ -32,18 +32,14 @@ USO 支持三种主要方法： ![工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - -

下载 JSON 工作流

- - - -

在 Comfy Cloud 上运行

- + + + 下载工作流 JSON 并将其拖拽到 ComfyUI 中 + + + 在 Comfy Cloud 上运行此工作流 + + 使用下面的图片作为输入 @@ -53,18 +49,18 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', **checkpoints** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) **loras** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **model_patches** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **clip_visions** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) 请下载所有模型并将它们放置在以下目录中： diff --git a/zh/tutorials/flux/flux-2-dev.mdx b/zh/tutorials/flux/flux-2-dev.mdx index 6d500db06..50d2120de 100644 --- a/zh/tutorials/flux/flux-2-dev.mdx +++ b/zh/tutorials/flux/flux-2-dev.mdx @@ -65,15 +65,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/flux/flux-2-klein.mdx b/zh/tutorials/flux/flux-2-klein.mdx index c3823aee0..97884c79b 100644 --- a/zh/tutorials/flux/flux-2-klein.mdx +++ b/zh/tutorials/flux/flux-2-klein.mdx @@ -45,19 +45,19 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 ## Flux.2 Klein 4B 模型下载 - + 4B 模型文本编码器。 - + 扩散模型（4B Base）。 - + 扩散模型（4B 蒸馏版）。 - + 4B 模型 VAE。 @@ -103,11 +103,11 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 扩散模型（9B 蒸馏版）。 - + 9B 模型文本编码器。 - + 9B 模型 VAE。 diff --git a/zh/tutorials/flux/flux1-krea-dev.mdx b/zh/tutorials/flux/flux1-krea-dev.mdx index 23f2514c3..4d5e62769 100644 --- a/zh/tutorials/flux/flux1-krea-dev.mdx +++ b/zh/tutorials/flux/flux1-krea-dev.mdx @@ -31,13 +31,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下载下面的图片或JSON，并拖入 ComfyUI 以加载对应工作流 ![Flux Krea Dev 工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

下载 JSON 格式工作流

- - - -

在 Comfy Cloud 上运行

- + + + 在 Comfy Cloud 上运行此工作流 + + + 下载 JSON 或在模板库中搜索 "Flux.1 Krea Dev" + + #### 2. 模型链接 @@ -47,7 +48,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面这个版本是原始权重，如果你追求更高质量有足够的显存，可以尝试这个版本 -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) - `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 的协议后才能使用浏览器进行下载。 @@ -55,12 +56,12 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 如果你使用过 Flux 相关的工作流，下面的模型是相同的，不需要重复下载 **Text encoders** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) 文件保存位置： ``` diff --git a/zh/tutorials/image/anima/anima.mdx b/zh/tutorials/image/anima/anima.mdx index d2691102f..5639ccb84 100644 --- a/zh/tutorials/image/anima/anima.mdx +++ b/zh/tutorials/image/anima/anima.mdx @@ -82,15 +82,15 @@ Anima 提供两个工作流——基础版适用于标准使用，预览版适 所有模型文件均可从 Hugging Face 的 [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima) 获取。 - + Anima Base v1 扩散模型（2B）。 - + 两个工作流共用的文本编码器（Qwen-3 0.6B）。 - + 两个工作流共用的 VAE。 @@ -111,7 +111,7 @@ Anima 提供两个工作流——基础版适用于标准使用，预览版适 如果您使用 Preview 工作流，请下载以下预览版扩散模型： - + Anima Preview 扩散模型（2B）。 diff --git a/zh/tutorials/image/boogu/boogu-image-0.1.mdx b/zh/tutorials/image/boogu/boogu-image-0.1.mdx index 7d70429c3..6fe4dc03c 100644 --- a/zh/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/zh/tutorials/image/boogu/boogu-image-0.1.mdx @@ -50,19 +50,19 @@ Boogu-Image-0.1-Turbo 工作流使用一个子图封装了扩散、文本编码 ### Boogu-Image-0.1-Turbo 模型下载 - + Boogu-Image-0.1-Turbo 的扩散模型。 - + Boogu-Image-0.1-Turbo 的文本编码器。 - + Boogu-Image-0.1-Turbo 的 VAE。 - + Boogu-Image-0.1-Turbo 的 LoRA 模块 (rank-128)。 @@ -100,15 +100,15 @@ Boogu-Image-0.1-Turbo 工作流使用一个子图封装了扩散、文本编码 ### Boogu-Image-0.1-Edit 模型下载 - + Boogu-Image-0.1-Edit 的扩散模型。 - + Boogu-Image-0.1-Edit 的文本编码器。 - + Boogu-Image-0.1-Edit 的 VAE。 diff --git a/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index ad33ee57c..620e11fdc 100644 --- a/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -37,17 +37,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **Diffusion model** -- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors) +- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_t2i.safetensors) 其它权重请访问 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) 进行下载 **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) 文件保存位置 diff --git a/zh/tutorials/image/ernie-image/ernie-image.mdx b/zh/tutorials/image/ernie-image/ernie-image.mdx index 51f8c668a..da10ec8c4 100644 --- a/zh/tutorials/image/ernie-image/ernie-image.mdx +++ b/zh/tutorials/image/ernie-image/ernie-image.mdx @@ -55,19 +55,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 所有重新打包的模型文件均可在 Hugging Face 的 [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) 获取。 - + ERNIE-Image 扩散模型。 - + ERNIE-Image 文本编码器。 - + ERNIE-Image 提示词增强器文本编码器。 - + ERNIE-Image VAE。 @@ -99,19 +99,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### ERNIE-Image-Turbo 模型下载 - + ERNIE-Image-Turbo 扩散模型。 - + ERNIE-Image-Turbo 文本编码器。 - + ERNIE-Image-Turbo 提示词增强器文本编码器。 - + ERNIE-Image-Turbo VAE。 diff --git a/zh/tutorials/image/hidream/hidream-e1.mdx b/zh/tutorials/image/hidream/hidream-e1.mdx index 8a9fd0afa..b281ca41d 100644 --- a/zh/tutorials/image/hidream/hidream-e1.mdx +++ b/zh/tutorials/image/hidream/hidream-e1.mdx @@ -40,8 +40,8 @@ HiDream-E1 是智象未来(HiDream-ai) 正式开源的交互式图像编辑大 **Diffusion Model** 你不用同时下载这两个模型，由于 E1.1 是基于 E1 的迭代版本，在实际测试中它的质量和效果较 E1 都有较大提升 -- [hidream_e1_1_bf16.safetensors(推荐)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors(推荐)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **Text Encoder**： @@ -73,6 +73,15 @@ HiDream-E1 是智象未来(HiDream-ai) 正式开源的交互式图像编辑大 ## HiDream E1.1 ComfyUI 原生工作流示例 + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "HiDream E1.1" + + + E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一百万分辨率，在工作流中使用了 `Scale Image to Total Pixels` 节点来将输入图片动态调整为 1百万像素 @@ -115,9 +124,14 @@ E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一 ## HiDream E1 ComfyUI 原生 工作流示例 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "HiDream E1 Full" + + E1 是于 2025 年 4 月 28 日发布的，这个模型只支持 768*768 的分辨率 diff --git a/zh/tutorials/image/hidream/hidream-i1.mdx b/zh/tutorials/image/hidream/hidream-i1.mdx index 040270b12..e72b3a8da 100644 --- a/zh/tutorials/image/hidream/hidream-i1.mdx +++ b/zh/tutorials/image/hidream/hidream-i1.mdx @@ -97,15 +97,20 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 full 版本工作流 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 上运行此工作流，无需任何设置 + + + 下载工作流 JSON 文件 + + #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 @@ -134,15 +139,20 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 dev 版本工作流 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 上运行此工作流，无需任何设置 + + + 下载工作流 JSON 文件 + + #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 @@ -170,15 +180,20 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 fast 版本工作流 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 上运行此工作流，无需任何设置 + + + 下载工作流 JSON 文件 + + #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 diff --git a/zh/tutorials/image/hidream/hidream-o1.mdx b/zh/tutorials/image/hidream/hidream-o1.mdx index 4ef55b41f..58f9b6bf8 100644 --- a/zh/tutorials/image/hidream/hidream-o1.mdx +++ b/zh/tutorials/image/hidream/hidream-o1.mdx @@ -51,19 +51,19 @@ HiDream-O1-Image 基于 [MIT 协议](https://github.com/HiDream-ai/HiDream-O1-Im **Checkpoint** — 经过重新打包和量化。所有版本均将最严重的离群值保留在 bf16，并移除了未使用的 deepstack 层： -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量化变体 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 全精度版（文件最大） +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量化变体 +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 全精度版（文件最大） **文本编码器（提示词优化）** — 所有版本通用： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA（可选）** — Dev 蒸馏也可以作为 LoRA 应用到 Full 模型中，让你可以调节蒸馏强度（由 [Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 全秩版 -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 剪枝版 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 基于 checkpoint 的替代蒸馏 +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 全秩版 +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 剪枝版 +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 基于 checkpoint 的替代蒸馏 ``` 📂 ComfyUI/ @@ -107,13 +107,13 @@ HiDream-O1-Image 基于 [MIT 协议](https://github.com/HiDream-ai/HiDream-O1-Im **Checkpoint（Dev 版）** — 经过重新打包和量化。所有版本均将最严重的离群值保留在 bf16，并移除了未使用的 deepstack 层： -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量化变体 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 全精度版（文件最大） +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量化变体 +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 全精度版（文件最大） **文本编码器（提示词优化）** — 所有版本通用： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/zh/tutorials/image/ideogram/ideogram-v4.mdx b/zh/tutorials/image/ideogram/ideogram-v4.mdx index bc4f330ed..73a743a1d 100644 --- a/zh/tutorials/image/ideogram/ideogram-v4.mdx +++ b/zh/tutorials/image/ideogram/ideogram-v4.mdx @@ -46,23 +46,23 @@ Ideogram 4.0 是 Ideogram 最新推出的文生图模型，已作为开源模型 你可以在 Hugging Face 的 [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) 找到所有重新打包的模型文件。 - + Ideogram 4.0 扩散模型（~13.8 GB）。放入 models/diffusion_models/ - + Ideogram 4.0 无条件扩散模型（~13.8 GB）。放入 models/diffusion_models/ - + Ideogram 4.0 文本编码器（~8 GB）。放入 models/text_encoders/ - + Ideogram 4.0 文本编码器（~2 GB）。放入 models/text_encoders/ - + Ideogram 4.0 VAE（~335 MB）。放入 models/vae/ diff --git a/zh/tutorials/image/krea/krea-2.mdx b/zh/tutorials/image/krea/krea-2.mdx index d9e9ff4c5..ba2115ce8 100644 --- a/zh/tutorials/image/krea/krea-2.mdx +++ b/zh/tutorials/image/krea/krea-2.mdx @@ -107,13 +107,13 @@ Krea 还发布了一组 Krea 2 的风格 LoRA。在 **CustomCombo** 节点中选 如需本地使用，请从 [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2) 下载 ComfyUI 优化版模型文件。 - + krea2_turbo_fp8_scaled.safetensors：Turbo FP8（推荐大多数用户使用） - + qwen3vl_4b_fp8_scaled.safetensors：Qwen3VL-4B 文本编码器 - + qwen_image_vae.safetensors diff --git a/zh/tutorials/image/lens/lens.mdx b/zh/tutorials/image/lens/lens.mdx index 14f673e34..bbb3c4780 100644 --- a/zh/tutorials/image/lens/lens.mdx +++ b/zh/tutorials/image/lens/lens.mdx @@ -94,19 +94,19 @@ Lens Turbo 是蒸馏版，只需较少的采样步数即可生成图像，推理 所有模型文件可在 Hugging Face 上的 [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens) 找到。 - + Lens 扩散模型 (BF16) - + Lens Turbo 扩散模型 (BF16) - + Lens 和 Lens Turbo 共用的文本编码器 (GPT-OSS-20B) - + Lens 和 Lens Turbo 共用的 VAE (FLUX.2) diff --git a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 969ac372c..48bd8f50a 100644 --- a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -30,13 +30,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## NewBie-image 文生图工作流 - -

下载 JSON 工作流文件

- - - -

在 ComfyUI Cloud 上运行

- +| +| +| 下载 JSON 或搜索"NewBie-image"模板 +| +| +| 在云端打开 +| +| @@ -44,16 +45,16 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **模型存放位置** diff --git a/zh/tutorials/image/omnigen/omnigen2.mdx b/zh/tutorials/image/omnigen/omnigen2.mdx index eb09cac91..667f22d59 100644 --- a/zh/tutorials/image/omnigen/omnigen2.mdx +++ b/zh/tutorials/image/omnigen/omnigen2.mdx @@ -42,13 +42,13 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 由于本文涉及不同工作流，对应的模型文件及安装位置如下，对应工作流中也已包含了模型文件下载信息： **Diffusion Models）** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **Text Encoders）** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) 文件保存位置： @@ -67,9 +67,11 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 ### 1. 工作流文件下载 - -

在 Comfy Cloud 上运行

- + + + 在 Comfy Cloud 上运行 + + ![文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -95,9 +97,11 @@ OmniGen2 有丰富的图像编辑能力，并且支持为图像添加文本 ### 1. 工作流文件下载 - -

在 Comfy Cloud 上运行

- + + + 在 Comfy Cloud 上运行 + + ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) 下载下面的图片，我们将使用它作为输入图片。 diff --git a/zh/tutorials/image/ovis/ovis-image.mdx b/zh/tutorials/image/ovis/ovis-image.mdx index b28c481ee..9aacd25c0 100644 --- a/zh/tutorials/image/ovis/ovis-image.mdx +++ b/zh/tutorials/image/ovis/ovis-image.mdx @@ -22,13 +22,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Ovis-Image 文生图工作流 - -

下载 JSON 工作流文件

- - - -

在 ComfyUI Cloud 上运行

- + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索"Ovis image" + + @@ -36,15 +37,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders（文本编码器）** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models（扩散模型）** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/pixeldit/pixeldit.mdx b/zh/tutorials/image/pixeldit/pixeldit.mdx index 9daacab20..00b5a3ee8 100644 --- a/zh/tutorials/image/pixeldit/pixeldit.mdx +++ b/zh/tutorials/image/pixeldit/pixeldit.mdx @@ -62,11 +62,11 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' PixelDiT 使用两个模型文件：文本编码器和扩散模型。 - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 文本编码器 - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 扩散模型 diff --git a/zh/tutorials/image/qwen/qwen-image-2512.mdx b/zh/tutorials/image/qwen/qwen-image-2512.mdx index c7e9877ee..0abcfebac 100644 --- a/zh/tutorials/image/qwen/qwen-image-2512.mdx +++ b/zh/tutorials/image/qwen/qwen-image-2512.mdx @@ -43,9 +43,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - + + 在 Comfy Cloud 上运行 - + + + 下载 JSON 或搜索 "Qwen-Image-2512" 在模板库中 + + ### 1. 工作流文件 @@ -55,28 +60,25 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **Text to Image (Qwen-Image 2512)**：标准 50 步生成 - **Text to Image (Qwen-Image 2512 4steps)**：使用 Lightning LoRA 的 4 步加速生成 - -

下载 JSON 工作流

- ### 2. 模型下载 **文本编码器** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（可选 - 用于 4 步 Lightning 加速）** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **扩散模型** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（推荐大多数用户使用） -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（如果您有足够的显存并想要更好的质量） +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（推荐大多数用户使用） +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（如果您有足够的显存并想要更好的质量） **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx index 345d68713..5d8fe82dc 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -33,31 +33,32 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载。 - -

下载 JSON 格式工作流

- - - -

在 ComfyUI Cloud 上运行

- + + + 在 ComfyUI Cloud 上运行 + + + 下载 JSON 格式工作流 + + ### 2. 模型下载 **Text Encoders（文本编码器）** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（可选 - 用于 4 步 Lightning 加速）** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **Diffusion Models（扩散模型）** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/qwen/qwen-image-edit.mdx b/zh/tutorials/image/qwen/qwen-image-edit.mdx index 97a4a2e7e..4ab071a5b 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit.mdx @@ -47,13 +47,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

下载 JSON 格式工作流

- - - -

在 ComfyUI Cloud 上运行

- + + + 下载 JSON 格式工作流或在模板库中搜索"image_qwen_image_edit" + + + 在 ComfyUI Cloud 上运行此工作流，零设置 + + 下载下面的图片作为输入 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -64,19 +65,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Diffusion model** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) Model Storage Location diff --git a/zh/tutorials/image/qwen/qwen-image-layered.mdx b/zh/tutorials/image/qwen/qwen-image-layered.mdx index 49f685ab0..2a835dc5b 100644 --- a/zh/tutorials/image/qwen/qwen-image-layered.mdx +++ b/zh/tutorials/image/qwen/qwen-image-layered.mdx @@ -30,13 +30,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 工作流 - -

下载 JSON 格式工作流

- - - -

在 ComfyUI Cloud 上运行

- +| +| +| 下载 JSON 格式工作流 +| +| +| +| 在 ComfyUI Cloud 上运行 +| +| @@ -44,15 +46,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **模型保存位置** @@ -87,4 +89,4 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 提示词（可选） -文本提示词用于描述输入图像的整体内容——包括可能被部分遮挡的元素（例如，你可以指定隐藏在前景物体后面的文字）。它不是用来明确控制各个图层的语义内容的。 +文本提示词用于描述输入图像的整体内容，包括可能被部分遮挡的元素（例如，你可以指定隐藏在前景物体后面的文字）。它不是用来明确控制各个图层的语义内容的。 diff --git a/zh/tutorials/image/qwen/qwen-image.mdx b/zh/tutorials/image/qwen/qwen-image.mdx index 86cca15d3..a004afd05 100644 --- a/zh/tutorials/image/qwen/qwen-image.mdx +++ b/zh/tutorials/image/qwen/qwen-image.mdx @@ -61,9 +61,12 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - - 在 Comfy Cloud 上运行 - + + + + + + 在本篇文档所附工作流中使用的不同模型有三种 1. Qwen-Image 原版模型 fp8_e4m3fn @@ -84,14 +87,11 @@ GPU: RTX4090D 24GB 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

下载原始版 JSON 格式工作流

- + 蒸馏版 - -

下载蒸馏版JSON 格式工作流

- + + ### 2. 模型下载 **你可以在 ComfyOrg 仓库找到的版本** @@ -104,12 +104,12 @@ GPU: RTX4090D 24GB **Diffusion model** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 蒸馏版本原始作者建议在 15 步 cfg 1.0 @@ -118,15 +118,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) 模型保存位置 @@ -165,18 +165,19 @@ Qwen_image_distill 这是一个 ControlNet 模型 - - 在 Comfy Cloud 上运行 - + + + + + + ### 1. 工作流及输入图片 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

下载 JSON 格式工作流

- + 下载下面的图片作为输入 ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -185,7 +186,7 @@ Qwen_image_distill 1. InstantX Controlnet -下载 [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) 并保存到 `ComfyUI/models/controlnet/` 文件夹下 +下载 [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) 并保存到 `ComfyUI/models/controlnet/` 文件夹下 2. **Lotus Depth model** @@ -199,11 +200,11 @@ Qwen_image_distill **Diffusion Model** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAE Model** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) 或者任意的 SD1.5 的 VAE 都可以使用 +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) 或者任意的 SD1.5 的 VAE 都可以使用 ``` ComfyUI/ @@ -225,9 +226,12 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches 工作流 - - 在 Comfy Cloud 上运行 - + + + + + + 这个模型实际上并不是一个 controlnet，而是一个 Model patch， 支持 canny、depth、inpaint 三种不同的控制模式 @@ -240,9 +244,7 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 下载下面的图片拖入 ComfyUI 中以加载对应的工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

下载 JSON 格式工作流

- + 下载下面的图片作为输入图片： @@ -252,9 +254,9 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 其它模型与 Qwen-Image 基础工作流一致，你只需下载下面的模型并保存到 `ComfyUI/models/model_patches` 文件夹中 -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. 工作流使用说明 @@ -296,9 +298,12 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http ## Qwen Image union ControlNet LoRA 工作流 - - 在 Comfy Cloud 上运行 - + + + + + + 原始模型地址：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 图像结构控制lora 支持 canny、depth、post、lineart、softedge、normal、openpose @@ -307,9 +312,7 @@ Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://h 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

下载 JSON 格式工作流

- + 下载下面的图片作为输入图片 diff --git a/zh/tutorials/image/z-image/z-image-turbo.mdx b/zh/tutorials/image/z-image/z-image-turbo.mdx index e8871670a..bc39c7864 100644 --- a/zh/tutorials/image/z-image/z-image-turbo.mdx +++ b/zh/tutorials/image/z-image/z-image-turbo.mdx @@ -46,15 +46,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### Z-Image-Turbo 模型下载 - + Z-Image-Turbo 文本编码器。 - + Z-Image-Turbo 扩散模型。 - + Z-Image-Turbo VAE。 @@ -81,7 +81,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### ControlNet 所需的额外模型 - + Z-Image-Turbo ControlNet 模型补丁。 diff --git a/zh/tutorials/image/z-image/z-image.mdx b/zh/tutorials/image/z-image/z-image.mdx index 6115cb4b8..da8e02180 100644 --- a/zh/tutorials/image/z-image/z-image.mdx +++ b/zh/tutorials/image/z-image/z-image.mdx @@ -37,15 +37,15 @@ Z-Image（Base）是非蒸馏基础模型，专为社区驱动的微调和自定 ## Z-Image 模型下载 - + Z-Image 文本编码器。 - + Z-Image 扩散模型。 - + Z-Image VAE。 diff --git a/zh/tutorials/llm/gemma4/gemma4.mdx b/zh/tutorials/llm/gemma4/gemma4.mdx index c953ec9f4..956ac5d5f 100644 --- a/zh/tutorials/llm/gemma4/gemma4.mdx +++ b/zh/tutorials/llm/gemma4/gemma4.mdx @@ -74,11 +74,11 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' Gemma 4 模型在 ComfyUI 中以文本编码器（text encoder）形式加载。下载对应的模型文件并放入正确的目录： - + 快速轻量，推荐消费级 GPU 使用。 - + 性能均衡，工作流默认使用此模型。 diff --git a/zh/tutorials/llm/qwen/qwen3.mdx b/zh/tutorials/llm/qwen/qwen3.mdx index 8e0bfa5bc..854a1d1ab 100644 --- a/zh/tutorials/llm/qwen/qwen3.mdx +++ b/zh/tutorials/llm/qwen/qwen3.mdx @@ -71,15 +71,15 @@ Qwen 3.0 非常适合在 ComfyUI 工作流中需要结构化文本生成和智 Qwen 3.0 模型以文本编码器的形式加载到 ComfyUI 中，模型文件与 Qwen3.5 共用。根据你的硬件选择合适的版本： - + 轻量版，约 4.5 GB。适合低显存环境和快速下载。 - + 大小和质量均衡。推荐大多数消费级 GPU。 - + 最大版本，约 19 GB。输出质量更高，需要更多显存。 diff --git a/zh/tutorials/llm/qwen/qwen3_5.mdx b/zh/tutorials/llm/qwen/qwen3_5.mdx index 1e14b68d0..56cb5f3b8 100644 --- a/zh/tutorials/llm/qwen/qwen3_5.mdx +++ b/zh/tutorials/llm/qwen/qwen3_5.mdx @@ -73,15 +73,15 @@ Qwen3.5 在需要将视觉理解与文本生成结合的 ComfyUI 场景中表现 Qwen3.5 模型以文本编码器的形式加载到 ComfyUI 中。根据你的硬件选择合适的版本： - + 轻量版，约 4.5 GB。适合低显存环境和快速下载。 - + 大小和质量均衡。推荐大多数消费级 GPU。 - + 最大版本，约 19 GB。输出质量更高，需要更多显存。 diff --git a/zh/tutorials/partner-nodes/google/gemini.mdx b/zh/tutorials/partner-nodes/google/gemini.mdx index 92c3a0fa7..2f619e998 100644 --- a/zh/tutorials/partner-nodes/google/gemini.mdx +++ b/zh/tutorials/partner-nodes/google/gemini.mdx @@ -22,13 +22,11 @@ Google Gemini 是 Google 推出的一款强大的 AI 模型，支持对话、文 请下载下面的 Json 文件并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + 下载 Json 格式工作流文件 + + ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx index f685102aa..6fc8db3e1 100644 --- a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -51,9 +51,8 @@ Kling 2.6 Motion Control 是由快手开发的专门多模态模型，能够实 ## Kling 2.6 Motion Control 工作流 - -

下载 Json 格式工作流文件

- + + ## 输入要求 diff --git a/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 7d82f0d0e..d479aed4a 100644 --- a/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -54,9 +54,11 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_text_to_video.mp4" > - -

下载 Json 格式工作流文件

- + + + 下载 Json 格式工作流文件 + + ### 2. 按步骤完成工作流的运行 @@ -80,9 +82,11 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_image_to_video.mp4" > - -

下载 Json 格式工作流文件

- + + + 下载 Json 格式工作流文件 + + 下载下面的图片作为输入图片 @@ -112,9 +116,11 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video.mp4" > - -

下载 Json 格式工作流文件

- + + + 下载 Json 格式工作流文件 + + 下载下面的视频作为输入视频： diff --git a/zh/tutorials/partner-nodes/openai/chat.mdx b/zh/tutorials/partner-nodes/openai/chat.mdx index 10107c75e..a5f94132c 100644 --- a/zh/tutorials/partner-nodes/openai/chat.mdx +++ b/zh/tutorials/partner-nodes/openai/chat.mdx @@ -22,13 +22,9 @@ OpenAI 是一家专注于生成式 AI 的科技公司，提供强大的对话功 请下载下面的 Json 文件并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- +| +| 下载 Json 格式工作流文件 +| ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/partner-nodes/rodin/model-generation.mdx b/zh/tutorials/partner-nodes/rodin/model-generation.mdx index 051d1f871..8d490add5 100644 --- a/zh/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/zh/tutorials/partner-nodes/rodin/model-generation.mdx @@ -32,13 +32,9 @@ Hyper3D Rodin (hyper3d.ai) 是一个专注于通过人工智能快速生成高 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + 单视角模型生成 (Json格式) + 下载下面的图片作为输入图片 @@ -66,13 +62,9 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + 多视角模型生成 (Json格式) + 下载下面的图片作为输入图片 diff --git a/zh/tutorials/partner-nodes/runway/video-generation.mdx b/zh/tutorials/partner-nodes/runway/video-generation.mdx index 7000f87c8..1bfb860f8 100644 --- a/zh/tutorials/partner-nodes/runway/video-generation.mdx +++ b/zh/tutorials/partner-nodes/runway/video-generation.mdx @@ -37,9 +37,7 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen3a_turbo_image_to_video/runway_image_to_video_gen3a_turbo.mp4" > - -

下载 Json 格式工作流文件

- + 下载下面的图片作为输入图片 @@ -67,9 +65,7 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen4_turbo_image_to_video/runway_gen4_turo_image_to_video.mp4" > - -

下载 Json 格式工作流文件

- + 下载下面的图片作为输入图片 @@ -99,9 +95,7 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/first_last_frame_to_video/runway_first_last_frame.mp4" > - -

下载 Json 格式工作流文件

- + 下载下面的图片作为输入图片 diff --git a/zh/tutorials/partner-nodes/tripo/model-generation.mdx b/zh/tutorials/partner-nodes/tripo/model-generation.mdx index dca3da2b4..a65091422 100644 --- a/zh/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/zh/tutorials/partner-nodes/tripo/model-generation.mdx @@ -35,9 +35,11 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + 文生模型工作流 + + ### 2. 按步骤完成工作流的运行 @@ -58,9 +60,11 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + 图生模型工作流 + + 下载下面的图片作为输入图片 @@ -86,9 +90,11 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + 多视图模型生成工作流 + + 下载下面的图片作为输入图片 diff --git a/zh/tutorials/utility/depth-anything-3.mdx b/zh/tutorials/utility/depth-anything-3.mdx index 26666a4db..6fdbca81a 100644 --- a/zh/tutorials/utility/depth-anything-3.mdx +++ b/zh/tutorials/utility/depth-anything-3.mdx @@ -39,10 +39,10 @@ ComfyUI 现已原生支持 Depth Anything 3 节点。开始前请确保已更新 下载 Depth Anything 3 的模型文件并将其保存到对应的 ComfyUI 文件夹： -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — 轻量快速推理 -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — 平衡性能 -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 最佳单目深度，含天空检测 -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 物理度量深度（米级） +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — 轻量快速推理 +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — 平衡性能 +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 最佳单目深度，含天空检测 +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 物理度量深度（米级） ``` ComfyUI/ diff --git a/zh/tutorials/utility/face-detection/mediapipe.mdx b/zh/tutorials/utility/face-detection/mediapipe.mdx index 7a7289a5d..0f5817700 100644 --- a/zh/tutorials/utility/face-detection/mediapipe.mdx +++ b/zh/tutorials/utility/face-detection/mediapipe.mdx @@ -55,7 +55,7 @@ MediaPipe Face Detection 已原生集成到 ComfyUI（PR [#14009](https://github MediaPipe Face Detection 模型托管在 [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe) 上。 -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) 将其放置在以下目录结构中： diff --git a/zh/tutorials/utility/moge.mdx b/zh/tutorials/utility/moge.mdx index 0c6c7223d..df34497e1 100644 --- a/zh/tutorials/utility/moge.mdx +++ b/zh/tutorials/utility/moge.mdx @@ -50,8 +50,8 @@ ComfyUI 现已原生支持 MoGe 节点。开始前请确保已更新到最新版 下载 MoGe 检查点并保存到相应的 ComfyUI 文件夹： -- **MoGe-2（推荐）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1（基线版本）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2（推荐）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1（基线版本）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/utility/pose-detection-sdpose.mdx b/zh/tutorials/utility/pose-detection-sdpose.mdx index 81f0c1d4d..1bc6257dc 100644 --- a/zh/tutorials/utility/pose-detection-sdpose.mdx +++ b/zh/tutorials/utility/pose-detection-sdpose.mdx @@ -87,11 +87,11 @@ SDPose + RT-DETRv4 已在 ComfyUI 中原生支持（PR [#12748](https://github.c SDPose 和 RT-DETRv4 模型文件托管在 [Comfy-Org SDPose 模型仓库](https://huggingface.co/Comfy-Org/SDPose) 中。 **checkpoints**（SDPose 模型）： -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models**（RT-DETRv4 检测器）： -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors)（推荐） -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors)（全精度，体积更大） +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors)（推荐） +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors)（全精度，体积更大） 将模型放置在以下目录结构中： diff --git a/zh/tutorials/utility/remove-background-birefnet.mdx b/zh/tutorials/utility/remove-background-birefnet.mdx index 9a248d3d2..8550f8625 100644 --- a/zh/tutorials/utility/remove-background-birefnet.mdx +++ b/zh/tutorials/utility/remove-background-birefnet.mdx @@ -48,7 +48,7 @@ BiRefNet 在 ComfyUI 中获得原生支持（PR [#12747](https://github.com/Comf BiRefNet 模型托管在 [Comfy-Org BiRefNet 模型仓库](https://huggingface.co/Comfy-Org/BiRefNet)。 -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) 放置到以下目录结构： diff --git a/zh/tutorials/utility/video-segment-sam3.mdx b/zh/tutorials/utility/video-segment-sam3.mdx index 1883a1d74..7ff2ad8d1 100644 --- a/zh/tutorials/utility/video-segment-sam3.mdx +++ b/zh/tutorials/utility/video-segment-sam3.mdx @@ -68,7 +68,7 @@ SAM 3.1 能根据文本提示在视频帧中分割并追踪物体。以上示例 SAM 3.1 模型托管在 [Comfy-Org SAM 3.1 模型仓库](https://huggingface.co/Comfy-Org/sam3.1)。 -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) 放置到以下目录结构： diff --git a/zh/tutorials/utility/void-video-inpainting.mdx b/zh/tutorials/utility/void-video-inpainting.mdx index bab1a6c0a..a9a6a49d3 100644 --- a/zh/tutorials/utility/void-video-inpainting.mdx +++ b/zh/tutorials/utility/void-video-inpainting.mdx @@ -65,24 +65,24 @@ VOID 在 ComfyUI 中获得原生支持（PR [#13403](https://github.com/Comfy-Or **扩散模型** — 核心的两阶段修复模型： -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 精炼阶段，时间稳定性更佳 -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 主要阶段 +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — 精炼阶段，时间稳定性更佳 +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — 主要阶段 **VAE：** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) **光流模型：** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 分割模型：** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **文本编码器：** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ diff --git a/zh/tutorials/video/bytedance/bernini-r.mdx b/zh/tutorials/video/bytedance/bernini-r.mdx index 8506530ac..de5dc5c8b 100644 --- a/zh/tutorials/video/bytedance/bernini-r.mdx +++ b/zh/tutorials/video/bytedance/bernini-r.mdx @@ -49,16 +49,16 @@ ComfyUI 现已原生支持 Bernini-R 节点。开始前请确保已更新到最 下载所需的模型权重并将其保存到对应的 ComfyUI 文件夹： **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index a61a79890..ecafc9cbe 100644 --- a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -12,24 +12,36 @@ Cosmos-Predict2 是由 NVIDIA 推出的新一代物理世界基础模型，专 该模型具备极高的物理准确性、环境交互性和细节还原能力，能够真实模拟复杂的物理现象与动态场景。 Cosmos-Predict2 支持文本到图像（Text2Image）和视频到世界（Video2World）等多种生成方式，广泛应用于工业仿真、自动驾驶、城市规划、科学研究等领域，是推动智能视觉与物理世界深度融合的重要基础工具。 -GitHub:[Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) -huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) + + + Cosmos-Predict2 源代码和文档 + + + Cosmos-Predict2 模型集合 + + 本篇指南将引导你完成在 ComfyUI 中 **图生视频** 的工作流 对于文生图部分，请参考下面的部分 - - 使用 Cosmos-Predict2 的进行文生图 - + + + 使用 Cosmos-Predict2 进行文生图 + + + 在 Comfy Cloud 上使用强大的 GPU 运行 Cosmos-Predict2 工作流 + + -{/* ## Cosmos Predict2 Video2World 工作流 对于 2B 版本，在我们测试使用时，大约占用 16GB 的显存 -#### 1.下载工作流文件 +### 1.下载工作流文件 + +请下载以下视频并将其拖入 ComfyUI 以加载工作流。工作流已嵌入模型下载链接。 - -

下载 Json 格式工作流文件

- - + + + 下载 JSON 格式工作流文件 + + + 在 Comfy Cloud 上运行此工作流，模型预装完毕 + + 请下载下面的图片作为输入文件： ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/input.png) -### 2.手动模型安装 +### 2.手动模型安装 + +如果模型未成功下载，您可以在此部分手动下载。 **Diffusion model** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) + + cosmos_predict2_2B_video2world_480p_16fps.safetensors + 其它权重请访问 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) 进行下载 **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) + + oldt5_xxl_fp8_e4m3fn_scaled.safetensors + **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - + + wan_2.1_vae.safetensors + 文件保存位置 ``` @@ -88,4 +111,4 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- 6. (可选) 你可以在 `ClipTextEncode` 节点中修改提示词 7. (可选) 修改 `CosmosPredict2ImageToVideoLatent` 节点中的尺寸和帧数 8. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行视频生成 -9. 生成完成后对应的视频会自动保存到 `ComfyUI/output/` 目录下，你也可以在 `save video` 节点中预览或者调整保存位置 */} +9. 生成完成后对应的视频会自动保存到 `ComfyUI/output/` 目录下，你也可以在 `save video` 节点中预览或者调整保存位置 diff --git a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 6d4a77890..ff50f788c 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -24,17 +24,17 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **diffusion_models** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **vae** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) Model Storage Location diff --git a/zh/tutorials/video/hunyuan/hunyuan-video.mdx b/zh/tutorials/video/hunyuan/hunyuan-video.mdx index 6f53ebd0a..38aa3e770 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video.mdx @@ -47,9 +47,9 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 在文生视频和图生视频的工作流中下面的这些模型是共有的，请完成下载并保存到指定目录中 -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors?download=true) +- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) 保存位置： @@ -74,7 +74,7 @@ ComfyUI/ ### 2. 混元文生图模型 -请下载 [hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) 并保存至 `ComfyUI/models/diffusion_models` 文件夹中 +请下载 [hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) 并保存至 `ComfyUI/models/diffusion_models` 文件夹中 确保包括共用模型文件夹有以下完整的模型文件： @@ -126,7 +126,7 @@ ComfyUI/ ### v1 及 v2 版本共用的模型 请下载下面的文件，并保存到 `ComfyUI/models/clip_vision` 目录中 -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) +- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) ### v1 “concat” 图生视频工作流 @@ -140,7 +140,7 @@ ComfyUI/ #### 2. v1 版本模型 -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) 确保包括共用模型文件夹有以下完整的模型文件： @@ -186,7 +186,7 @@ v2 版本的工作流与 v1 版本的工作流基本相同，你只需要下载 #### 2. v2 版本模型 -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) 确保包括共用模型文件夹有以下完整的模型文件： diff --git a/zh/tutorials/video/kandinsky/kandinsky-5.mdx b/zh/tutorials/video/kandinsky/kandinsky-5.mdx index c7987f9df..03674838e 100644 --- a/zh/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/zh/tutorials/video/kandinsky/kandinsky-5.mdx @@ -51,21 +51,39 @@ Kandinsky 5.0 使用带有 Flow Matching 的潜在扩散管道，具有以下特 请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `视频` 找到 "Kandinsky 5.0 T2V" 以加载工作流。 - -

下载 JSON 格式工作流

- + + + 下载 T2V 工作流以本地使用 + + + 在 Comfy Cloud 中打开 + + ### 2. 手动下载模型 **Text Encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B 文本编码器 (FP8) + + + CLIP-L 文本编码器 + + **Diffusion Model** -- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) + + + Kandinsky 5.0 T2V Lite SFT 扩散模型 (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ @@ -85,21 +103,39 @@ ComfyUI/ 请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `视频` 找到 "Kandinsky 5.0 I2V" 以加载工作流。 - -

下载 JSON 格式工作流

- + + + 下载 I2V 工作流以本地使用 + + + 在 Comfy Cloud 中打开 + + ### 2. 手动下载模型 **Text Encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) + + + + Qwen2.5-VL 7B 文本编码器 (FP8) + + + CLIP-L 文本编码器 + + **Diffusion Model** -- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) + + + Kandinsky 5.0 I2V Lite 扩散模型 (5s) + **VAE** -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) + + + HunyuanVideo 3D VAE + ``` ComfyUI/ diff --git a/zh/tutorials/video/ltxv.mdx b/zh/tutorials/video/ltxv.mdx index a777f7cdd..890d8bfa4 100644 --- a/zh/tutorials/video/ltxv.mdx +++ b/zh/tutorials/video/ltxv.mdx @@ -26,9 +26,17 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 通过首帧图像控制视频生成：[示例首帧](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)。 - -

在 Comfy 云上运行

- + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "LTX-Video" + + + 获取此工作流的示例输入图片 + + LTX-Video 图生视频 @@ -38,6 +46,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 文生视频 + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "LTX-Video" + + + LTX-Video 文生视频 @@ -48,8 +65,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 将以下模型下载并放到指定位置： -- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true)（如尚未下载） + + 下载并放置到 ComfyUI/models/checkpoints/ + + + + 下载并放置到 ComfyUI/models/text_encoders/（如尚未下载） + ``` ├── checkpoints/ diff --git a/zh/tutorials/video/wan/fun-camera.mdx b/zh/tutorials/video/wan/fun-camera.mdx index 88be391c5..e0d9387c6 100644 --- a/zh/tutorials/video/wan/fun-camera.mdx +++ b/zh/tutorials/video/wan/fun-camera.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Performance Reference": 32425486 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 Wan2.1 Fun Camera @@ -36,21 +35,49 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面的所有模型你可以在 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 找到 -**Diffusion Models** 选择 1.3B 或 14B： -- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) -- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) +### Diffusion Models + +选择 1.3B 或 14B： + + + + Wan2.1 Fun Camera 1.3B 扩散模型 + + + Wan2.1 Fun Camera 14B 扩散模型 + + 下面的模型，如果你使用过 Wan2.1 的相关模型，那么你应该已经有了下面的模型，如果没有，请下载下面的模型： -**Text Encoders** 选择其中一个： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +### Text Encoders + +选择其中一个： + + + + 全精度文本编码器 + + + FP8 量化文本编码器（推荐用于更低显存） + + + +### VAE -**VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + + Wan2.1 VAE 模型 + + -**CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +### CLIP Vision + + + + CLIP 视觉编码器 + + 文件保存位置： @@ -70,9 +97,16 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 1.3B 原生工作流示例 -### 1. 工作流相关文件下载 +### 1. 下载工作流 -#### 1.1 工作流文件 + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Wan 2.1 Fun Camera 1.3B" + + 下载下面的视频，并拖入 ComfyUI 中以加载对应的工作流： @@ -82,21 +116,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > - -

下载 Json 格式工作流文件

- 如果你想使用 14B 版本，只需要将模型文件替换为 14B 版本即可，但请注意显存要求。 -#### 1.2 输入图片下载 - - -请下载下面的图片，我们将作为起始帧： +### 2. 下载输入素材 -![输入参考图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) + + + 下载下面的图片作为 1.3B 工作流的起始帧 + + -### 2. 按步骤完成工作流 +### 3. 按步骤完成工作流 ![Wan2.1 Fun Camera 工作流步骤](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -113,18 +145,30 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 14B 工作流及输入图片 +### 1. 下载工作流 + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Wan 2.1 Fun Camera 14B" + + + - -

下载 Json 格式工作流文件

- +### 2. 下载输入素材 -**输入图片** -![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) + + + 下载下面的图片作为 14B 工作流的起始帧 + + ## 性能参考 diff --git a/zh/tutorials/video/wan/fun-control.mdx b/zh/tutorials/video/wan/fun-control.mdx index a45cec2a1..e63c29035 100644 --- a/zh/tutorials/video/wan/fun-control.mdx +++ b/zh/tutorials/video/wan/fun-control.mdx @@ -54,18 +54,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 点击对应链接进行下载，如果你之前使用过 Wan 相关的工作流，那么你仅需要下载 **Diffusino models** **Diffusion models** 选择 1.3B 或 14B, 14B 的文件体积更大（32GB）但是对于运行显存要求也较高， -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-Control.safetensors` **Text encoders** 选择下面两个模型中的一个，fp16 精度体积较大对性能要求高 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/fun-inp.mdx b/zh/tutorials/video/wan/fun-inp.mdx index d0e8741aa..dcaa210d8 100644 --- a/zh/tutorials/video/wan/fun-inp.mdx +++ b/zh/tutorials/video/wan/fun-inp.mdx @@ -54,18 +54,18 @@ Wan-Fun InP 是阿里巴巴推出的开源视频生成模型，属于 ​​Wan2 下面的模型你可以在 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 和 [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) 找到 **Diffusion models** 选择 1.3B 或 14B, 14B 的文件体积更大（32GB）但是对于运行显存要求也较高， -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-InP.safetensors` +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-InP.safetensors` **Text encoders** 选择下面两个模型中的一个，fp16 精度体积较大对性能要求高 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/vace.mdx b/zh/tutorials/video/wan/vace.mdx index 4d6fc967f..4bd6dfe2a 100644 --- a/zh/tutorials/video/wan/vace.mdx +++ b/zh/tutorials/video/wan/vace.mdx @@ -59,19 +59,19 @@ VACE 14B 是阿里通义万相团队推出的开源视频编辑统一模型。 ### 模型下载 **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 如果你之前使用过 Wan Video 相关的工作流，下面的模型文件你已经下载过了。 **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) 从**Text encoders** 选择一个版本进行下载 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/wan-ati.mdx b/zh/tutorials/video/wan/wan-ati.mdx index 240fdb0fd..3ab064d08 100644 --- a/zh/tutorials/video/wan/wan-ati.mdx +++ b/zh/tutorials/video/wan/wan-ati.mdx @@ -47,17 +47,17 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 如果你没有成功下载工作流中的模型文件，可以尝试使用下面的链接手动下载 **Diffusion Model** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **Text encoders** Chose one of following model -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) File save location diff --git a/zh/tutorials/video/wan/wan-causal-forcing.mdx b/zh/tutorials/video/wan/wan-causal-forcing.mdx index 62324a6a1..854b7a722 100644 --- a/zh/tutorials/video/wan/wan-causal-forcing.mdx +++ b/zh/tutorials/video/wan/wan-causal-forcing.mdx @@ -83,10 +83,10 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B 检查点 - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B 检查点（最低 8GB 显存） @@ -94,10 +94,10 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### CLIP 和 VAE - + google-bert/bert-base-uncased — CLIP 文本编码器 - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/zh/tutorials/video/wan/wan-dancer.mdx b/zh/tutorials/video/wan/wan-dancer.mdx index 21ae06fa7..a03c899b5 100644 --- a/zh/tutorials/video/wan/wan-dancer.mdx +++ b/zh/tutorials/video/wan/wan-dancer.mdx @@ -61,20 +61,20 @@ Wan万相 Dancer 工作流接受两个输入：角色参考图像和一个音频 ### 3. 手动下载模型 **扩散模型** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **文本编码器** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP 视觉** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan-flf.mdx b/zh/tutorials/video/wan/wan-flf.mdx index a624005e3..55a47107c 100644 --- a/zh/tutorials/video/wan/wan-flf.mdx +++ b/zh/tutorials/video/wan/wan-flf.mdx @@ -56,7 +56,7 @@ Wan FLF2V（首尾帧视频生成）是由阿里通义万相团队推出的开 **diffusion_models** 根据你的硬件情况选择一个版本进行下载，FP8 版本对显存要求低一些 -- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8:[wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -64,14 +64,14 @@ Wan FLF2V（首尾帧视频生成）是由阿里通义万相团队推出的开 从**Text encoders** 选择一个版本进行下载， -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/wan-move.mdx b/zh/tutorials/video/wan/wan-move.mdx index 93a027390..793153134 100644 --- a/zh/tutorials/video/wan/wan-move.mdx +++ b/zh/tutorials/video/wan/wan-move.mdx @@ -28,37 +28,37 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Wan-Move 图生视频工作流 - -

下载 JSON 工作流文件

- + + 下载工作流 + - -

在 ComfyUI Cloud 上运行

- + + 在云端运行 + ## 模型下载链接 -**text_encoders** + + **text_encoders** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors + -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + **clip_vision** -- clip_vision_h.safetensors + -**clip_vision** + + **loras** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors + -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) + + **diffusion_models** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors + -**loras** - -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) - -**diffusion_models** - -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) - -**vae** - -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + **vae** -- wan_2.1_vae.safetensors + **模型存放位置** diff --git a/zh/tutorials/video/wan/wan-video.mdx b/zh/tutorials/video/wan/wan-video.mdx index 5ee12bf95..2bd273b2f 100644 --- a/zh/tutorials/video/wan/wan-video.mdx +++ b/zh/tutorials/video/wan/wan-video.mdx @@ -34,14 +34,14 @@ Wan2.1 Video 系列为阿里巴巴于 2025年2月开源的视频生成模型， 本篇指南涉及的所有模型你都可以在[这里](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)找到, 下面是本篇示例中将会使用到的共用的模型，你可以提前进行下载： 从**Text encoders** 选择一个版本进行下载， -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 @@ -64,7 +64,7 @@ ComfyUI/ ## Wan2.1 文生视频工作流 -在开始工作流前请下载 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下。 +在开始工作流前请下载 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下。 > 如果你需要其它的 t2v 精度版本，请访问[这里](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models)进行下载 @@ -101,7 +101,7 @@ ComfyUI/ ![Wan2.1 图生视频工作流 14B 480P Workflow 输入图片示例](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) #### 2. 模型下载 -请下载[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 +请下载[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 #### 3. 按步骤完成工作流的运行 @@ -129,7 +129,7 @@ ComfyUI/ #### 2. 模型下载 -请下载[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 +请下载[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 #### 3. 按步骤完成工作流的运行 diff --git a/zh/tutorials/video/wan/wan2-2-animate.mdx b/zh/tutorials/video/wan/wan2-2-animate.mdx index 23407f486..ce39bd583 100644 --- a/zh/tutorials/video/wan/wan2-2-animate.mdx +++ b/zh/tutorials/video/wan/wan2-2-animate.mdx @@ -54,13 +54,14 @@ Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架 下载以下工作流文件并将其拖入 ComfyUI 以加载工作流。 - -

下载工作流

- - - -

在 Comfy 云上运行

- + + + 在 Comfy 云上运行 + + + 下载 JSON 或在模板库中搜索 "Wan2.2 Animate" + + 下载以下素材作为输入： @@ -76,20 +77,40 @@ Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架 ### 2. 模型链接 -**diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) 这是来自 Kijai 仓库的模型 -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 原始模型权重 +**Diffusion Models** + + + + Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: 来自 Kijai 仓库的缩放 FP8 版本 + + + wan2.2_animate_14B_bf16.safetensors: 原始 bf16 模型权重 + + + +**CLIP Vision** + + + clip_vision_h.safetensors: CLIP Vision 编码器 + + +**LoRAs** + + + lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4 步加速 LoRA + + +**VAE** -**clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) + + wan_2.1_vae.safetensors: 用于编码和解码的 Wan2.1 VAE + -**loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 这是一个 4 步的加速 lora -**vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +**Text Encoders** -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: 缩放 FP8 文本编码器 + ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan2-2-fun-camera.mdx b/zh/tutorials/video/wan/wan2-2-fun-camera.mdx index 8702967cb..31fb21cf4 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -19,7 +19,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [🤗Wan2.2-Fun-A14B-Control-Camera](https://huggingface.co/alibaba-pai/Wan2.2-Fun-A14B-Control-Camera) - 代码仓库：[VideoX-Fun](https://github.com/aigc-apps/VideoX-Fun) - ## Wan2.2 Fun Camera Control 相机控制视频生成工作流示例 @@ -37,7 +36,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 由于使用了4 步 LoRA 对于初次使用工作流的用户体验较好， 但可能导致生成的视频动态会有损失， 我们默认启用了使用了加速 LoRA 版本，如果你需要启用另一组的工作流，框选后使用 **Ctrl+B** 即可启用 - ### 1. 工作流及素材下载 下载下面的视频或者 JSON 文件并拖入 ComfyUI 中以加载对应的工作流，工作流会提示下载模型 @@ -48,32 +46,60 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

下载 JSON 格式工作流

- + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Wan2.2 Fun Camera" + + 请下载下面的图片，我们将作为输入。 -![输入起始图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) + + + 视频生成的起始帧。下载并使用此图片，或替换为您自己的图片。 + + ### 2. 模型链接 下面的模型你可以在 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 找到 -**Diffusion Model** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) +**Diffusion Models** + + + + Wan2.2 Fun Camera 高噪声扩散模型 + + + Wan2.2 Fun Camera 低噪声扩散模型 + + -**Wan2.2-Lightning LoRA (可选，用于加速)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +**Wan2.2-Lightning LoRA（可选，用于加速）** + + + + 高噪声模型用的 4 步加速 LoRA + + + 低噪声模型用的 4 步加速 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + 用于编码/解码的 Wan2.1 VAE + + +**Text Encoder** + + FP8 缩放版文本编码器 + File save location @@ -92,7 +118,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. 按步骤完成工作流 ![Wan2.2 Fun Camera Control 工作流步骤](/images/tutorial/video/wan/wan_2.2_14b_fun_camera.jpg) diff --git a/zh/tutorials/video/wan/wan2-2-fun-control.mdx b/zh/tutorials/video/wan/wan2-2-fun-control.mdx index 1447efb01..147242751 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-control.mdx @@ -58,27 +58,27 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 1. 工作流及素材下载 -下载下面的视频或者 JSON 文件并拖入 ComfyUI 中以加载对应的工作流 +更新您的 ComfyUI 到最新版本，然后下载工作流文件并拖入 ComfyUI 中，或在模板库中通过 `Workflow` → `Browse Templates` → `Video` 找到 "Wan2.2 Fun Control"。 - - - -

下载 JSON 格式工作流

- + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 文件，或在模板库中搜索 "Wan2.2 Fun Control" + + 请下载下面的图片及视频，我们将作为输入。 -![输入起始图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) - - + + + 视频生成的起始帧。下载并使用此图片，或替换为您自己的图片。 + + + 预处理后的姿态控制视频。下载并使用此视频，或替换为您自己的视频。 + + > 这里我们使用了经过预处理的视频, 可以直接用于控制视频生成 @@ -86,19 +86,39 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面的模型你可以在 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 找到 -**Diffusion Model** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) +**Diffusion Models** + + + + wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors：高噪声扩散模型 + + + wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors：低噪声扩散模型 + + -** Wan2.2-Lightning LoRA (可选，用于加速)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +**Wan2.2-Lightning LoRA（可选，用于加速）** + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors：高噪声 4 步加速 LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors：低噪声 4 步加速 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors：Wan2.1 VAE，用于编码/解码 + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors：缩放 FP8 文本编码器 + File save location diff --git a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx index 50dd30fa8..97d98b2aa 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -61,13 +61,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 或者，在更新 ComfyUI 至最新版本后，下载下面的工作流并拖入 ComfyUI 中加载。 - -

下载 JSON 格式工作流

- - - -

在 Comfy 云上运行

- + + + 下载 JSON 或在模板库中搜索 "Wan2.2 Fun Inp" + + + 在 Comfy Cloud 中打开 + + 使用下面的素材作为首尾帧 @@ -76,19 +77,39 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 2. 手动下载模型 -**Diffusion Model** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) +**Diffusion Models** + + + + wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: 用于首尾帧修复的高噪点扩散模型 + + + wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: 用于首尾帧修复的低噪点扩散模型 + + + +**Lightning LoRA（可选，用于加速）** -**Lightning LoRA (可选，用于加速)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 用于高噪点模型的 4 步加速 LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 用于低噪点模型的 4 步加速 LoRA + + **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + wan_2.1_vae.safetensors: 用于编码和解码的 Wan2.1 VAE + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: 缩放后的 FP8 文本编码器 + ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan2-2-s2v.mdx b/zh/tutorials/video/wan/wan2-2-s2v.mdx index 447201527..34520f445 100644 --- a/zh/tutorials/video/wan/wan2-2-s2v.mdx +++ b/zh/tutorials/video/wan/wan2-2-s2v.mdx @@ -33,37 +33,58 @@ Wan2.2 S2V 模型仓库：[Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Wan2.2 S2V" + + 下载下面的图片及音频作为输入: -![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - -

下载输入音频

- + + + 下载默认输入图片，或使用你自己的图片。 + + + 下载默认输入音频，或使用你自己的音频。 + + ### 2. 模型链接 你可以在 [我们的仓库](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 中找到所有模型。 -**diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +**diffusion_models** + + + + FP8 scaled diffusion model。放置在 ComfyUI/models/diffusion_models/ + + + BF16 diffusion model。放置在 ComfyUI/models/diffusion_models/ + + **audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) + + + Audio encoder model。放置在 ComfyUI/models/audio_encoders/ + **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -**text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + Wan2.1 VAE model。放置在 ComfyUI/models/vae/ + + +**text_encoders** + + + FP8 scaled text encoder。放置在 ComfyUI/models/text_encoders/ + ``` @@ -92,8 +113,14 @@ ComfyUI/ 你可以在 [这里](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) 找到两种模型： -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) + + + FP8 scaled diffusion model + + + BF16 diffusion model + + 本模板使用 `wan2.2_s2v_14B_fp8_scaled.safetensors`，它需要更少的显存。但你可以尝试 `wan2.2_s2v_14B_bf16.safetensors` 来减少质量损失。 diff --git a/zh/tutorials/video/wan/wan2_2.mdx b/zh/tutorials/video/wan/wan2_2.mdx index 3c2588525..72e167ffd 100644 --- a/zh/tutorials/video/wan/wan2_2.mdx +++ b/zh/tutorials/video/wan/wan2_2.mdx @@ -101,24 +101,25 @@ Wan2.2 5B 版本配合 ComfyUI 原生 offloading功能，能很好地适配 8GB src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - -

下载 JSON 格式工作流

- - - -

Run on Comfy Cloud

- + + + 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 5B" + + + 在 Comfy Cloud 中打开 + + ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) +- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) +- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan2.2_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -156,25 +157,26 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - -

下载 JSON 格式工作流

- - - -

Run on Comfy Cloud

- + + + 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 14B T2V" + + + 在 Comfy Cloud 中打开 + + ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -215,13 +217,14 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - -

下载 JSON 格式工作流

- - - -

Run on Comfy Cloud

- + + + 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 14B I2V" + + + 在 Comfy Cloud 中打开 + + 你可以使用下面的图片作为输入 ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -229,14 +232,14 @@ ComfyUI/ ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) +- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -274,13 +277,14 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - -

下载 JSON 格式工作流

- - - -

Run on Comfy Cloud

- + + + 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 14B FLF2V" + + + 在 Comfy Cloud 中打开 + + 下载下面的素材作为输入 diff --git a/zh/tutorials/video/zai/scail2.mdx b/zh/tutorials/video/zai/scail2.mdx index ff1e51c4d..03684d603 100644 --- a/zh/tutorials/video/zai/scail2.mdx +++ b/zh/tutorials/video/zai/scail2.mdx @@ -105,23 +105,23 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 所需模型 **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders**（选择其一） -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### 文件存放位置 From 2e044271e30491f487b4e4baf31dfd0f4523bb56 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 21:08:15 +0800 Subject: [PATCH 02/19] fix: add preview images, input material cards, model storage paths --- tutorials/video/kandinsky/kandinsky-5.mdx | 42 ++++++++------ tutorials/video/wan/fun-control.mdx | 39 ++++++++----- tutorials/video/wan/fun-inp.mdx | 42 +++++++------- tutorials/video/wan/vace.mdx | 58 ++++++++++++++++++-- tutorials/video/wan/wan-ati.mdx | 67 +++++++++++++---------- tutorials/video/wan/wan-flf.mdx | 48 ++++++++-------- tutorials/video/wan/wan-move.mdx | 59 ++++++++++++-------- tutorials/video/wan/wan-video.mdx | 38 +++++++------ 8 files changed, 244 insertions(+), 149 deletions(-) diff --git a/tutorials/video/kandinsky/kandinsky-5.mdx b/tutorials/video/kandinsky/kandinsky-5.mdx index c6a91682b..710d04f8b 100644 --- a/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/tutorials/video/kandinsky/kandinsky-5.mdx @@ -41,6 +41,8 @@ All models are available in 5-second and 10-second video generation versions. Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 T2V" to load the workflow. +![Kandinsky 5.0 T2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_t2v-1.webp) + Download the T2V workflow to use locally @@ -55,24 +57,24 @@ Please update your ComfyUI to the latest version, and through the menu `Workflow **Text Encoders** - - Qwen2.5-VL 7B text encoder (FP8) + + Qwen2.5-VL 7B text encoder (FP8). Place in ComfyUI/models/text_encoders/ - - CLIP-L text encoder + + CLIP-L text encoder. Place in ComfyUI/models/text_encoders/ **Diffusion Model** - - Kandinsky 5.0 T2V Lite SFT diffusion model (5s) + + Kandinsky 5.0 T2V Lite SFT diffusion model (5s). Place in ComfyUI/models/diffusion_models/ **VAE** - - HunyuanVideo 3D VAE + + HunyuanVideo 3D VAE. Place in ComfyUI/models/vae/ ``` @@ -93,6 +95,8 @@ ComfyUI/ Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 I2V" to load the workflow. +![Kandinsky 5.0 I2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_i2v-1.webp) + Download the I2V workflow to use locally @@ -102,29 +106,35 @@ Please update your ComfyUI to the latest version, and through the menu `Workflow +**Input Image** + + + Default input image for the I2V workflow. Download and use this image, or replace with your own. + + ### 2. Manually download models **Text Encoders** - - Qwen2.5-VL 7B text encoder (FP8) + + Qwen2.5-VL 7B text encoder (FP8). Place in ComfyUI/models/text_encoders/ - - CLIP-L text encoder + + CLIP-L text encoder. Place in ComfyUI/models/text_encoders/ **Diffusion Model** - - Kandinsky 5.0 I2V Lite diffusion model (5s) + + Kandinsky 5.0 I2V Lite diffusion model (5s). Place in ComfyUI/models/diffusion_models/ **VAE** - - HunyuanVideo 3D VAE + + HunyuanVideo 3D VAE. Place in ComfyUI/models/vae/ ``` diff --git a/tutorials/video/wan/fun-control.mdx b/tutorials/video/wan/fun-control.mdx index f6dbbc9fb..e2966606c 100644 --- a/tutorials/video/wan/fun-control.mdx +++ b/tutorials/video/wan/fun-control.mdx @@ -18,10 +18,18 @@ Model versions: - **1.3B** Lightweight: Suitable for local deployment and quick inference with **lower VRAM requirements** - **14B** High-performance: Model size reaches 32GB+, offering better results but **requiring higher VRAM** -Here are the relevant code repositories: -- [Wan2.1-Fun-1.3B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-1.3B-Control) -- [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control) -- Code repository: [VideoX-Fun](https://github.com/aigc-apps/VideoX-Fun) +**Model Weights Download**: + + + + Lightweight version, lower VRAM requirements + + + High-performance version, 32GB+ file size, higher VRAM requirements + + + +**Code Repository**: [VideoX-Fun](https://github.com/aigc-apps/VideoX-Fun) ComfyUI now **natively supports** the Wan2.1 Fun Control model. Before starting this tutorial, please update your ComfyUI to ensure you're using a version after [this commit](https://github.com/Comfy-Org/ComfyUI/commit/3661c833bcc41b788a7c9f0e7bc48524f8ee5f82). @@ -83,13 +91,14 @@ File storage location: 📂 ComfyUI/ ├── 📂 models/ │ ├── 📂 diffusion_models/ -│ │ └── wan2.1_fun_control_1.3B_bf16.safetensors +│ │ ├── wan2.1_fun_control_1.3B_bf16.safetensors +│ │ └── Wan2.1-Fun-14B-Control.safetensors │ ├── 📂 text_encoders/ -│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors -│ └── 📂 vae/ +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── 📂 vae/ │ │ └── wan_2.1_vae.safetensors │ └── 📂 clip_vision/ -│ └── clip_vision_h.safetensors +│ └── clip_vision_h.safetensors ``` ## ComfyUI Native Workflow @@ -101,6 +110,8 @@ Thanks to our powerful ComfyUI authors who provide feature-rich nodes. If you wa ### 1. Workflow File Download +![Wan2.1 Fun Control Native Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/wan2.1_fun_control_native.webp) + Open in Comfy Cloud @@ -113,8 +124,8 @@ Thanks to our powerful ComfyUI authors who provide feature-rich nodes. If you wa #### Input Materials - - Input reference image for the native workflow. Download and use this image, or replace with your own. + + Input start frame for the native workflow. Download and use this image, or replace with your own. Input control video in WebP format for the native workflow. The native workflow requires WebP format since Load Image does not support mp4. @@ -153,6 +164,8 @@ You can use [ComfyUI Manager](https://github.com/Comfy-Org/ComfyUI-Manager) to i ### 1. Workflow File Download +![Wan2.1 Fun Control Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/wan2.1_fun_control-1.webp) + Open in Comfy Cloud @@ -165,10 +178,10 @@ You can use [ComfyUI Manager](https://github.com/Comfy-Org/ComfyUI-Manager) to i #### Input Materials - - Input reference image for the custom nodes workflow. Download and use this image, or replace with your own. + + Input start frame for the custom nodes workflow. Download and use this image, or replace with your own. - + Input control video in MP4 format for the custom nodes workflow. The custom Load Video node supports mp4 format. diff --git a/tutorials/video/wan/fun-inp.mdx b/tutorials/video/wan/fun-inp.mdx index 655087099..d0770f209 100644 --- a/tutorials/video/wan/fun-inp.mdx +++ b/tutorials/video/wan/fun-inp.mdx @@ -34,6 +34,10 @@ Currently, ComfyUI natively supports the Wan2.1 Fun InP model. Before starting t ### 1. Download Workflow +Download the image below and drag it into ComfyUI to load the workflow: + +![Wan2.1 Fun InP Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/wan2.1_fun_inp-1.webp) + Open in Comfy Cloud @@ -43,10 +47,6 @@ Currently, ComfyUI natively supports the Wan2.1 Fun InP model. Before starting t -Download the image below and drag it into ComfyUI to load the workflow: - -![Workflow File](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_inp/wan2.1_fun_inp.webp) - Starting frame for the video generation. Download and use this image, or replace with your own. @@ -65,35 +65,35 @@ All models involved in this guide can be found at [Wan_2.1_ComfyUI_repackaged](h **Diffusion Models** — Choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: - - Wan2.1 Fun InP 1.3B diffusion model — lower VRAM requirements + + `models/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors` - - Wan2.1 Fun InP 14B diffusion model — rename to Wan2.1-Fun-14B-InP.safetensors after downloading + + `models/diffusion_models/Wan2.1-Fun-14B-InP.safetensors` (rename after download) **Text Encoders** — Choose one of the following models (fp16 precision has a larger size and higher performance requirements): - - Full precision text encoder + + `models/text_encoders/umt5_xxl_fp16.safetensors` - - FP8 quantized text encoder (recommended for lower VRAM) + + `models/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors` **VAE** - - Wan2.1 VAE for encoding/decoding + + `models/vae/wan_2.1_vae.safetensors` **CLIP Vision** - - CLIP vision encoder + + `models/clip_vision/clip_vision_h.safetensors` File storage location: @@ -101,13 +101,15 @@ File storage location: 📂 ComfyUI/ ├── 📂 models/ │ ├── 📂 diffusion_models/ -│ │ └── wan2.1_fun_inp_1.3B_bf16.safetensors +│ │ ├── wan2.1_fun_inp_1.3B_bf16.safetensors +│ │ └── Wan2.1-Fun-14B-InP.safetensors │ ├── 📂 text_encoders/ -│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors -│ └── 📂 vae/ +│ │ ├── umt5_xxl_fp16.safetensors +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── 📂 vae/ │ │ └── wan_2.1_vae.safetensors │ └── 📂 clip_vision/ -│ └── clip_vision_h.safetensors +│ └── clip_vision_h.safetensors ``` ### 3. Complete the Workflow Step by Step diff --git a/tutorials/video/wan/vace.mdx b/tutorials/video/wan/vace.mdx index 6f74256f6..82dbb9fd3 100644 --- a/tutorials/video/wan/vace.mdx +++ b/tutorials/video/wan/vace.mdx @@ -63,10 +63,10 @@ The model download information is already embedded in the workflow information i - VACE 14B diffusion model (recommended, ~32GB) + VACE 14B diffusion model (recommended, ~32GB). Place in ComfyUI/models/diffusion_models/ - VACE 1.3B diffusion model (lighter, ~7GB) + VACE 1.3B diffusion model (lighter, ~7GB). Place in ComfyUI/models/diffusion_models/ @@ -78,7 +78,7 @@ If you have used Wan Video related workflows before, you have already downloaded - Wan VAE model + Wan VAE model. Place in ComfyUI/models/vae/ @@ -86,10 +86,10 @@ If you have used Wan Video related workflows before, you have already downloaded - Full precision text encoder (higher quality, larger size) + Full precision text encoder (higher quality, larger size). Place in ComfyUI/models/text_encoders/ - FP8 text encoder (lower VRAM usage) + FP8 text encoder (lower VRAM usage). Place in ComfyUI/models/text_encoders/ @@ -190,6 +190,12 @@ You can continue using the workflow above, just unbypass the `Load image` node i Please download the image below as input: + + + Default input image for I2V. Download and use in the Load Image node, or provide your own image. + + + ![vace-i2v-input](https://github.com/Comfy-Org/example_workflows/raw/refs/heads/main/video/wan/vace/i2v/input.jpg) ### Steps to Run @@ -235,6 +241,15 @@ VACE also supports inputting multiple reference images in a single image to gene We will use the following materials as input: + + + Input reference image for style and content. + + + Preprocessed control video (depth). Use directly in Load Video under Load control video. + + + 1. Input image for reference: ![v2v-input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/vace/v2v/input.jpg) @@ -284,6 +299,20 @@ Please follow the numbered steps in the image to ensure smooth workflow executio +### Input Materials + + + + Input video for inpainting. Load in the Load Video node. + + + First frame with mask overlay for inpainting region. + + + Reference image for inpainting content. + + + [To be updated] --- @@ -301,6 +330,14 @@ Please follow the numbered steps in the image to ensure smooth workflow executio +### Input Materials + + + + Input video for outpainting. Load in the Load Video node. + + + [To be updated] --- @@ -318,6 +355,17 @@ Please follow the numbered steps in the image to ensure smooth workflow executio +### Input Materials + + + + Starting frame image. Load in the Load Image node. + + + Ending frame image. Load in the Load Image node. + + + To ensure that the first and last frames are effective, the video `length` setting must satisfy that `length-1` is divisible by 4. The corresponding `Batch_size` setting must satisfy `Batch_size = length - 2`. diff --git a/tutorials/video/wan/wan-ati.mdx b/tutorials/video/wan/wan-ati.mdx index 0e0628782..333de9cd6 100644 --- a/tutorials/video/wan/wan-ati.mdx +++ b/tutorials/video/wan/wan-ati.mdx @@ -30,14 +30,16 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ### 1. Workflow Download -Download the video below and drag it into ComfyUI to load the corresponding workflow. +Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan2.1 ATI" in the Template Library under `Workflow` → `Browse Templates` → `Video`. + +![Wan2.1 ATI Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_wan_ati-1.webp) Open in Comfy Cloud - - Download the workflow video and drag it into ComfyUI to load the workflow + + Download JSON or search "Wan2.1 ATI" in Template Library @@ -48,7 +50,12 @@ Download the video below and drag it into ComfyUI to load the corresponding work > We will use the following image as input: -![v2v-input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/ati/input.jpg) + + + + Download the default input image, or use your own image as the starting frame. + + -### 2. Model Download - -If you haven't successfully downloaded the model files from the workflow, you can try downloading them manually using the links below. - -#### Diffusion Model - - - Wan2.1 I2V ATI 14B diffusion model (fp8 precision) - +### 2. Manual Model Installation -#### VAE +All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files). - - Wan2.1 VAE model - +**Diffusion Model** — Choose one version: -#### Text Encoders + + + Wan2.1 I2V ATI 14B diffusion model (fp8 precision). Place in ComfyUI/models/diffusion_models/ + + -Choose one of the following models: +**Text Encoders** — Choose one version: - - Full precision text encoder (higher quality, larger size) + + Full precision text encoder (higher quality, larger size). Place in ComfyUI/models/text_encoders/ - - FP8 text encoder (lower VRAM usage) + + FP8 text encoder (lower VRAM usage). Place in ComfyUI/models/text_encoders/ -#### CLIP Vision +**VAE** + + + Wan2.1 VAE model. Place in ComfyUI/models/vae/ + + +**CLIP Vision** - - CLIP Vision model for processing reference images + + CLIP Vision model for processing reference images. Place in ComfyUI/models/clip_vision/ -#### File Save Location +**File Storage Location** ``` 📂 ComfyUI/ @@ -99,11 +106,11 @@ Choose one of the following models: │ ├── 📂 diffusion_models/ │ │ └── Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors │ ├── 📂 text_encoders/ -│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors # or other version +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors # or your chosen version │ ├── 📂 clip_vision/ -│ │ └─── clip_vision_h.safetensors +│ │ └── clip_vision_h.safetensors │ └── 📂 vae/ -│ └── wan_2.1_vae.safetensors +│ └── wan_2.1_vae.safetensors ``` ### 3. Complete the Workflow Execution Step by Step diff --git a/tutorials/video/wan/wan-flf.mdx b/tutorials/video/wan/wan-flf.mdx index af781a447..c7305710d 100644 --- a/tutorials/video/wan/wan-flf.mdx +++ b/tutorials/video/wan/wan-flf.mdx @@ -41,6 +41,8 @@ If needed, please adjust the video generation size for testing. A small generati Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan2.1 FLF2V 720P" in the Template Library under `Workflow` → `Browse Templates` → `Video`. +![Wan2.1 FLF2V 720P Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/wan2.1_flf2v_720_f16-1.webp) + Open in Comfy Cloud @@ -51,11 +53,11 @@ Update your ComfyUI to the latest version, then download and drag the workflow f - - Starting frame for the video generation. Download and use this image, or replace with your own. + + Starting frame for the video generation (LoadImage node 52). Download and use this image, or replace with your own. - - Ending frame for the video generation. Download and use this image, or replace with your own. + + Ending frame for the video generation (LoadImage node 72). Download and use this image, or replace with your own. @@ -66,11 +68,11 @@ All models involved in this guide can be found [here](https://huggingface.co/Com **Diffusion Models** — Choose one version based on your hardware - - wan2.1_flf2v_720p_14B_fp16.safetensors — Full precision, requires more VRAM + + wan2.1_flf2v_720p_14B_fp16.safetensors — Full precision, requires more VRAM. Place in ComfyUI/models/diffusion_models/ - - wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors — Quantized version, lower VRAM usage + + wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors — Quantized version, lower VRAM usage. Place in ComfyUI/models/diffusion_models/ @@ -81,38 +83,38 @@ If you have previously tried Wan Video related workflows, you may already have t **Text Encoders** — Choose one version - - umt5_xxl_fp16.safetensors — Full precision text encoder + + umt5_xxl_fp16.safetensors — Full precision text encoder. Place in ComfyUI/models/text_encoders/ - - umt5_xxl_fp8_e4m3fn_scaled.safetensors — Quantized text encoder + + umt5_xxl_fp8_e4m3fn_scaled.safetensors — Quantized text encoder. Place in ComfyUI/models/text_encoders/ **VAE** - - wan_2.1_vae.safetensors — Wan2.1 VAE for encoding/decoding + + wan_2.1_vae.safetensors — Wan2.1 VAE for encoding/decoding. Place in ComfyUI/models/vae/ **CLIP Vision** - - clip_vision_h.safetensors — CLIP Vision encoder + + clip_vision_h.safetensors — CLIP Vision encoder. Place in ComfyUI/models/clip_vision/ File Storage Location ``` ComfyUI/ -├── models/ -│ ├── diffusion_models/ +├───📂 models/ +│ ├───📂 diffusion_models/ │ │ └─── wan2.1_flf2v_720p_14B_fp16.safetensors # or FP8 version -│ ├── text_encoders/ +│ ├───📂 text_encoders/ │ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors # or your chosen version -│ ├── vae/ -│ │ └── wan_2.1_vae.safetensors -│ └── clip_vision/ -│ └── clip_vision_h.safetensors +│ ├───📂 vae/ +│ │ └─── wan_2.1_vae.safetensors +│ └───📂 clip_vision/ +│ └─── clip_vision_h.safetensors ``` ### 3. Complete Workflow Execution Step by Step diff --git a/tutorials/video/wan/wan-move.mdx b/tutorials/video/wan/wan-move.mdx index 1db485be7..e39bd8b62 100644 --- a/tutorials/video/wan/wan-move.mdx +++ b/tutorials/video/wan/wan-move.mdx @@ -21,37 +21,48 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Wan-Move image-to-video workflow - - Download workflow - +Preview the workflow output: - - Open in cloud - +![Wan-Move preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_wanmove_480p-1.webp) - - -## Model links + + + Open in Comfy Cloud + + + Download JSON or search "Wan-Move Motion-Control" in Template Library + + - - **text_encoders** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors - +### Input Image - - **clip_vision** -- clip_vision_h.safetensors - + + + Download the default input image, or use your own image as the starting frame. + + - - **loras** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors - + - - **diffusion_models** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors - +## Model links - - **vae** -- wan_2.1_vae.safetensors - + + + Place in ComfyUI/models/diffusion_models/ + + + Place in ComfyUI/models/loras/ + + + Place in ComfyUI/models/text_encoders/ + + + Place in ComfyUI/models/clip_vision/ + + + Place in ComfyUI/models/vae/ + + **Model Storage Location** diff --git a/tutorials/video/wan/wan-video.mdx b/tutorials/video/wan/wan-video.mdx index 6514a89ac..03ad0037a 100644 --- a/tutorials/video/wan/wan-video.mdx +++ b/tutorials/video/wan/wan-video.mdx @@ -60,7 +60,7 @@ File storage locations: ComfyUI/ ├── models/ │ ├── diffusion_models/ -│ │ └── ... (download per workflow) +│ │ └── ... (download per workflow below) │ ├── text_encoders/ │ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors │ ├── vae/ @@ -75,7 +75,12 @@ For diffusion models, we'll use the fp16 precision models in this guide because ## Wan2.1 Text-to-Video Workflow (1.3B) - +![Wan2.1 Text-to-Video Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_t2v_1.3b.webp) + + + + Open in Comfy Cloud + Download the workflow image and drag it into ComfyUI to load the workflow @@ -89,8 +94,6 @@ For diffusion models, we'll use the fp16 precision models in this guide because > If you need other t2v precision versions, please visit [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) to download them. -![Wan2.1 Text-to-Video Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_t2v_1.3b.webp) - ### Steps to Run ![ComfyUI Wan2.1 Workflow Steps](/images/tutorial/video/wan/wan2.1_t2v_1.3b_flow_diagram.jpg) @@ -108,8 +111,10 @@ For diffusion models, we'll use the fp16 precision models in this guide because ### 480P Version +![Wan2.1 Image-to-Video Workflow 14B 480P](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_480P.webp) + - + Open in Comfy Cloud @@ -117,13 +122,9 @@ For diffusion models, we'll use the fp16 precision models in this guide because - - + Download the default input image, or use your own image. - - - -![Wan2.1 Image-to-Video Workflow 14B 480P](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_480P.webp) + #### Model Downloads @@ -146,19 +147,20 @@ For diffusion models, we'll use the fp16 precision models in this guide because ### 720P Version - +![Wan2.1 Image-to-Video Workflow 14B 720P](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_720P.webp) + + + + Open in Comfy Cloud + Download the workflow image and drag it into ComfyUI to load the workflow - - + Download the default input image, or use your own image. - - - -![Wan2.1 Image-to-Video Workflow 14B 720P](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/wan2.1_i2v_14b_720P.webp) + #### Model Downloads From fe29d4a810b1ec203bc8dbf7abb42e559b46bc9e Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 21:09:35 +0800 Subject: [PATCH 03/19] fix: add previews, input cards, model paths to hunyuan and kandinsky --- tutorials/video/hunyuan/hunyuan-video-1-5.mdx | 113 +++++++++--------- tutorials/video/hunyuan/hunyuan-video.mdx | 92 +++++++++----- 2 files changed, 119 insertions(+), 86 deletions(-) diff --git a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index a5a2d5802..5aa507f36 100644 --- a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -22,27 +22,27 @@ import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; The following models are used in both Text-to-Video and Image-to-Video workflows. Download and save them to the specified directories: - - qwen_2.5_vl_7b_fp8_scaled.safetensors — text encoder shared across all workflows - - - - byt5_small_glyphxl_fp16.safetensors — text encoder shared across all workflows - - - - hunyuanvideo15_vae_fp16.safetensors — VAE shared across all workflows - + + + qwen_2.5_vl_7b_fp8_scaled.safetensors — save to ComfyUI/models/text_encoders/ + + + byt5_small_glyphxl_fp16.safetensors — save to ComfyUI/models/text_encoders/ + + + hunyuanvideo15_vae_fp16.safetensors — save to ComfyUI/models/vae/ + + -Storage location: +#### Storage ``` ComfyUI/ -├── models/ -│ ├── text_encoders/ +├── 📂 models/ +│ ├── 📂 text_encoders/ │ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors │ │ └── byt5_small_glyphxl_fp16.safetensors -│ └── vae/ +│ └── 📂 vae/ │ └── hunyuanvideo15_vae_fp16.safetensors ``` @@ -50,41 +50,40 @@ ComfyUI/ HunyuanVideo 1.5 Text-to-Video generates 5-10 second videos from natural language descriptions with enhanced quality and reduced VRAM requirements. -### 1. Workflow +![ComfyUI Workflow - Hunyuan Video 1.5 T2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_t2v-1.webp) Open in Comfy Cloud - Download JSON or search "Hunyuan Video 1.5 T2V" in Template Library + Download JSON or search "Hunyuan Video 1.5 T2V" in Template Library -Download the image below and drag it into ComfyUI to load the workflow: -![ComfyUI Workflow - Hunyuan Video 1.5 T2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_t2v-1.webp) - -### 2. Manual models installation +### Model downloads - - hunyuanvideo1.5_720p_t2v_fp16.safetensors — save to ComfyUI/models/diffusion_models - + + + hunyuanvideo1.5_720p_t2v_fp16.safetensors — save to ComfyUI/models/diffusion_models/ + + -Ensure you have all these model files in the correct locations: +#### Model storage ``` ComfyUI/ -├── models/ -│ ├── text_encoders/ +├── 📂 models/ +│ ├── 📂 text_encoders/ │ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // Shared model │ │ └── byt5_small_glyphxl_fp16.safetensors // Shared model -│ ├── vae/ +│ ├── 📂 vae/ │ │ └── hunyuanvideo15_vae_fp16.safetensors // Shared model -│ └── diffusion_models/ +│ └── 📂 diffusion_models/ │ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V model ``` -### 3. Steps to run the workflow +### Steps to run the workflow 1. Ensure the `DualCLIPLoader` node has loaded these models: - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` @@ -101,47 +100,53 @@ The workflow includes a super-resolution upscaler node. When enabled, it uses `h HunyuanVideo 1.5 Image-to-Video transforms static images into smooth, high-quality videos with improved consistency and motion dynamics. -### 1. Workflow +![ComfyUI Workflow - Hunyuan Video 1.5 I2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_i2v-1.webp) Open in Comfy Cloud - Download JSON or search "Hunyuan Video 1.5 I2V" in Template Library + Download JSON or search "Hunyuan Video 1.5 I2V" in Template Library -Download the image below and drag it into ComfyUI to load the workflow: -![ComfyUI Workflow - Hunyuan Video 1.5 I2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_i2v-1.webp) +#### Input materials -### 2. Related models manual installation + + + Download the default input image, or use your own image as the starting frame. + + - - sigclip_vision_patch14_384.safetensors — save to ComfyUI/models/clip_vision - +### Model downloads - - hunyuanvideo1.5_720p_i2v_fp16.safetensors — save to ComfyUI/models/diffusion_models - + + + sigclip_vision_patch14_384.safetensors — save to ComfyUI/models/clip_vision/ + + + hunyuanvideo1.5_720p_i2v_fp16.safetensors — save to ComfyUI/models/diffusion_models/ + + -Ensure you have all these model files in the correct locations: +#### Model storage ``` ComfyUI/ -├── models/ -│ ├── clip_vision/ +├── 📂 models/ +│ ├── 📂 clip_vision/ │ │ └── sigclip_vision_patch14_384.safetensors // I2V vision encoder -│ ├── text_encoders/ +│ ├── 📂 text_encoders/ │ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // Shared model │ │ └── byt5_small_glyphxl_fp16.safetensors // Shared model -│ ├── vae/ +│ ├── 📂 vae/ │ │ └── hunyuanvideo15_vae_fp16.safetensors // Shared model -│ └── diffusion_models/ +│ └── 📂 diffusion_models/ │ └── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V model ``` -### 3. Steps to run the workflow +### Steps to run the workflow 1. Ensure the `DualCLIPLoader` node has loaded these models: - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` @@ -156,22 +161,22 @@ ComfyUI/ Both workflows include a super-resolution node that can upscale the output video from 720p to 1080p. This optional upscaler uses a distilled model for efficient high-resolution output. - hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors — save to ComfyUI/models/diffusion_models + hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors — save to ComfyUI/models/diffusion_models/ All shared and workflow-specific models combined: ``` ComfyUI/ -├── models/ -│ ├── clip_vision/ +├── 📂 models/ +│ ├── 📂 clip_vision/ │ │ └── sigclip_vision_patch14_384.safetensors -│ ├── text_encoders/ +│ ├── 📂 text_encoders/ │ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors │ │ └── byt5_small_glyphxl_fp16.safetensors -│ ├── vae/ +│ ├── 📂 vae/ │ │ └── hunyuanvideo15_vae_fp16.safetensors -│ └── diffusion_models/ +│ └── 📂 diffusion_models/ │ ├── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V model │ ├── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V model │ └── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors // Upscaler diff --git a/tutorials/video/hunyuan/hunyuan-video.mdx b/tutorials/video/hunyuan/hunyuan-video.mdx index ccefbc727..51d4e77f8 100644 --- a/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/tutorials/video/hunyuan/hunyuan-video.mdx @@ -37,16 +37,17 @@ Alternatively, this guide provides direct model links if automatic downloads fai The following models are used in both Text-to-Video and Image-to-Video workflows. Download and save them to the specified directories: - - clip_l.safetensors — text encoder shared across all workflows - - - - llava_llama3_fp8_scaled.safetensors — text encoder shared across all workflows - + + + Save to ComfyUI/models/text_encoders + + + Save to ComfyUI/models/text_encoders + + - - hunyuan_video_vae_bf16.safetensors — VAE shared across all workflows + + Save to ComfyUI/models/vae Storage location: @@ -57,8 +58,8 @@ ComfyUI/ │ ├── text_encoders/ │ │ ├── clip_l.safetensors │ │ └── llava_llama3_fp8_scaled.safetensors -│ ├── vae/ -│ │ └── hunyuan_video_vae_bf16.safetensors +│ └── vae/ +│ └── hunyuan_video_vae_bf16.safetensors ``` ## Hunyuan Text-to-Video Workflow @@ -67,6 +68,10 @@ Hunyuan Text-to-Video was open-sourced in December 2024, supporting 5-second sho ### 1. Workflow +Download the workflow image below and drag it into ComfyUI to load the workflow: + +![ComfyUI Workflow - Hunyuan Text-to-Video](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/hunyuan_video_text_to_video-1.webp) + Open in Comfy Cloud @@ -76,13 +81,10 @@ Hunyuan Text-to-Video was open-sourced in December 2024, supporting 5-second sho -Download the image below and drag it into ComfyUI to load the workflow: -![ComfyUI Workflow - Hunyuan Text-to-Video](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/t2v/kitchen.webp) - -### 2. Manual Models Installation +### 2. Manual models installation - - hunyuan_video_t2v_720p_bf16.safetensors — save to ComfyUI/models/diffusion_models + + Save to ComfyUI/models/diffusion_models Ensure you have all these model files in the correct locations: @@ -99,7 +101,7 @@ ComfyUI/ │ └── hunyuan_video_t2v_720p_bf16.safetensors // T2V model ``` -### 3. Steps to Run the Workflow +### 3. Steps to run the workflow ![ComfyUI Hunyuan Video T2V Workflow](/images/tutorial/advanced/hunyuanvideo/flow_diagram_t2v.jpg) @@ -135,8 +137,8 @@ Currently, the Hunyuan Image-to-Video model has two versions: ### Shared Model for v1 and v2 Versions - - llava_llama3_vision.safetensors — save to ComfyUI/models/clip_vision + + Save to ComfyUI/models/clip_vision ### V1 "concat" Image-to-Video Workflow @@ -144,15 +146,28 @@ Currently, the Hunyuan Image-to-Video model has two versions: #### 1. Workflow and Asset Download the workflow image below and drag it into ComfyUI to load the workflow: + ![ComfyUI Workflow - Hunyuan Image-to-Video v1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/i2v/v1_robot.webp) -Download the image below, which we will use as the starting frame for the image-to-video generation: -![Starting Frame](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/hunyuan-video/i2v/robot-ballet.png) + + + Download the workflow image and drag into ComfyUI + + + Open in Comfy Cloud + + + +Download the image below and use it as the starting frame for the image-to-video generation: -#### 2. Related Models Manual Installation + + Save and use as input image for I2V generation + + +#### 2. Related models manual installation - - hunyuan_video_image_to_video_720p_bf16.safetensors — save to ComfyUI/models/diffusion_models + + Save to ComfyUI/models/diffusion_models Ensure you have all these model files in the correct locations: @@ -171,7 +186,7 @@ ComfyUI/ │ └── hunyuan_video_image_to_video_720p_bf16.safetensors // I2V v1 "concat" version model ``` -#### 3. Steps to Run the Workflow +#### 3. Steps to run the workflow ![ComfyUI Hunyuan Video I2V v1 Workflow](/images/tutorial/advanced/hunyuanvideo/flow_diagram_i2v_v1.jpg) @@ -191,15 +206,28 @@ The v2 workflow is essentially the same as the v1 workflow. You just need to dow #### 1. Workflow and Asset Download the workflow image below and drag it into ComfyUI to load the workflow: + ![ComfyUI Workflow - Hunyuan Image-to-Video v2](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/i2v/v2_fennec_gril.webp) -Download the image below, which we will use as the starting frame for the image-to-video generation: -![Starting Frame](https://comfyanonymous.github.io/ComfyUI_examples/flux/flux_dev_example.png) + + + Download the workflow image and drag into ComfyUI + + + Open in Comfy Cloud + + + +Download the image below and use it as the starting frame for the image-to-video generation: + + + Save and use as input image for I2V generation + -#### 2. Related Models Manual Installation +#### 2. Related models manual installation - - hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors — save to ComfyUI/models/diffusion_models + + Save to ComfyUI/models/diffusion_models Ensure you have all these model files in the correct locations: @@ -218,7 +246,7 @@ ComfyUI/ │ └── hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors // V2 "replace" version model ``` -#### 3. Steps to Run the Workflow +#### 3. Steps to run the workflow ![ComfyUI Hunyuan Video I2V v2 Workflow](/images/tutorial/advanced/hunyuanvideo/flow_diagram_i2v_v2.jpg) From faf5430cb0e6475ca806fb5880fa3d252de1d3a1 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 21:58:00 +0800 Subject: [PATCH 04/19] fix: add missing Cloud cards to 15+ tutorial pages --- tutorials/flux/flux-2-klein.mdx | 24 ++++++++++ .../partner-nodes/bria/background-removal.mdx | 10 ++--- .../bytedance/seedance-2-0-real-human.mdx | 8 ++++ tutorials/partner-nodes/google/gemini.mdx | 5 ++- .../happyhorse/happyhorse1-0.mdx | 44 ++++++++++--------- .../kling/kling-motion-control.mdx | 10 +++-- tutorials/partner-nodes/luma/luma-uni-1.mdx | 14 +++--- tutorials/partner-nodes/meshy/meshy-6.mdx | 18 ++++---- tutorials/partner-nodes/openai/chat.mdx | 11 +++-- .../partner-nodes/openai/gpt-image-2.mdx | 4 ++ .../partner-nodes/rodin/model-generation.mdx | 22 +++++++--- .../partner-nodes/tripo/model-generation.mdx | 33 ++++++++++---- tutorials/partner-nodes/tripo/tripo-3-1.mdx | 21 +++++---- tutorials/utility/moge.mdx | 24 +++++++--- 14 files changed, 171 insertions(+), 77 deletions(-) diff --git a/tutorials/flux/flux-2-klein.mdx b/tutorials/flux/flux-2-klein.mdx index a6ebbc250..5e1f25059 100644 --- a/tutorials/flux/flux-2-klein.mdx +++ b/tutorials/flux/flux-2-klein.mdx @@ -21,14 +21,26 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a ## Flux.2 Klein 4B Workflows + + Run this workflow on Comfy Cloud + + Download the text-to-image workflow for Flux.2 Klein 4B. + + Run this workflow on Comfy Cloud + + Download the image editing workflow using the 4B base model. + + Run this workflow on Comfy Cloud + + Download the fast distilled 4B image editing workflow. @@ -67,14 +79,26 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a ## Flux.2 Klein 9B Workflows + + Run this workflow on Comfy Cloud + + Download the text-to-image workflow for Flux.2 Klein 9B. + + Run this workflow on Comfy Cloud + + Download the image editing workflow using the 9B base model. + + Run this workflow on Comfy Cloud + + Download the fast distilled 9B image editing workflow. diff --git a/tutorials/partner-nodes/bria/background-removal.mdx b/tutorials/partner-nodes/bria/background-removal.mdx index a86cd2c20..9ba6b7bc9 100644 --- a/tutorials/partner-nodes/bria/background-removal.mdx +++ b/tutorials/partner-nodes/bria/background-removal.mdx @@ -20,7 +20,7 @@ Remove the background from any image using Bria's AI service. The result is retu Download the workflow. - + Try it on Comfy Cloud. @@ -37,7 +37,7 @@ Replace a video's background with a solid color of your choice. Download the workflow. - + Try it on Comfy Cloud. @@ -50,7 +50,7 @@ Remove the background from a video and output it with transparency. Download the workflow. - + Try it on Comfy Cloud. @@ -63,7 +63,7 @@ Apply a professional chroma-key green or blue screen effect to your video, with Download the workflow. - + Try it on Comfy Cloud. @@ -76,7 +76,7 @@ Replace a video's background with a custom image or another video. Download the workflow. - + Try it on Comfy Cloud. diff --git a/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index bf351f2e6..617a6405b 100644 --- a/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -90,6 +90,10 @@ Two prebuilt Seedance 2.0 Real Human templates are available in ComfyUI. Both wi Use a verified portrait (plus optional additional reference images, videos, or audio) to drive a Seedance 2.0 video with a consistent real-person identity. + + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. + + Get the Seedance 2.0 Real Human Reference-to-Video workflow file. @@ -98,6 +102,10 @@ Use a verified portrait (plus optional additional reference images, videos, or a Provide a verified starting frame and ending frame to generate the video between them while preserving the real person's identity. + + Try the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow instantly on Comfy Cloud. + + Get the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow file. diff --git a/tutorials/partner-nodes/google/gemini.mdx b/tutorials/partner-nodes/google/gemini.mdx index f404b5245..093b5aae4 100644 --- a/tutorials/partner-nodes/google/gemini.mdx +++ b/tutorials/partner-nodes/google/gemini.mdx @@ -20,10 +20,13 @@ In this guide, we will walk you through completing the corresponding conversatio Please download the Json file below and drag it into ComfyUI to load the corresponding workflow. - + Download Json Format Workflow File + + Open in Comfy Cloud + ### 2. Complete the Workflow Execution Step by Step diff --git a/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 4c8332a52..06c45e021 100644 --- a/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -38,13 +38,14 @@ Generate cinematic video from a single image with strong aesthetic control and m > - - Get the HappyHorse 1.0 Image-to-Video workflow file. - - - + + Try the Image-to-Video workflow instantly on Comfy Cloud. + + Get the HappyHorse 1.0 Image-to-Video workflow file. + + ## HappyHorse 1.0 text-to-video @@ -56,34 +57,37 @@ Generate cinematic video from pure text prompts, with multi-shot sequencing and src="https://github.com/Comfy-Org/example_workflows/raw/refs/heads/main/api_nodes/happy_horse/1.0/t2v_1.mp4" > - - Get the HappyHorse 1.0 Text-to-Video workflow file. - - - + + Try the Text-to-Video workflow instantly on Comfy Cloud. + + Get the HappyHorse 1.0 Text-to-Video workflow file. + + ## HappyHorse 1.0 reference-to-video Use a reference subject to drive video generation, preserving identity across cinematic multi-shot sequences. - - Get the HappyHorse 1.0 Reference-to-Video workflow file. - - - + + Try the Reference-to-Video workflow instantly on Comfy Cloud. + + Get the HappyHorse 1.0 Reference-to-Video workflow file. + + ## HappyHorse 1.0 video edit Transform existing footage or replace/insert subjects while preserving motion and composition with V2V and SV2V editing workflows. - - Get the HappyHorse 1.0 Video Edit workflow file. - - - + + Try the Video Edit workflow instantly on Comfy Cloud. + + Get the HappyHorse 1.0 Video Edit workflow file. + + diff --git a/tutorials/partner-nodes/kling/kling-motion-control.mdx b/tutorials/partner-nodes/kling/kling-motion-control.mdx index ba213883a..c51dc8673 100644 --- a/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -40,9 +40,13 @@ The `character_orientation` parameter determines how the model interprets spatia ## Kling 2.6 Motion Control workflow - -

Download the workflow file in JSON format

- + + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. + + + + Download the workflow JSON file for local use. + ## Input requirements diff --git a/tutorials/partner-nodes/luma/luma-uni-1.mdx b/tutorials/partner-nodes/luma/luma-uni-1.mdx index 8b3ed92c5..573cb6aec 100644 --- a/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -55,23 +55,25 @@ When in doubt: ### Image Create workflow - + + Try the Image Create workflow instantly on Comfy Cloud. - - + Download JSON or search "Luma UNI-1 Image Create" in Template Library + ### Image Edit workflow - + + Try the Image Edit workflow instantly on Comfy Cloud. - - + Download JSON or search "Luma UNI-1 Image Edit" in Template Library + The workflow is simple: **prompt → evaluate → refine**. Leave the seed blank while exploring. Once you find something strong, lock the seed and iterate from there. diff --git a/tutorials/partner-nodes/meshy/meshy-6.mdx b/tutorials/partner-nodes/meshy/meshy-6.mdx index e5f96f2c0..6cf0bc2d9 100644 --- a/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -27,9 +27,9 @@ Meshy 6 is the latest generation of Meshy's 3D model generation technology, feat Generate 3D models directly from text descriptions using Meshy 6. - - Run the text-to-model workflow instantly on Comfy Cloud. - + + Run the text-to-model workflow instantly on Comfy Cloud. + Download the workflow JSON file for local use. @@ -39,9 +39,9 @@ Generate 3D models directly from text descriptions using Meshy 6. Convert 2D images into detailed 3D models with Meshy 6's image-to-3D capabilities. - - Run the image-to-model workflow instantly on Comfy Cloud. - + + Run the image-to-model workflow instantly on Comfy Cloud. + Download the workflow JSON file for local use. @@ -51,9 +51,9 @@ Convert 2D images into detailed 3D models with Meshy 6's image-to-3D capabilitie Generate 3D models from multiple view images for more accurate geometry and texture reconstruction. - - Run the multi-view workflow instantly on Comfy Cloud. - + + Run the multi-view workflow instantly on Comfy Cloud. + Download the workflow JSON file for local use. diff --git a/tutorials/partner-nodes/openai/chat.mdx b/tutorials/partner-nodes/openai/chat.mdx index e8e3a0eea..e96926d54 100644 --- a/tutorials/partner-nodes/openai/chat.mdx +++ b/tutorials/partner-nodes/openai/chat.mdx @@ -20,9 +20,14 @@ In this guide, we will walk you through completing the corresponding conversatio Please download the Json file below and drag it into ComfyUI to load the corresponding workflow. -| -| Download the JSON format workflow file. -| + + + Open in Comfy Cloud + + + Download the JSON format workflow file. + + ### 2. Complete the Workflow Execution Step by Step diff --git a/tutorials/partner-nodes/openai/gpt-image-2.mdx b/tutorials/partner-nodes/openai/gpt-image-2.mdx index c32bb1b14..a9fabc53e 100644 --- a/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -34,6 +34,7 @@ GPT-Image-2 is selected as a `model` option on the **OpenAI GPT Image 1.5** node Generate an image from a text prompt with GPT-Image-2's reasoning-driven composition. + Try the Text-to-Image workflow instantly on Comfy Cloud. @@ -41,6 +42,7 @@ Generate an image from a text prompt with GPT-Image-2's reasoning-driven composi Download the workflow JSON. + ![GPT-Image-2 Text-to-Image example](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_t2i_1.jpg) @@ -48,6 +50,7 @@ Generate an image from a text prompt with GPT-Image-2's reasoning-driven composi Edit an input image with high structural fidelity at up to 2K resolution. + Try the Image Edit workflow instantly on Comfy Cloud. @@ -55,6 +58,7 @@ Edit an input image with high structural fidelity at up to 2K resolution. Download the workflow JSON. + ![GPT-Image-2 Image-to-Image example](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_i2i_1.jpg) diff --git a/tutorials/partner-nodes/rodin/model-generation.mdx b/tutorials/partner-nodes/rodin/model-generation.mdx index 0863d781a..d19cea301 100644 --- a/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/tutorials/partner-nodes/rodin/model-generation.mdx @@ -24,9 +24,14 @@ Currently, ComfyUI's Partner nodes support the following Rodin model generation Download the file below and drag it into ComfyUI to load the corresponding workflow. - - Single-view Model Generation (Json Format) - + + + Single-view Model Generation (Json Format) + + + Open in Comfy Cloud + + Download the image below as input image @@ -54,9 +59,14 @@ You can modify the single-view workflow to a multi-view workflow, or directly do Download the file below and drag it into ComfyUI to load the corresponding workflow. - - Multi-view Model Generation (Json Format) - + + + Multi-view Model Generation (Json Format) + + + Open in Comfy Cloud + + Download the images below as input images diff --git a/tutorials/partner-nodes/tripo/model-generation.mdx b/tutorials/partner-nodes/tripo/model-generation.mdx index 3980a89e3..414136fd2 100644 --- a/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/tutorials/partner-nodes/tripo/model-generation.mdx @@ -26,9 +26,14 @@ Currently, ComfyUI's Partner nodes support the following Tripo model generation Download the file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- + + + Try the Text-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + ### 2. Complete the Workflow Execution Step by Step @@ -49,9 +54,14 @@ You can refer to the numbers in the image to complete the basic text-to-model wo Download the file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- + + + Try the Image-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + Download the image below as input image @@ -77,9 +87,14 @@ You can refer to the numbers in the image to complete the basic image-to-model w Download the file below and drag it into ComfyUI to load the corresponding workflow. - -

Download Json Format Workflow File

- + + + Try the Multiview-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + Download the images below as input images diff --git a/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/tutorials/partner-nodes/tripo/tripo-3-1.mdx index 6fcd1d761..dd87aab7c 100644 --- a/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -32,33 +32,36 @@ Compared to previous Tripo model versions, 3.1 provides the following improvemen ### Text-to-Model - + + Try the Text-to-Model workflow instantly on Comfy Cloud. - - + Download JSON or search "Tripo 3.1 Text-to-Model" in Template Library. + ### Image-to-Model - + + Try the Image-to-Model workflow instantly on Comfy Cloud. - - + Download JSON or search "Tripo 3.1 Image-to-Model" in Template Library. + ### Multiview-to-Model - + + Try the Multiview-to-Model workflow instantly on Comfy Cloud. - - + Download JSON or search "Tripo 3.1 Multiview-to-Model" in Template Library. + ### Version Comparison diff --git a/tutorials/utility/moge.mdx b/tutorials/utility/moge.mdx index 20e7fccfa..73d3b6598 100644 --- a/tutorials/utility/moge.mdx +++ b/tutorials/utility/moge.mdx @@ -61,8 +61,12 @@ ComfyUI/ MoGe also estimates the camera's field of view (FOV) from the image, which can be optionally overridden with a ground-truth value for even more accurate results. - Download JSON or search "MoGe Depth Estimation" in Template Library - + Download JSON or search "MoGe Depth Estimation" in Template Library + + + + Open in Comfy Cloud + Get the example input image for this workflow @@ -87,8 +91,12 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b **What it does:** Converts a single perspective photo into a textured GLB mesh with normal and depth previews. MoGe estimates point maps, depth, and normals from the visible scene, then converts them to a mesh. This is **monocular geometry estimation** — occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. - Download JSON or search "3D MoGe Perspective to Mesh" in Template Library - + Download JSON or search "3D MoGe Perspective to Mesh" in Template Library + + + + Open in Comfy Cloud + Get the example input image for this workflow @@ -110,8 +118,12 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b **What it does:** Converts an equirectangular (360°) panorama into a textured GLB mesh. The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, runs monocular geometry estimation on each view independently, then merges them into a single mesh. Each segment is still single-view estimation, so the result is a rough scene reconstruction — useful for getting a spatial overview of a 360° scene, but occluded areas and geometry behind surfaces will be missing or fragmented. - Download JSON or search "3D MoGe Panorama to Mesh" in Template Library - + Download JSON or search "3D MoGe Panorama to Mesh" in Template Library + + + + Open in Comfy Cloud + Get the example input image for this workflow From 7c444bff2bcb1854d9526ec5b7f55ef444ac66fa Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 21:59:17 +0800 Subject: [PATCH 05/19] fix: add missing Cloud cards to remaining pages --- tutorials/3d/triposplat.mdx | 7 ++++++- tutorials/image/pixeldit/pixeldit.mdx | 4 ++++ tutorials/image/z-image/z-image-turbo.mdx | 4 ++++ 3 files changed, 14 insertions(+), 1 deletion(-) diff --git a/tutorials/3d/triposplat.mdx b/tutorials/3d/triposplat.mdx index b6a9c2d73..5af78a21a 100644 --- a/tutorials/3d/triposplat.mdx +++ b/tutorials/3d/triposplat.mdx @@ -14,9 +14,14 @@ Unlike traditional 3D reconstruction methods that require multiple views or gene - + + + Run this workflow instantly on Comfy Cloud + + Download JSON or search "TripoSplat" in Template Library + ## How it works diff --git a/tutorials/image/pixeldit/pixeldit.mdx b/tutorials/image/pixeldit/pixeldit.mdx index 5f547000c..04913bbf6 100644 --- a/tutorials/image/pixeldit/pixeldit.mdx +++ b/tutorials/image/pixeldit/pixeldit.mdx @@ -27,6 +27,10 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' Download JSON or search "PixelDiT" in Template Library + + Open in cloud + + The workflow consists of three main nodes: diff --git a/tutorials/image/z-image/z-image-turbo.mdx b/tutorials/image/z-image/z-image-turbo.mdx index 1d9f1aea0..264770ce4 100644 --- a/tutorials/image/z-image/z-image-turbo.mdx +++ b/tutorials/image/z-image/z-image-turbo.mdx @@ -72,6 +72,10 @@ This workflow uses the Z-Image-Turbo Fun Union ControlNet model to generate imag Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file. + + Run this workflow directly on ComfyUI Cloud. + + ### Additional model for ControlNet From 51cb3b7ff0bdc8569a6949bf46e6e5c6987041ec Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 22:19:10 +0800 Subject: [PATCH 06/19] fix: add previews and input cards to partner-nodes pages --- tutorials/partner-nodes/anthropic/claude.mdx | 9 +- .../partner-nodes/beeble/beeble-switchx.mdx | 14 +-- .../bytedance/seedance-2-0-real-human.mdx | 45 ++++++++-- .../partner-nodes/bytedance/seedance-2-0.mdx | 67 ++++++++++++-- .../bytedance/seedream-5-lite.mdx | 33 +++++-- .../bytedance/seedream-5-pro.mdx | 38 +++++--- .../google/gemini-omni-flash.mdx | 6 ++ .../partner-nodes/google/nano-banana-2.mdx | 6 ++ .../happyhorse/happyhorse1-0.mdx | 28 ++++++ .../happyhorse/happyhorse1-1.mdx | 4 + .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 45 ++++++++++ .../hunyuan3d/model-generation.mdx | 39 ++++++++ tutorials/partner-nodes/meshy/meshy-6.mdx | 22 +++++ .../partner-nodes/openai/gpt-image-2.mdx | 8 ++ .../partner-nodes/recraft/recraft-v4.mdx | 4 + .../partner-nodes/rodin/model-generation.mdx | 36 ++++++-- .../partner-nodes/tripo/model-generation.mdx | 37 ++++++-- tutorials/partner-nodes/tripo/tripo-3-1.mdx | 12 +++ tutorials/video/bytedance/bernini-r.mdx | 22 +++++ tutorials/video/ltx/ltx-2.mdx | 90 +++++++++++++++++-- tutorials/video/ltxv.mdx | 11 ++- tutorials/video/wan/wan-alpha.mdx | 2 + tutorials/video/wan/wan-causal-forcing.mdx | 8 ++ tutorials/video/wan/wan-dancer.mdx | 2 + tutorials/video/wan/wan2-2-fun-camera.mdx | 8 +- tutorials/video/wan/wan2-2-fun-control.mdx | 8 +- tutorials/video/wan/wan2-2-fun-inp.mdx | 14 ++- tutorials/video/zai/scail2.mdx | 13 +++ 28 files changed, 563 insertions(+), 68 deletions(-) diff --git a/tutorials/partner-nodes/anthropic/claude.mdx b/tutorials/partner-nodes/anthropic/claude.mdx index 49b511de6..142b74e64 100644 --- a/tutorials/partner-nodes/anthropic/claude.mdx +++ b/tutorials/partner-nodes/anthropic/claude.mdx @@ -28,13 +28,16 @@ All models support both text-only and multimodal (text + image) inputs with up t ## Anthropic Claude Chat Workflow - +Anthropic Claude Chat workflow preview + + + Open in Comfy Cloud - - + Download JSON or search "Anthropic Claude" in Template Library + In the corresponding template, we have built a prompt for analyzing and generating role prompts, used to interpret your images into corresponding drawing prompts. diff --git a/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 67f3a92fa..8a1bc5ea5 100644 --- a/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -20,13 +20,14 @@ Replace the background environment and relight your subject with a reference ima Beeble SwitchX Image Edit Workflow - + + Open in Comfy Cloud - - + Download JSON or search "Beeble SwitchX: Image Edit" in Template Library + ### How It Works @@ -60,13 +61,14 @@ Apply environment relighting to a video while preserving the original motion and Beeble SwitchX Video Edit Workflow - + + Open in Comfy Cloud - - + Download JSON or search "Beeble SwitchX: Video Edit" in Template Library + ### How It Works diff --git a/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 617a6405b..4920828af 100644 --- a/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -90,23 +90,56 @@ Two prebuilt Seedance 2.0 Real Human templates are available in ComfyUI. Both wi Use a verified portrait (plus optional additional reference images, videos, or audio) to drive a Seedance 2.0 video with a consistent real-person identity. - +Seedance 2.0 Real Human R2V workflow preview + + + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. - - + Get the Seedance 2.0 Real Human Reference-to-Video workflow file. + + +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample portrait for verification + + +
### Seedance 2.0 Real Human first-last-frame to video Provide a verified starting frame and ending frame to generate the video between them while preserving the real person's identity. - +Seedance 2.0 Real Human FLF2V workflow preview + + + Try the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow instantly on Comfy Cloud. - - + Get the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow file. + + +
+Input materials + +Download these sample input images to try the workflow: + + + + Download sample first frame + + + Download sample last frame + + +
diff --git a/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index d573b4bd1..c785caaa9 100644 --- a/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -28,37 +28,82 @@ With these capabilities, Seedance 2.0 fits a wide range of use cases including a Generate a video from a text prompt, with Seedance 2.0 handling scene, motion, and pacing. - +Seedance 2.0 Text-to-Video workflow preview + + + Try the Text-to-Video workflow instantly on Comfy Cloud. - - + Download the workflow JSON. + ### Reference to video (R2V) Use reference images, video, or audio to guide look, motion, and rhythm while keeping results coherent. - +Seedance 2.0 Reference-to-Video workflow preview + + + Try the Reference-to-Video workflow instantly on Comfy Cloud. - - + Download the workflow JSON. + + +
+Input materials + +Download these sample input images to try the workflow: + + + + Download sample outfit image 1 + + + Download sample outfit image 2 + + + Download sample outfit image 3 + + + Download sample subject reference + + +
### First-last-frame to video (FLF2V) Provide a starting frame and ending frame, and Seedance 2.0 generates the motion and transitions between them. - +Seedance 2.0 FLF2V workflow preview + + + Try the First-Last-Frame-to-Video workflow instantly on Comfy Cloud. - - + Download the workflow JSON. + + +
+Input materials + +Download these sample input images to try the workflow: + + + + Download sample first frame + + + Download sample last frame + + +
## Seedance 2.0 Mini @@ -74,6 +119,8 @@ Seedance 2.0 Mini is on the same `Seedance 2.0` pricing line as the full model. Generate a cinematic video from a text prompt with AI camera controls and consistent character generation across scenes. +Seedance 2.0 Mini Text-to-Video workflow preview + Try the Mini Text-to-Video workflow instantly on Comfy Cloud. @@ -87,6 +134,8 @@ Generate a cinematic video from a text prompt with AI camera controls and consis Generate a cinematic video from two reference images (subject + scene) with AI-driven camera motion and consistent character generation. +Seedance 2.0 Mini Reference-to-Video workflow preview + Try the Mini Reference-to-Video workflow instantly on Comfy Cloud. diff --git a/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index 961cb7e14..e660c7c28 100644 --- a/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -21,13 +21,31 @@ Seedream 5.0 lite is the latest image generation model from BytePlus. It is the ## Seedream 5.0 lite image edit workflow - +Seedream 5.0 Lite Image Edit workflow preview + + + Open in Comfy Cloud - - + Download JSON or search "Seedream 5.0 lite" in Template Library + + +
+Input materials + +Download these sample input images to try the workflow: + + + + Download sample input image + + + Download sample style reference + + +
### Image edit showcase @@ -56,13 +74,16 @@ Seedream 5.0 lite is the latest image generation model from BytePlus. It is the ## Seedream 5.0 lite text to image workflow - +Seedream 5.0 Lite Text-to-Image workflow preview + + + Open in Comfy Cloud - - + Download JSON or search "Seedream 5.0 lite" in Template Library + ### World knowledge showcase diff --git a/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index d51805f39..114b12811 100644 --- a/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -24,13 +24,16 @@ Seedream 5.0 Pro is ByteDance's professional-grade image generation model, built Generate a high-quality image from a text prompt, with Seedream 5.0 Pro handling composition, lighting, and detail. - - Open in Comfy Cloud - +Seedream 5.0 Pro Text-to-Image workflow preview - + + + Open in Comfy Cloud + + Download JSON or search "Seedream 5.0 Pro" in Template Library - + + ![Seedream 5.0 Pro text to image preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/Seedream5.0_Pro_T2I.png) @@ -38,13 +41,28 @@ Generate a high-quality image from a text prompt, with Seedream 5.0 Pro handling Edit existing images with text instructions. Change objects, swap styles, adjust colors, or recompose scenes while keeping the original structure intact. - - Open in Comfy Cloud - +Seedream 5.0 Pro Image Edit workflow preview - + + + Open in Comfy Cloud + + Download JSON or search "Seedream 5.0 Pro" in Template Library - + + + +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample input image + + +
![Seedream 5.0 Pro image edit input](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/input/interior_college.png) diff --git a/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/tutorials/partner-nodes/google/gemini-omni-flash.mdx index c07c9cebb..a63d908e7 100644 --- a/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -33,6 +33,8 @@ Gemini Omni Flash is Google DeepMind's high-quality, cost-efficient video genera +![Gemini Omni Flash Text to Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_t2v-1.webp) + Generate cinematic video from natural language prompts. Transform text descriptions into video output with world-aware motion, lighting, and sound. Ideal for social media content creation, rapid video prototyping, and iterative visual storytelling. ### Image to Video @@ -52,6 +54,8 @@ Generate cinematic video from natural language prompts. Transform text descripti +![Gemini Omni Flash Image to Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_i2v-1.webp) + Generate a video from two images using Gemini Omni Flash. Interpret natural language prompts to control duration and aspect ratio. Perfect for creating short brand clips, dynamic social media content, and iterative video edits through conversational prompting. ### Video Edit @@ -68,6 +72,8 @@ Generate a video from two images using Gemini Omni Flash. Interpret natural lang +![Gemini Omni Flash Video Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_video_edit-1.webp) + Edit videos with natural language using Gemini Omni Flash. Transform a single input video into one edited output based on your descriptive instructions. Specify the duration and aspect ratio in your prompt. Ideal for quick social media remixes, cinematic scene adjustments, and iterative video refinements. ## Get started diff --git a/tutorials/partner-nodes/google/nano-banana-2.mdx b/tutorials/partner-nodes/google/nano-banana-2.mdx index b524ac55c..2ded81bda 100644 --- a/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -30,6 +30,12 @@ Nano Banana 2 is now available in ComfyUI through Partner Nodes. This release fu Download JSON or search "Nano Banana 2" in Template Library +![Nano Banana 2 workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_nano_banana2_image_edit-1.webp) + + + Get the example input image for this workflow + + ### Pro-level quality showcase ![Quality comparison](https://substack-post-media.s3.amazonaws.com/public/images/1f92ae2e-14d8-4a3b-9ed7-e57dacd2584f_1825x1696.png) diff --git a/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 06c45e021..0a2649889 100644 --- a/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -47,6 +47,12 @@ Generate cinematic video from a single image with strong aesthetic control and m +![HappyHorse 1.0 Image-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_i2v-1.webp) + + + Get the example input image for this workflow. + + ## HappyHorse 1.0 text-to-video Generate cinematic video from pure text prompts, with multi-shot sequencing and refined visual atmosphere. @@ -66,6 +72,8 @@ Generate cinematic video from pure text prompts, with multi-shot sequencing and +![HappyHorse 1.0 Text-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_t2v-1.webp) + ## HappyHorse 1.0 reference-to-video Use a reference subject to drive video generation, preserving identity across cinematic multi-shot sequences. @@ -79,6 +87,16 @@ Use a reference subject to drive video generation, preserving identity across ci +![HappyHorse 1.0 Reference-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_r2v-1.webp) + + + Get the example car reference image. + + + + Get the example person reference image. + + ## HappyHorse 1.0 video edit Transform existing footage or replace/insert subjects while preserving motion and composition with V2V and SV2V editing workflows. @@ -91,3 +109,13 @@ Transform existing footage or replace/insert subjects while preserving motion an Get the HappyHorse 1.0 Video Edit workflow file. + +![HappyHorse 1.0 Video Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_video_edit-1.webp) + + + Get the example input image for the video edit workflow. + + + + Get the example input video for the video edit workflow. + diff --git a/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index 6b7acfe55..ea570898b 100644 --- a/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -41,6 +41,8 @@ Build a complete scene from scratch. You control style, shot size, lighting, act +![HappyHorse 1.1 Text-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_1_t2v-1.webp) + ## HappyHorse 1.1 image-to-video Animate a static first frame. The image already carries the look, so you describe the motion and the camera move. HappyHorse 1.1 returns a video with audio baked in. @@ -57,6 +59,8 @@ Animate a static first frame. The image already carries the look, so you describ +![HappyHorse 1.1 Image-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_1_i2v-1.webp) + ## HappyHorse 1.1 reference-to-video Orchestrate a multi-character stage play. Map characters and scenes to reference images, then direct them through a timestamped storyboard with per-character dialogue. Up to 9 reference images per generation, with character and scene references separated so characters stay consistent across background changes. diff --git a/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index b9ed778f1..5c95f47d1 100644 --- a/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -46,6 +46,8 @@ Enter a text description to generate a 3D model in one click. The model accurate **Example prompt**: "Mechanical dolphin with gears, steampunk" +Hunyuan 3D Text-to-3D workflow preview + Run on Comfy Cloud @@ -54,18 +56,55 @@ Enter a text description to generate a 3D model in one click. The model accurate Upload one or more images to generate high-quality 3D models. Support for 2–4 multi-view images improves geometry and material fidelity. +Hunyuan 3D Image-to-3D workflow preview + Run on Comfy Cloud +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample input image + + +
+ ## Multi-view-to-3D workflow Provide multiple view images (front, back, side) to generate 3D models with improved accuracy and detail. This uses the same workflow as Image-to-3D—simply upload 2–4 images from different angles. +Hunyuan 3D Multi-view-to-3D workflow preview + Run on Comfy Cloud +
+Input materials + +Download these sample multi-view input images to try the workflow: + + + + Download front view + + + Download back view + + + Download left view + + + Download right view + + +
+ ## Advanced features Following the [initial HY 3D 3.0 integration](https://blog.comfy.org/p/hunyuan-3d-30-in-comfyui-state-of), Hunyuan 3D's advanced processing features are now available via Partner Nodes. These workflows help close the gap between generation and production by bringing key post-processing steps into ComfyUI. @@ -74,6 +113,8 @@ Following the [initial HY 3D 3.0 integration](https://blog.comfy.org/p/hunyuan-3 Split a complete 3D model into meaningful structural parts, such as armor pieces, accessories, wheels, or other distinct components. This makes it easier to edit specific regions of an asset, swap parts for variations, and prepare models for modular workflows, animation, or downstream assembly. +Hunyuan 3D parts decomposition workflow preview + Run the workflow @@ -86,6 +127,8 @@ Split a complete 3D model into meaningful structural parts, such as armor pieces Automatically generate UV layouts for supported 3D models, turning raw geometry into assets that are much easier to texture. Instead of manually cutting seams and organizing UV islands, creators can move more quickly into painting, baking, and material work with a cleaner starting point. +Hunyuan 3D UV unwrapping workflow preview + Run the workflow @@ -98,6 +141,8 @@ Automatically generate UV layouts for supported 3D models, turning raw geometry Convert dense geometry into cleaner meshes with more organized edge flow, helping generated models become easier to optimize and reuse in real production pipelines. This is especially useful when preparing assets for game engines, real-time rendering, or any workflow that benefits from lower-density, better-structured geometry. +Hunyuan 3D smart topology workflow preview + Run the workflow diff --git a/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index 070f38b48..d3ff29a5b 100644 --- a/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -46,6 +46,8 @@ Enter a text description to generate a 3D model in one click. The model accurate **Example prompt**: "Mechanical dolphin with gears, steampunk" +Hunyuan 3D Text-to-3D workflow preview + Run on Comfy Cloud @@ -54,14 +56,51 @@ Enter a text description to generate a 3D model in one click. The model accurate Upload one or more images to generate high-quality 3D models. Support for 2–4 multi-view images improves geometry and material fidelity. +Hunyuan 3D Image-to-3D workflow preview + Run on Comfy Cloud +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample input image + + +
+ ## Multi-view-to-3D workflow Provide multiple view images (front, back, side) to generate 3D models with improved accuracy and detail. This uses the same workflow as Image-to-3D—simply upload 2–4 images from different angles. +Hunyuan 3D Multi-view-to-3D workflow preview + Run on Comfy Cloud + +
+Input materials + +Download these sample multi-view input images to try the workflow: + + + + Download front view + + + Download back view + + + Download left view + + + Download right view + + +
diff --git a/tutorials/partner-nodes/meshy/meshy-6.mdx b/tutorials/partner-nodes/meshy/meshy-6.mdx index 6cf0bc2d9..36767485d 100644 --- a/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -35,6 +35,8 @@ Generate 3D models directly from text descriptions using Meshy 6. Download the workflow JSON file for local use. +![Meshy 6 Text-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_text_to_model-1.webp) + ## Image-to-Model Workflow Convert 2D images into detailed 3D models with Meshy 6's image-to-3D capabilities. @@ -47,6 +49,12 @@ Convert 2D images into detailed 3D models with Meshy 6's image-to-3D capabilitie Download the workflow JSON file for local use. +![Meshy 6 Image-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_image_to_model-1.webp) + + + Get the example input image for this workflow. + + ## Multi-view to Model Workflow Generate 3D models from multiple view images for more accurate geometry and texture reconstruction. @@ -58,3 +66,17 @@ Generate 3D models from multiple view images for more accurate geometry and text Download the workflow JSON file for local use. + +![Meshy 6 Multi-view to Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_multi_image_to_model-1.webp) + + + + Get the example front view image. + + + Get the example back view image. + + + Get the example side view image. + + diff --git a/tutorials/partner-nodes/openai/gpt-image-2.mdx b/tutorials/partner-nodes/openai/gpt-image-2.mdx index a9fabc53e..dad603bb0 100644 --- a/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -44,6 +44,8 @@ Generate an image from a text prompt with GPT-Image-2's reasoning-driven composi +![GPT-Image-2 Text-to-Image workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_openai_gpt_image_2_t2i-1.webp) + ![GPT-Image-2 Text-to-Image example](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_t2i_1.jpg) ### Image edit @@ -60,6 +62,12 @@ Edit an input image with high structural fidelity at up to 2K resolution. +![GPT-Image-2 Image Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_openai_gpt_image_2_image_edit-1.webp) + + + Get the example input image for this workflow + + ![GPT-Image-2 Image-to-Image example](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_i2i_1.jpg) ![GPT-Image-2 Image Edit example 1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_image_edit_1.jpg) diff --git a/tutorials/partner-nodes/recraft/recraft-v4.mdx b/tutorials/partner-nodes/recraft/recraft-v4.mdx index 726f872be..f4d3802b8 100644 --- a/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -32,6 +32,8 @@ Recraft V4 is a new image generation model built for professional design work. I Download JSON or search "Recraft V4 Text to Image" in Template Library +![Recraft V4 Text to Image workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_t2i-1.webp) + ### Steps to run the workflow 1. (Optional) Modify the `Recraft Style` node to control the visual style @@ -63,6 +65,8 @@ Recraft V4 can generate production-ready SVG vector outputs directly. This is us Download JSON or search "Recraft V4 Text to Vector" in Template Library +![Recraft V4 Text to Vector workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_text_to_vector-1.webp) + ### Vector showcase ![Rooftop illustration](https://substack-post-media.s3.amazonaws.com/public/images/571c5214-5c7f-43f9-81c4-80cad1e5578b_1792x1024.png) diff --git a/tutorials/partner-nodes/rodin/model-generation.mdx b/tutorials/partner-nodes/rodin/model-generation.mdx index d19cea301..5a5121655 100644 --- a/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/tutorials/partner-nodes/rodin/model-generation.mdx @@ -20,9 +20,9 @@ Currently, ComfyUI's Partner nodes support the following Rodin model generation ## Single-view Model Generation Workflow -### 1. Workflow File Download +Generate a 3D model from a single image input with Rodin. -Download the file below and drag it into ComfyUI to load the corresponding workflow. +Rodin Single-view Model Generation workflow preview @@ -33,7 +33,17 @@ Download the file below and drag it into ComfyUI to load the corresponding workf -Download the image below as input image +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample input image + + +
![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/doll.jpg) @@ -57,7 +67,7 @@ The corresponding `Rodin 3D Generate - Regular Generate` allows up to 5 image in You can modify the single-view workflow to a multi-view workflow, or directly download the workflow file below -Download the file below and drag it into ComfyUI to load the corresponding workflow. +Rodin Multi-view Model Generation workflow preview @@ -68,7 +78,23 @@ Download the file below and drag it into ComfyUI to load the corresponding workf -Download the images below as input images +
+Input materials + +Download these sample input images to try the workflow: + + + + Download front view + + + Download back view + + + Download left view + + +
![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/front.jpg) ![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/back.jpg) diff --git a/tutorials/partner-nodes/tripo/model-generation.mdx b/tutorials/partner-nodes/tripo/model-generation.mdx index 414136fd2..c802f0849 100644 --- a/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/tutorials/partner-nodes/tripo/model-generation.mdx @@ -22,9 +22,9 @@ Currently, ComfyUI's Partner nodes support the following Tripo model generation ## Text-to-Model Workflow -### 1. Workflow File Download +Generate a 3D model from a text prompt with Tripo's text-to-model capabilities. -Download the file below and drag it into ComfyUI to load the corresponding workflow. +Tripo Text-to-Model workflow preview @@ -50,9 +50,9 @@ You can refer to the numbers in the image to complete the basic text-to-model wo ## Image-to-Model Workflow -### 1. Workflow File Download +Generate a 3D model from a single image input. -Download the file below and drag it into ComfyUI to load the corresponding workflow. +Tripo Image-to-Model workflow preview @@ -63,8 +63,17 @@ Download the file below and drag it into ComfyUI to load the corresponding workf +
+Input materials -Download the image below as input image +Download this sample input image to try the workflow: + + + + Download sample input image + + +
![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/image_to_model/panda.jpg) ### 2. Complete the Workflow Execution Step by Step @@ -83,9 +92,9 @@ You can refer to the numbers in the image to complete the basic image-to-model w ## Multi-view Model Generation Workflow -### 1. Workflow File Download +Generate a 3D model from multiple view images for enhanced accuracy. -Download the file below and drag it into ComfyUI to load the corresponding workflow. +Tripo Multi-view Model Generation workflow preview @@ -96,8 +105,20 @@ Download the file below and drag it into ComfyUI to load the corresponding workf +
+Input materials + +Download these sample input images to try the workflow: -Download the images below as input images + + + Download front view input + + + Download back view input + + +
![Front View](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/multiview_to_image/front.jpg) ![Back View](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/multiview_to_image/back.jpg) diff --git a/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/tutorials/partner-nodes/tripo/tripo-3-1.mdx index dd87aab7c..9dd2c73ea 100644 --- a/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -32,6 +32,10 @@ Compared to previous Tripo model versions, 3.1 provides the following improvemen ### Text-to-Model +Generate a high-detail 3D model from a text prompt using Tripo 3.1. + +Tripo 3.1 Text-to-Model workflow preview + Try the Text-to-Model workflow instantly on Comfy Cloud. @@ -43,6 +47,10 @@ Compared to previous Tripo model versions, 3.1 provides the following improvemen ### Image-to-Model +Generate a high-detail 3D model from an image input using Tripo 3.1. + +Tripo 3.1 Image-to-Model workflow preview + Try the Image-to-Model workflow instantly on Comfy Cloud. @@ -54,6 +62,10 @@ Compared to previous Tripo model versions, 3.1 provides the following improvemen ### Multiview-to-Model +Generate a high-detail 3D model from multiple view images using Tripo 3.1. + +Tripo 3.1 Multiview-to-Model workflow preview + Try the Multiview-to-Model workflow instantly on Comfy Cloud. diff --git a/tutorials/video/bytedance/bernini-r.mdx b/tutorials/video/bytedance/bernini-r.mdx index 66a3c5e5b..0f2a388cd 100644 --- a/tutorials/video/bytedance/bernini-r.mdx +++ b/tutorials/video/bytedance/bernini-r.mdx @@ -80,6 +80,17 @@ ComfyUI/ +#### Input materials + + + + Download the default input image, or use your own. + + + Download the default reference image, or use your own. + + +
Bernini-R Image Editing output Bernini-R Image Editing comparison @@ -115,6 +126,17 @@ ComfyUI/ +#### Input materials + + + + Download the default input video, or use your own. + + + Download the default reference image, or use your own. + + + ![Bernini-R Video Editing preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_bernini_r_video_editing-1.webp) ### Steps to Run diff --git a/tutorials/video/ltx/ltx-2.mdx b/tutorials/video/ltx/ltx-2.mdx index 05778e3c8..66a900902 100644 --- a/tutorials/video/ltx/ltx-2.mdx +++ b/tutorials/video/ltx/ltx-2.mdx @@ -45,6 +45,8 @@ LTX-2 is natively supported in ComfyUI. To get started: Generate videos from text prompts. +![ComfyUI Workflow - LTX-2 T2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_t2v-1.webp) + Download workflow @@ -55,14 +57,24 @@ Generate videos from text prompts. **Distilled version** (faster, 8 steps): - - Download workflow - + +![ComfyUI Workflow - LTX-2 T2V Distilled](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_t2v_distilled-1.webp) + + + + Download workflow + + + Open in cloud + + ### Image-to-video Generate videos from an input image. +![ComfyUI Workflow - LTX-2 I2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_i2v-1.webp) + Download workflow @@ -72,16 +84,43 @@ Generate videos from an input image. +#### Input materials + + + + Download the default input image, or use your own image. + + + **Distilled version** (faster, 8 steps): - - Download workflow - + +![ComfyUI Workflow - LTX-2 I2V Distilled](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_i2v_distilled-1.webp) + + + + Download workflow + + + Open in cloud + + + +#### Input materials + + + + Download the default input image, or use your own image. + + ### Control-to-video Generate videos with structural control using IC-LoRAs. **Depth control:** + +![ComfyUI Workflow - LTX-2 Depth to Video](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_depth_to_video-1.webp) + Download workflow @@ -91,7 +130,21 @@ Generate videos with structural control using IC-LoRAs. +#### Input materials + + + + Download the default depth input image, or use your own. + + + Download the default input video, or use your own. + + + **Canny control:** + +![ComfyUI Workflow - LTX-2 Canny to Video](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_canny_to_video-1.webp) + Download workflow @@ -101,7 +154,21 @@ Generate videos with structural control using IC-LoRAs. +#### Input materials + + + + Download the default canny input image, or use your own. + + + Download the default input video, or use your own. + + + **Pose control:** + +![ComfyUI Workflow - LTX-2 Pose to Video](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_pose_to_video-1.webp) + Download workflow @@ -111,6 +178,17 @@ Generate videos with structural control using IC-LoRAs. +#### Input materials + + + + Download the default pose input image, or use your own. + + + Download the default input video, or use your own. + + + ## Prompting tips When writing prompts for LTX-2, focus on detailed, chronological descriptions of actions and scenes. Include specific movements, appearances, camera angles, and environmental details in a single flowing paragraph. Start directly with the action and keep descriptions literal and precise. diff --git a/tutorials/video/ltxv.mdx b/tutorials/video/ltxv.mdx index 7d8f75de8..7ead602e7 100644 --- a/tutorials/video/ltxv.mdx +++ b/tutorials/video/ltxv.mdx @@ -22,7 +22,7 @@ Drag the video directly into ComfyUI to run the workflow. ## Image to Video -Allows you to control the video with a first [frame image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png). +Allows you to control the video with a first frame image. @@ -31,8 +31,13 @@ Allows you to control the video with a first [frame image](https://raw.githubuse Download JSON or search "LTX-Video" in Template Library - - Get the example input image for this workflow + + +#### Input materials + + + + Download the default input image, or use your own image. diff --git a/tutorials/video/wan/wan-alpha.mdx b/tutorials/video/wan/wan-alpha.mdx index 0a2d6c9b7..b10b24783 100644 --- a/tutorials/video/wan/wan-alpha.mdx +++ b/tutorials/video/wan/wan-alpha.mdx @@ -37,6 +37,8 @@ The model excels at generating transparent backgrounds, semi-transparent objects Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan 2.1 Alpha T2V 14B" in the Template Library under `Workflow` → `Browse Templates` → `Video`. +Wan-Alpha T2V workflow + Open in Comfy Cloud diff --git a/tutorials/video/wan/wan-causal-forcing.mdx b/tutorials/video/wan/wan-causal-forcing.mdx index cf3d71ce7..94b9dac4d 100644 --- a/tutorials/video/wan/wan-causal-forcing.mdx +++ b/tutorials/video/wan/wan-causal-forcing.mdx @@ -23,6 +23,14 @@ This workflow uses **Wan2.1** and supports both **Causal Forcing** (standard) an +### Input materials + + + + Input image for the first frame. Download and use this image, or replace with your own. + + + ## How it works Unlike standard video generation which processes all frames in parallel, Causal Forcing treats video generation as a **sequential process**: diff --git a/tutorials/video/wan/wan-dancer.mdx b/tutorials/video/wan/wan-dancer.mdx index 6d7a9592b..cc40e2d11 100644 --- a/tutorials/video/wan/wan-dancer.mdx +++ b/tutorials/video/wan/wan-dancer.mdx @@ -30,6 +30,8 @@ The Wan Dancer workflow takes two inputs: a reference image of the character and Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan Dancer" in the Template Library under `Workflow` → `Browse Templates` → `Video`. +Wan Dancer workflow + Open in Comfy Cloud diff --git a/tutorials/video/wan/wan2-2-fun-camera.mdx b/tutorials/video/wan/wan2-2-fun-camera.mdx index 69163f42a..4204c50b1 100644 --- a/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -44,6 +44,8 @@ Download the video or JSON file below and drag it into ComfyUI to load the corre src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > +Wan2.2 Fun Camera Control workflow + Open in Comfy Cloud @@ -55,8 +57,10 @@ Download the video or JSON file below and drag it into ComfyUI to load the corre Please download the image below, which we will use as input. - - +### Input materials + + + Starting frame for the video generation. Download and use this image, or replace with your own. diff --git a/tutorials/video/wan/wan2-2-fun-control.mdx b/tutorials/video/wan/wan2-2-fun-control.mdx index 024ebec5b..4d5d31b04 100644 --- a/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/tutorials/video/wan/wan2-2-fun-control.mdx @@ -49,6 +49,8 @@ Since using the 4-step LoRA provides a better experience for first-time workflow Update your ComfyUI to the latest version, then download and drag the workflow file into ComfyUI, or find "Wan2.2 Fun Control" in the Template Library under `Workflow` → `Browse Templates` → `Video`. +Wan2.2 Fun Control workflow + Open in Comfy Cloud @@ -60,11 +62,13 @@ Update your ComfyUI to the latest version, then download and drag the workflow f Please download the following images and videos as input materials. +### Input materials + - + Start frame for video generation. Download and use this image, or replace with your own. - + Preprocessed pose control video. Download and use this video, or replace with your own. diff --git a/tutorials/video/wan/wan2-2-fun-inp.mdx b/tutorials/video/wan/wan2-2-fun-inp.mdx index a744c578a..2c6fb2aa7 100644 --- a/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -54,6 +54,8 @@ Please update your ComfyUI to the latest version, and find "**Wan2.2 Fun Inp**" Or, after updating ComfyUI to the latest version, download the workflow below and drag it into ComfyUI to load. +Wan2.2 Fun Inp workflow + Download JSON or search "Wan2.2 Fun Inp" in Template Library @@ -65,8 +67,16 @@ Or, after updating ComfyUI to the latest version, download the workflow below an Use the following materials as the start and end frames -![Wan2.2 Fun Control ComfyUI Workflow Start Frame Material](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_inp/start_image.png) -![Wan2.2 Fun Control ComfyUI Workflow End Frame Material](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_inp/end_image.png) +### Input materials + + + + Start frame for video generation. Download and use this image, or replace with your own. + + + End frame for video generation. Download and use this image, or replace with your own. + + ### 2. Models diff --git a/tutorials/video/zai/scail2.mdx b/tutorials/video/zai/scail2.mdx index e969fef1c..a56b86d06 100644 --- a/tutorials/video/zai/scail2.mdx +++ b/tutorials/video/zai/scail2.mdx @@ -29,6 +29,19 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' +![ComfyUI Workflow - SCAIL-2 Character Replacement](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_wan21_scail2_character_replacement-1.webp) + +#### Input materials + + + + Download the default reference character image, or use your own. + + + Download the default driving video, or use your own. + + + ## How the Workflow Works From 8d998af952e8871fba25b65e3264d505a5edf90f Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 22:19:26 +0800 Subject: [PATCH 07/19] fix: add previews and input cards to video + partner-nodes pages --- .../grok/grok-imagine-video-1-5.mdx | 4 +++ .../happyhorse/happyhorse1-1.mdx | 2 ++ tutorials/partner-nodes/wan/wan2-7.mdx | 36 ++++++++++++------- 3 files changed, 30 insertions(+), 12 deletions(-) diff --git a/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index 757048c33..98ac05114 100644 --- a/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -32,6 +32,10 @@ Both variants generate **native audio** — sound effects, ambience, and dialogu + + Get the example input image for this workflow. + + ### Workflow Overview This workflow uses three nodes: diff --git a/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index ea570898b..052701160 100644 --- a/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -80,6 +80,8 @@ Orchestrate a multi-character stage play. Map characters and scenes to reference +![HappyHorse 1.1 Reference-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_1_r2v-1.webp) + ## Getting started 1. Update ComfyUI to the latest version diff --git a/tutorials/partner-nodes/wan/wan2-7.mdx b/tutorials/partner-nodes/wan/wan2-7.mdx index 16d3b46e3..762a63724 100644 --- a/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/tutorials/partner-nodes/wan/wan2-7.mdx @@ -33,46 +33,58 @@ This release brings a fully upgraded multimodal video pipeline directly into you Generate video from image inputs. Supports first-frame, first+last-frame, and audio-driven generation modes. - +Wan2.7 I2V workflow preview + + + Get the Wan2.7 Image-to-Video workflow file. - - + Try the Image-to-Video workflow instantly on Comfy Cloud. + ## Wan2.7 text-to-video Generate video from pure text prompts. Optionally include audio input and multi-shot narration for richer storytelling. - +Wan2.7 T2V workflow preview + + + Get the Wan2.7 Text-to-Video workflow file. - - + Try the Text-to-Video workflow instantly on Comfy Cloud. + ## Wan2.7 reference-to-video Use reference images of a subject's visual appearance along with an optional vocal timbre reference. Supports up to 5 real-person inputs for multi-character interaction scenes. - +Wan2.7 R2V workflow preview + + + Get the Wan2.7 Reference-to-Video workflow file. - - + Try the Reference-to-Video workflow instantly on Comfy Cloud. + ## Wan2.7 video edit Edit or replicate existing videos using text prompts, a reference image, or style transfer. - +Wan2.7 Video Edit workflow preview + + + Get the Wan2.7 Video Edit workflow file. - - + Try the Video Edit workflow instantly on Comfy Cloud. + From 225a7ad306ea5b0c81772173c9fa1501cc739c47 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Thu, 16 Jul 2026 22:20:16 +0800 Subject: [PATCH 08/19] fix: add previews and input cards to tripo, rodin, wan2-7 --- .../partner-nodes/ideogram/ideogram-v4.mdx | 2 ++ .../kling/kling-motion-control.mdx | 17 +++++++++++ tutorials/partner-nodes/luma/luma-uni-1.mdx | 28 +++++++++++++++++++ tutorials/partner-nodes/topaz/astra-2.mdx | 14 ++++++++++ 4 files changed, 61 insertions(+) diff --git a/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 486e9ca90..7fe08ea6c 100644 --- a/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -14,6 +14,8 @@ Ideogram 4.0 is the latest text-to-image model from Ideogram, offering superior ## Ideogram 4.0 Partner Node Text-to-Image Workflow +Ideogram 4.0 Text-to-Image workflow preview + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/kling/kling-motion-control.mdx b/tutorials/partner-nodes/kling/kling-motion-control.mdx index c51dc8673..16de4a9a3 100644 --- a/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -40,6 +40,8 @@ The `character_orientation` parameter determines how the model interprets spatia ## Kling 2.6 Motion Control workflow +Kling 2.6 Motion Control workflow preview + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. @@ -48,6 +50,21 @@ The `character_orientation` parameter determines how the model interprets spatia Download the workflow JSON file for local use. +
+Input materials + +Download these sample input files to try the workflow: + + + + Download sample reference image + + + Download sample motion reference video + + +
+ ## Input requirements - **Image formats**: JPG, PNG, WEBP, GIF, AVIF (max 10MB) diff --git a/tutorials/partner-nodes/luma/luma-uni-1.mdx b/tutorials/partner-nodes/luma/luma-uni-1.mdx index 573cb6aec..fa35daacb 100644 --- a/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -55,6 +55,8 @@ When in doubt: ### Image Create workflow +Luma Uni-1 Image Create workflow preview + Try the Image Create workflow instantly on Comfy Cloud. @@ -64,8 +66,22 @@ When in doubt: +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample style reference + + +
+ ### Image Edit workflow +Luma Uni-1 Image Edit workflow preview + Try the Image Edit workflow instantly on Comfy Cloud. @@ -75,6 +91,18 @@ When in doubt: +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample source image + + +
+ The workflow is simple: **prompt → evaluate → refine**. Leave the seed blank while exploring. Once you find something strong, lock the seed and iterate from there. ## Core parameters diff --git a/tutorials/partner-nodes/topaz/astra-2.mdx b/tutorials/partner-nodes/topaz/astra-2.mdx index ffcb9a5aa..066e6eba1 100644 --- a/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/tutorials/partner-nodes/topaz/astra-2.mdx @@ -36,6 +36,8 @@ Optional **Prompt**: max **450** frames (~15 s @ 30 fps). Empty prompt: up t **Load Video** → **Topaz Video Enhance** → **Save Video**. After opening the template, pick **Astra 2** on the upscaler if another preset is selected. +Astra 2 workflow preview + Open the workflow instantly on Comfy Cloud. @@ -44,6 +46,18 @@ Optional **Prompt**: max **450** frames (~15 s @ 30 fps). Empty prompt: up t Download the workflow JSON for local ComfyUI. +
+Input materials + +Download this sample input video to try the workflow: + + + + Download sample input video + + +
+ ## Related - [Video upscaling in ComfyUI](/tutorials/utility/video-upscale) — guidance on choosing an upscaling approach. From fbe22b8c833fbfdaedfd8346ccedf5330d0a683b Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Wed, 22 Jul 2026 14:01:16 +0800 Subject: [PATCH 09/19] fix: remove translation files and em dashes from tutorial-standardize PR - Revert all ja, zh, ko translation file changes (246 files) Translation file changes must go through the pipeline - Replace 16 em dashes with colons in 5 English tutorial files per project prose style guide --- ja/tutorials/3d/hunyuan3D-2.mdx | 39 +++----- ja/tutorials/3d/triposplat.mdx | 10 +- ja/tutorials/audio/ace-step/ace-step-v1-5.mdx | 10 +- ja/tutorials/audio/ace-step/ace-step-v1.mdx | 36 ++----- .../audio/stable-audio/stable-audio-1.mdx | 4 +- .../audio/stable-audio/stable-audio-3.mdx | 8 +- ja/tutorials/basic/inpaint.mdx | 4 +- ja/tutorials/basic/outpaint.mdx | 2 +- ja/tutorials/controlnet/controlnet.mdx | 4 +- ja/tutorials/controlnet/depth-controlnet.mdx | 2 +- ja/tutorials/controlnet/depth-t2i-adapter.mdx | 2 +- .../controlnet/mixing-controlnets.mdx | 6 +- .../controlnet/pose-controlnet-2-pass.mdx | 4 +- ja/tutorials/flux/flux-1-controlnet.mdx | 40 +++----- ja/tutorials/flux/flux-1-fill-dev.mdx | 32 +++--- ja/tutorials/flux/flux-1-kontext-dev.mdx | 15 +-- ja/tutorials/flux/flux-1-text-to-image.mdx | 70 +++++-------- ja/tutorials/flux/flux-1-uso.mdx | 28 +++--- ja/tutorials/flux/flux-2-dev.mdx | 6 +- ja/tutorials/flux/flux-2-klein.mdx | 12 +-- ja/tutorials/flux/flux1-krea-dev.mdx | 25 +++-- ja/tutorials/image/anima/anima.mdx | 8 +- ja/tutorials/image/boogu/boogu-image-0.1.mdx | 14 +-- .../image/cosmos/cosmos-predict2-t2i.mdx | 6 +- .../image/ernie-image/ernie-image.mdx | 16 +-- ja/tutorials/image/hidream/hidream-e1.mdx | 24 +---- ja/tutorials/image/hidream/hidream-i1.mdx | 45 +++------ ja/tutorials/image/hidream/hidream-o1.mdx | 22 ++--- ja/tutorials/image/ideogram/ideogram-v4.mdx | 10 +- ja/tutorials/image/krea/krea-2.mdx | 6 +- ja/tutorials/image/lens/lens.mdx | 8 +- .../newbie-image/newbie-image-exp-0-1.mdx | 23 +++-- ja/tutorials/image/omnigen/omnigen2.mdx | 24 ++--- ja/tutorials/image/ovis/ovis-image.mdx | 21 ++-- ja/tutorials/image/pixeldit/pixeldit.mdx | 4 +- ja/tutorials/image/qwen/qwen-image-2512.mdx | 22 ++--- .../image/qwen/qwen-image-edit-2511.mdx | 23 +++-- ja/tutorials/image/qwen/qwen-image-edit.mdx | 23 +++-- .../image/qwen/qwen-image-layered.mdx | 24 +++-- ja/tutorials/image/qwen/qwen-image.mdx | 81 ++++++++------- ja/tutorials/image/z-image/z-image-turbo.mdx | 8 +- ja/tutorials/image/z-image/z-image.mdx | 6 +- ja/tutorials/llm/gemma4/gemma4.mdx | 4 +- ja/tutorials/llm/qwen/qwen3.mdx | 6 +- ja/tutorials/llm/qwen/qwen3_5.mdx | 6 +- ja/tutorials/partner-nodes/google/gemini.mdx | 12 ++- .../kling/kling-motion-control.mdx | 5 +- .../moonvalley-video-generation.mdx | 24 ++--- ja/tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 ++-- .../partner-nodes/runway/video-generation.mdx | 12 ++- .../partner-nodes/tripo/model-generation.mdx | 24 ++--- ja/tutorials/utility/depth-anything-3.mdx | 8 +- .../utility/face-detection/mediapipe.mdx | 2 +- ja/tutorials/utility/moge.mdx | 4 +- .../utility/pose-detection-sdpose.mdx | 6 +- .../utility/remove-background-birefnet.mdx | 2 +- ja/tutorials/utility/video-segment-sam3.mdx | 2 +- .../utility/void-video-inpainting.mdx | 12 +-- ja/tutorials/video/bytedance/bernini-r.mdx | 8 +- .../cosmos/cosmos-predict2-video2world.mdx | 47 +++------ .../video/hunyuan/hunyuan-video-1-5.mdx | 10 +- ja/tutorials/video/hunyuan/hunyuan-video.mdx | 14 +-- ja/tutorials/video/kandinsky/kandinsky-5.mdx | 64 +++--------- ja/tutorials/video/ltxv.mdx | 32 +----- ja/tutorials/video/wan/fun-camera.mdx | 98 +++++-------------- ja/tutorials/video/wan/fun-control.mdx | 10 +- ja/tutorials/video/wan/fun-inp.mdx | 12 +-- ja/tutorials/video/wan/vace.mdx | 10 +- ja/tutorials/video/wan/wan-ati.mdx | 10 +- ja/tutorials/video/wan/wan-causal-forcing.mdx | 8 +- ja/tutorials/video/wan/wan-dancer.mdx | 12 +-- ja/tutorials/video/wan/wan-flf.mdx | 10 +- ja/tutorials/video/wan/wan-move.mdx | 42 ++++---- ja/tutorials/video/wan/wan-video.mdx | 14 +-- ja/tutorials/video/wan/wan2-2-animate.mdx | 56 ++++------- ja/tutorials/video/wan/wan2-2-fun-camera.mdx | 53 +++------- ja/tutorials/video/wan/wan2-2-fun-control.mdx | 72 +++++--------- ja/tutorials/video/wan/wan2-2-fun-inp.mdx | 49 +++------- ja/tutorials/video/wan/wan2-2-s2v.mdx | 72 +++++--------- ja/tutorials/video/wan/wan2_2.mdx | 82 ++++++++-------- ja/tutorials/video/zai/scail2.mdx | 14 +-- ko/tutorials/3d/hunyuan3D-2.mdx | 39 +++----- ko/tutorials/3d/triposplat.mdx | 10 +- ko/tutorials/audio/ace-step/ace-step-v1-5.mdx | 10 +- ko/tutorials/audio/ace-step/ace-step-v1.mdx | 36 ++----- .../audio/stable-audio/stable-audio-1.mdx | 4 +- .../audio/stable-audio/stable-audio-3.mdx | 8 +- ko/tutorials/basic/inpaint.mdx | 4 +- ko/tutorials/basic/outpaint.mdx | 2 +- ko/tutorials/controlnet/controlnet.mdx | 4 +- ko/tutorials/controlnet/depth-controlnet.mdx | 2 +- ko/tutorials/controlnet/depth-t2i-adapter.mdx | 2 +- .../controlnet/mixing-controlnets.mdx | 6 +- .../controlnet/pose-controlnet-2-pass.mdx | 4 +- ko/tutorials/flux/flux-1-controlnet.mdx | 40 +++----- ko/tutorials/flux/flux-1-fill-dev.mdx | 32 +++--- ko/tutorials/flux/flux-1-kontext-dev.mdx | 21 ++-- ko/tutorials/flux/flux-1-text-to-image.mdx | 70 +++++-------- ko/tutorials/flux/flux-1-uso.mdx | 28 +++--- ko/tutorials/flux/flux-2-dev.mdx | 6 +- ko/tutorials/flux/flux-2-klein.mdx | 12 +-- ko/tutorials/flux/flux1-krea-dev.mdx | 25 +++-- ko/tutorials/image/anima/anima.mdx | 8 +- ko/tutorials/image/boogu/boogu-image-0.1.mdx | 14 +-- .../image/cosmos/cosmos-predict2-t2i.mdx | 6 +- .../image/ernie-image/ernie-image.mdx | 16 +-- ko/tutorials/image/hidream/hidream-e1.mdx | 24 +---- ko/tutorials/image/hidream/hidream-i1.mdx | 45 +++------ ko/tutorials/image/hidream/hidream-o1.mdx | 22 ++--- ko/tutorials/image/ideogram/ideogram-v4.mdx | 10 +- ko/tutorials/image/krea/krea-2.mdx | 6 +- ko/tutorials/image/lens/lens.mdx | 8 +- .../newbie-image/newbie-image-exp-0-1.mdx | 23 +++-- ko/tutorials/image/omnigen/omnigen2.mdx | 22 ++--- ko/tutorials/image/ovis/ovis-image.mdx | 21 ++-- ko/tutorials/image/pixeldit/pixeldit.mdx | 4 +- ko/tutorials/image/qwen/qwen-image-2512.mdx | 22 ++--- .../image/qwen/qwen-image-edit-2511.mdx | 23 +++-- ko/tutorials/image/qwen/qwen-image-edit.mdx | 23 +++-- .../image/qwen/qwen-image-layered.mdx | 22 ++--- ko/tutorials/image/qwen/qwen-image.mdx | 81 ++++++++------- ko/tutorials/image/z-image/z-image-turbo.mdx | 8 +- ko/tutorials/image/z-image/z-image.mdx | 6 +- ko/tutorials/llm/gemma4/gemma4.mdx | 4 +- ko/tutorials/llm/qwen/qwen3.mdx | 6 +- ko/tutorials/llm/qwen/qwen3_5.mdx | 6 +- ko/tutorials/partner-nodes/google/gemini.mdx | 12 ++- .../kling/kling-motion-control.mdx | 5 +- .../moonvalley-video-generation.mdx | 24 ++--- ko/tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 ++-- .../partner-nodes/runway/video-generation.mdx | 12 ++- .../partner-nodes/tripo/model-generation.mdx | 24 ++--- ko/tutorials/utility/depth-anything-3.mdx | 8 +- .../utility/face-detection/mediapipe.mdx | 2 +- ko/tutorials/utility/moge.mdx | 4 +- .../utility/pose-detection-sdpose.mdx | 6 +- .../utility/remove-background-birefnet.mdx | 2 +- ko/tutorials/utility/video-segment-sam3.mdx | 2 +- .../utility/void-video-inpainting.mdx | 12 +-- ko/tutorials/video/bytedance/bernini-r.mdx | 8 +- .../cosmos/cosmos-predict2-video2world.mdx | 47 +++------ .../video/hunyuan/hunyuan-video-1-5.mdx | 10 +- ko/tutorials/video/hunyuan/hunyuan-video.mdx | 14 +-- ko/tutorials/video/kandinsky/kandinsky-5.mdx | 64 +++--------- ko/tutorials/video/ltxv.mdx | 32 +----- ko/tutorials/video/wan/fun-camera.mdx | 98 +++++-------------- ko/tutorials/video/wan/fun-control.mdx | 10 +- ko/tutorials/video/wan/fun-inp.mdx | 12 +-- ko/tutorials/video/wan/vace.mdx | 10 +- ko/tutorials/video/wan/wan-ati.mdx | 10 +- ko/tutorials/video/wan/wan-causal-forcing.mdx | 8 +- ko/tutorials/video/wan/wan-dancer.mdx | 12 +-- ko/tutorials/video/wan/wan-flf.mdx | 10 +- ko/tutorials/video/wan/wan-move.mdx | 42 ++++---- ko/tutorials/video/wan/wan-video.mdx | 14 +-- ko/tutorials/video/wan/wan2-2-animate.mdx | 56 ++++------- ko/tutorials/video/wan/wan2-2-fun-camera.mdx | 53 +++------- ko/tutorials/video/wan/wan2-2-fun-control.mdx | 70 +++++-------- ko/tutorials/video/wan/wan2-2-fun-inp.mdx | 49 +++------- ko/tutorials/video/wan/wan2-2-s2v.mdx | 68 ++++--------- ko/tutorials/video/wan/wan2_2.mdx | 82 ++++++++-------- ko/tutorials/video/zai/scail2.mdx | 14 +-- tutorials/flux/flux-1-kontext-dev.mdx | 6 +- tutorials/video/wan/fun-inp.mdx | 4 +- tutorials/video/wan/wan-alpha.mdx | 2 +- tutorials/video/wan/wan-ati.mdx | 4 +- tutorials/video/wan/wan-flf.mdx | 16 +-- zh/tutorials/3d/hunyuan3D-2.mdx | 39 +++----- zh/tutorials/3d/triposplat.mdx | 10 +- zh/tutorials/audio/ace-step/ace-step-v1-5.mdx | 10 +- zh/tutorials/audio/ace-step/ace-step-v1.mdx | 36 ++----- .../audio/stable-audio/stable-audio-1.mdx | 4 +- .../audio/stable-audio/stable-audio-3.mdx | 8 +- zh/tutorials/basic/inpaint.mdx | 4 +- zh/tutorials/basic/outpaint.mdx | 2 +- zh/tutorials/controlnet/controlnet.mdx | 4 +- zh/tutorials/controlnet/depth-controlnet.mdx | 2 +- zh/tutorials/controlnet/depth-t2i-adapter.mdx | 2 +- .../controlnet/mixing-controlnets.mdx | 6 +- .../controlnet/pose-controlnet-2-pass.mdx | 4 +- zh/tutorials/flux/flux-1-controlnet.mdx | 40 +++----- zh/tutorials/flux/flux-1-fill-dev.mdx | 34 +++---- zh/tutorials/flux/flux-1-kontext-dev.mdx | 21 ++-- zh/tutorials/flux/flux-1-text-to-image.mdx | 70 +++++-------- zh/tutorials/flux/flux-1-uso.mdx | 28 +++--- zh/tutorials/flux/flux-2-dev.mdx | 6 +- zh/tutorials/flux/flux-2-klein.mdx | 12 +-- zh/tutorials/flux/flux1-krea-dev.mdx | 25 +++-- zh/tutorials/image/anima/anima.mdx | 8 +- zh/tutorials/image/boogu/boogu-image-0.1.mdx | 14 +-- .../image/cosmos/cosmos-predict2-t2i.mdx | 6 +- .../image/ernie-image/ernie-image.mdx | 16 +-- zh/tutorials/image/hidream/hidream-e1.mdx | 24 +---- zh/tutorials/image/hidream/hidream-i1.mdx | 45 +++------ zh/tutorials/image/hidream/hidream-o1.mdx | 22 ++--- zh/tutorials/image/ideogram/ideogram-v4.mdx | 10 +- zh/tutorials/image/krea/krea-2.mdx | 6 +- zh/tutorials/image/lens/lens.mdx | 8 +- .../newbie-image/newbie-image-exp-0-1.mdx | 23 +++-- zh/tutorials/image/omnigen/omnigen2.mdx | 22 ++--- zh/tutorials/image/ovis/ovis-image.mdx | 21 ++-- zh/tutorials/image/pixeldit/pixeldit.mdx | 4 +- zh/tutorials/image/qwen/qwen-image-2512.mdx | 22 ++--- .../image/qwen/qwen-image-edit-2511.mdx | 23 +++-- zh/tutorials/image/qwen/qwen-image-edit.mdx | 23 +++-- .../image/qwen/qwen-image-layered.mdx | 24 +++-- zh/tutorials/image/qwen/qwen-image.mdx | 81 ++++++++------- zh/tutorials/image/z-image/z-image-turbo.mdx | 8 +- zh/tutorials/image/z-image/z-image.mdx | 6 +- zh/tutorials/llm/gemma4/gemma4.mdx | 4 +- zh/tutorials/llm/qwen/qwen3.mdx | 6 +- zh/tutorials/llm/qwen/qwen3_5.mdx | 6 +- zh/tutorials/partner-nodes/google/gemini.mdx | 12 ++- .../kling/kling-motion-control.mdx | 5 +- .../moonvalley-video-generation.mdx | 24 ++--- zh/tutorials/partner-nodes/openai/chat.mdx | 10 +- .../partner-nodes/rodin/model-generation.mdx | 20 ++-- .../partner-nodes/runway/video-generation.mdx | 12 ++- .../partner-nodes/tripo/model-generation.mdx | 24 ++--- zh/tutorials/utility/depth-anything-3.mdx | 8 +- .../utility/face-detection/mediapipe.mdx | 2 +- zh/tutorials/utility/moge.mdx | 4 +- .../utility/pose-detection-sdpose.mdx | 6 +- .../utility/remove-background-birefnet.mdx | 2 +- zh/tutorials/utility/video-segment-sam3.mdx | 2 +- .../utility/void-video-inpainting.mdx | 12 +-- zh/tutorials/video/bytedance/bernini-r.mdx | 8 +- .../cosmos/cosmos-predict2-video2world.mdx | 57 ++++------- .../video/hunyuan/hunyuan-video-1-5.mdx | 10 +- zh/tutorials/video/hunyuan/hunyuan-video.mdx | 14 +-- zh/tutorials/video/kandinsky/kandinsky-5.mdx | 64 +++--------- zh/tutorials/video/ltxv.mdx | 32 +----- zh/tutorials/video/wan/fun-camera.mdx | 98 +++++-------------- zh/tutorials/video/wan/fun-control.mdx | 10 +- zh/tutorials/video/wan/fun-inp.mdx | 12 +-- zh/tutorials/video/wan/vace.mdx | 10 +- zh/tutorials/video/wan/wan-ati.mdx | 10 +- zh/tutorials/video/wan/wan-causal-forcing.mdx | 8 +- zh/tutorials/video/wan/wan-dancer.mdx | 12 +-- zh/tutorials/video/wan/wan-flf.mdx | 10 +- zh/tutorials/video/wan/wan-move.mdx | 42 ++++---- zh/tutorials/video/wan/wan-video.mdx | 14 +-- zh/tutorials/video/wan/wan2-2-animate.mdx | 57 ++++------- zh/tutorials/video/wan/wan2-2-fun-camera.mdx | 57 +++-------- zh/tutorials/video/wan/wan2-2-fun-control.mdx | 72 +++++--------- zh/tutorials/video/wan/wan2-2-fun-inp.mdx | 53 +++------- zh/tutorials/video/wan/wan2-2-s2v.mdx | 67 ++++--------- zh/tutorials/video/wan/wan2_2.mdx | 82 ++++++++-------- zh/tutorials/video/zai/scail2.mdx | 14 +-- 251 files changed, 2091 insertions(+), 3162 deletions(-) diff --git a/ja/tutorials/3d/hunyuan3D-2.mdx b/ja/tutorials/3d/hunyuan3D-2.mdx index d815f9748..35f163849 100644 --- a/ja/tutorials/3d/hunyuan3D-2.mdx +++ b/ja/tutorials/3d/hunyuan3D-2.mdx @@ -57,14 +57,9 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して - - - Comfy Cloud でこのワークフローをすぐに実行 - - - ワークフロー JSON ファイルをダウンロード - - + +

Run on Comfy Cloud

+ ### 1. ワークフロー @@ -86,7 +81,7 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます ``` ComfyUI/ @@ -109,14 +104,9 @@ ComfyUI/ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを使用して 3D モデルを生成します。このモデルは Hunyuan3D-2mv のステップ蒸留バージョンで、より高速な 3D モデル生成を可能にします。このバージョンのワークフローでは、`cfg` を 1.0 に設定し、`flux guidance` ノードを追加して `distilled cfg` 生成を制御します。 - - - Comfy Cloud でこのワークフローをすぐに実行 - - - ワークフロー JSON ファイルをダウンロード - - + +

Run on Comfy Cloud

+ ### 1. ワークフロー @@ -135,7 +125,7 @@ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます ``` ComfyUI/ @@ -156,14 +146,9 @@ ComfyUI/ Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D モデルを生成します。このモデルはマルチビューモデルではありません。このワークフローでは、`Hunyuan3Dv2ConditioningMultiView` ノードの代わりに `Hunyuan3Dv2Conditioning` ノードを使用します。 - - - Comfy Cloud でこのワークフローをすぐに実行 - - - ワークフロー JSON ファイルをダウンロード - - + +

Run on Comfy Cloud

+ ### 1. ワークフロー @@ -178,7 +163,7 @@ Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます ``` ComfyUI/ diff --git a/ja/tutorials/3d/triposplat.mdx b/ja/tutorials/3d/triposplat.mdx index 0db4907a4..3c1dbffce 100644 --- a/ja/tutorials/3d/triposplat.mdx +++ b/ja/tutorials/3d/triposplat.mdx @@ -106,23 +106,23 @@ TripoSplat は **フィードフォワードアーキテクチャ** を使用し TripoSplat モデルと必要なファイルをダウンロードします。対応する `models/` サブディレクトリに配置してください。 - + triposplat_fp16.safetensors — TripoSplat 拡散モデルチェックポイント - + triposplat_vae_decoder_fp16.safetensors — VAE デコーダー - + flux2-vae.safetensors — Flux.2 VAE、潜在表現エンコード用 - + dino_v3_vit_h.safetensors — CLIP ビジョンエンコーダー（DINOv2） - + birefnet.safetensors — 前処理用の背景除去モデル diff --git a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx index 72d03af91..97dd4f880 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -45,7 +45,7 @@ AIO（All-in-One）版は、すべてのモデルを単一のチェックポイ ### AIOモデルのダウンロード - + オールインワンチェックポイントファイル（大多数のユーザーに推奨）。 @@ -74,19 +74,19 @@ AIO（All-in-One）版は、すべてのモデルを単一のチェックポイ ### 分割モデルのダウンロード - + 拡散モデル（Diffusion Model）。 - + テキストエンコーダー（0.6B）。 - + テキストエンコーダー（1.7B）。 - + VAEモデル。 diff --git a/ja/tutorials/audio/ace-step/ace-step-v1.mdx b/ja/tutorials/audio/ace-step/ace-step-v1.mdx index 1e36359f5..d558882bf 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1.mdx @@ -33,15 +33,9 @@ ACE-Step は、中国のチーム StepFun と ACE Studio が共同開発した 以下のボタンをクリックして、対応するワークフローファイルをダウンロードしてください。ダウンロードしたファイルを ComfyUI にドラッグ＆ドロップすることで、ワークフロー情報が読み込まれます。このワークフローには、モデルのダウンロード情報も含まれています。 - - - JSON 形式のワークフローファイルをダウンロード - - + +

JSON 形式のワークフローファイルをダウンロード

+ また、[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors) を手動でダウンロードし、`ComfyUI/models/checkpoints` フォルダーに保存することもできます。 @@ -64,27 +58,15 @@ ACE-Step は、中国のチーム StepFun と ACE Studio が共同開発した 以下のボタンをクリックして、対応するワークフローファイルをダウンロードしてください。ダウンロードしたファイルを ComfyUI にドラッグ＆ドロップすることで、ワークフロー情報が読み込まれます。 - - - JSON 形式のワークフローファイルをダウンロード - - + +

JSON 形式のワークフローファイルをダウンロード

+ 以下の音声ファイルを入力音声としてダウンロードしてください： - - - 入力用のサンプル音声ファイルをダウンロード - - + +

入力用のサンプル音声ファイルをダウンロード

+ ### 2. ワークフローをステップごとに実行 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx index 258bd2f9f..a0c3bcfa4 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -41,7 +41,7 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" ### チェックポイント - + 2.3GB。models/checkpoints/ に配置 @@ -56,7 +56,7 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" ### テキストエンコーダー - + プロンプト処理用テキストエンコーダー。models/text_encoders/ に配置 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx index c21d44138..e638d2d56 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -77,11 +77,11 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 ### チェックポイント - + Medium ワークフロー用。models/checkpoints/ に配置 - + Medium Base ワークフロー用。models/checkpoints/ に配置 @@ -97,11 +97,11 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 ### テキストエンコーダー - + すべての Stable Audio 3 ワークフローで必要。models/text_encoders/ に配置 - + Medium ワークフローで必要（Qwen リプロンプト）。models/text_encoders/ に配置 diff --git a/ja/tutorials/basic/inpaint.mdx b/ja/tutorials/basic/inpaint.mdx index 2d4a7298f..d97d65da2 100644 --- a/ja/tutorials/basic/inpaint.mdx +++ b/ja/tutorials/basic/inpaint.mdx @@ -35,7 +35,7 @@ AIによる画像生成において、全体としては満足できる画像が #### 1. モデルのインストール 以下のファイルをダウンロードし、`ComfyUI/models/checkpoints` フォルダに保存してください： -[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) +[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) #### 2. 局部再描画用の入力画像 @@ -69,7 +69,7 @@ AIによる画像生成において、全体としては満足できる画像が ![SD1.5 インペインティング結果](/images/tutorial/basic/inpaint/inpaint_sd1.5_pruned_emaonly.png) -一方、[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) モデルを用いた結果は、より高品質なインペインティング効果と、自然な境界のトランジションが得られます。 +一方、[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) モデルを用いた結果は、より高品質なインペインティング効果と、自然な境界のトランジションが得られます。 これは、当該モデルが**インペインティング専用に設計・最適化**されているためであり、生成領域をより正確に制御でき、結果として優れた再描画品質を実現します。 先ほどご紹介した「画家」というアナロジーを思い出してください。異なるモデルは、それぞれ得意分野や限界を持つ「画家」のような存在です。適切なモデルを選択することで、より理想的な生成結果を得ることが可能になります。 diff --git a/ja/tutorials/basic/outpaint.mdx b/ja/tutorials/basic/outpaint.mdx index 4d09d7120..9a96f5004 100644 --- a/ja/tutorials/basic/outpaint.mdx +++ b/ja/tutorials/basic/outpaint.mdx @@ -34,7 +34,7 @@ AI画像生成においては、既存の画像の構図が優れているもの #### 1. モデルのインストール 以下のモデルファイルをダウンロードし、`ComfyUI/models/checkpoints` ディレクトリに保存してください： -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) #### 2. 入力画像 diff --git a/ja/tutorials/controlnet/controlnet.mdx b/ja/tutorials/controlnet/controlnet.mdx index bacd4e2ec..42acac85e 100644 --- a/ja/tutorials/controlnet/controlnet.mdx +++ b/ja/tutorials/controlnet/controlnet.mdx @@ -76,8 +76,8 @@ ControlNet の登場により、追加の条件を導入することで画像生 - [dreamCreationVirtual3DECommerce_v10.safetensors](https://civitai.com/api/download/models/731340?type=Model&format=SafeTensor&size=full&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/depth-controlnet.mdx b/ja/tutorials/controlnet/depth-controlnet.mdx index ea0da6d71..19637bc3d 100644 --- a/ja/tutorials/controlnet/depth-controlnet.mdx +++ b/ja/tutorials/controlnet/depth-controlnet.mdx @@ -54,7 +54,7 @@ Depth ControlNet は、深度マップの情報を理解・活用するために - [architecturerealmix_v11.safetensors](https://civitai.com/api/download/models/431755?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) +- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/depth-t2i-adapter.mdx b/ja/tutorials/controlnet/depth-t2i-adapter.mdx index 2541de169..f6ae26c4c 100644 --- a/ja/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/ja/tutorials/controlnet/depth-t2i-adapter.mdx @@ -76,7 +76,7 @@ ComfyUI における T2I Adapter の使用方法は、インターフェース - [interiordesignsuperm_v2.safetensors](https://civitai.com/api/download/models/93152?type=Model&format=SafeTensor&size=full&fp=fp16) -- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_depth_sd15v2.pth?download=true) +- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/resolve/main/models/t2iadapter_depth_sd15v2.pth?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/mixing-controlnets.mdx b/ja/tutorials/controlnet/mixing-controlnets.mdx index 7f65a43fd..a9990bd85 100644 --- a/ja/tutorials/controlnet/mixing-controlnets.mdx +++ b/ja/tutorials/controlnet/mixing-controlnets.mdx @@ -54,9 +54,9 @@ AI による画像生成において、単一の制御条件では複雑なシ - [awpainting_v14.safetensors](https://civitai.com/api/download/models/624939?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx b/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx index b4dca8f4d..aacef9020 100644 --- a/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -49,10 +49,10 @@ AI による画像生成において、OpenPose で生成された骨格構造 ネットワーク環境によっては、対応するモデルの自動ダウンロードが失敗する場合があります。その場合は、以下のモデルを手動でダウンロードし、指定されたディレクトリに配置してください： -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) - [majicmixRealistic_v7.safetensors](https://civitai.com/api/download/models/176425?type=Model&format=SafeTensor&size=pruned&fp=fp16) - [japaneseStyleRealistic_v20.safetensors](https://civitai.com/api/download/models/85426?type=Model&format=SafeTensor&size=pruned&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ja/tutorials/flux/flux-1-controlnet.mdx b/ja/tutorials/flux/flux-1-controlnet.mdx index 9448ed209..39c15df7e 100644 --- a/ja/tutorials/flux/flux-1-controlnet.mdx +++ b/ja/tutorials/flux/flux-1-controlnet.mdx @@ -49,14 +49,9 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 ## FLUX.1-Canny-dev 完全版ワークフロー - - - JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Canny」を検索してください - - - Comfy Cloud で開く - - + +

Comfy Cloud で実行

+ ### 1. ワークフローおよび関連アセット @@ -77,10 +72,10 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） ファイルの保存先ディレクトリ構成： ``` @@ -120,14 +115,9 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora ワークフロー - - - JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Depth LoRA」を検索してください - - - Comfy Cloud で開く - - + +

Comfy Cloud で実行

+ LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを追加したものであり、[Flux ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image) と比較して、対応する LoRA モデルを読み込むためのノードが追加されています。 @@ -148,11 +138,11 @@ LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) ファイルの保存先ディレクトリ構成： ``` diff --git a/ja/tutorials/flux/flux-1-fill-dev.mdx b/ja/tutorials/flux/flux-1-fill-dev.mdx index b6bb401b4..9e8b9f12e 100644 --- a/ja/tutorials/flux/flux-1-fill-dev.mdx +++ b/ja/tutorials/flux/flux-1-fill-dev.mdx @@ -39,10 +39,10 @@ Inpainting や Outpainting のワークフローについてまだご存じな ![Flux Agreement](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) ファイルの保存場所： ``` @@ -61,14 +61,13 @@ ComfyUI/ ### 1. Inpainting ワークフローおよび関連アセット - - - Download JSON or search "flux_fill_inpaint" in Template Library - - - Open in Comfy Cloud - - + +

ワークフロー画像をダウンロード

+ + + +

Comfy Cloud で実行

+ 以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 ![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -97,15 +96,6 @@ ComfyUI/ ### 1. Outpainting ワークフローおよび関連アセット - - - Download JSON or search "flux_fill_outpaint" in Template Library - - - Open in Comfy Cloud - - - 以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 ![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/ja/tutorials/flux/flux-1-kontext-dev.mdx b/ja/tutorials/flux/flux-1-kontext-dev.mdx index 5933115ec..b9537bda5 100644 --- a/ja/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ja/tutorials/flux/flux-1-kontext-dev.mdx @@ -51,7 +51,7 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： **Diffusion Model（拡散モデル）** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) オリジナルの重み（weights）を使用したい場合は、Black Forest Labs の関連リポジトリからオリジナルモデルの重みを取得・利用できます。 @@ -62,7 +62,7 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： **Text Encoder（テキストエンコーダー）** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) または [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) または [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) モデルの保存先 @@ -80,14 +80,9 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： ## Flux.1 Kontext Dev ワークフロー - - - JSONをダウンロードするか、テンプレートライブラリで「Flux Kontext Dev」を検索してください - - - Comfy Cloud で開く - - + +

Comfy Cloud で実行

+ このワークフローでは、編集対象の画像を読み込むために `Load Image(from output)` ノードを採用しており、編集後の画像を容易に取得・再利用できるため、複数回の反復編集がよりスムーズに行えます。 diff --git a/ja/tutorials/flux/flux-1-text-to-image.mdx b/ja/tutorials/flux/flux-1-text-to-image.mdx index 1e19bd6f7..51b98dd27 100644 --- a/ja/tutorials/flux/flux-1-text-to-image.mdx +++ b/ja/tutorials/flux/flux-1-text-to-image.mdx @@ -50,30 +50,25 @@ Flux は、優れた画像品質と高い柔軟性で知られており、高品 #### 1. ワークフロー・ファイル - - - Comfy Cloud でこのワークフローを実行 - - - JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Dev」を検索 - - - 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Dev オリジナルバージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) + +

Comfy Cloud で実行

+ + #### 2. モデルの手動インストール - `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) のライセンス契約に同意する必要があります。 -- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 +- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 以下のモデルファイルをダウンロードしてください： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) 保存先のディレクトリ構成： ``` @@ -108,19 +103,14 @@ Flux の優れたプロンプト追従能力により、負のプロンプト（ #### 1. ワークフロー・ファイル - - - Comfy Cloud でこのワークフローを実行 - - - JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Schnell」を検索 - - - 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Schnell バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) + +

Comfy Cloud で実行

+ + #### 2. モデルの手動インストール @@ -130,10 +120,10 @@ Flux の優れたプロンプト追従能力により、負のプロンプト（ 完全なモデルファイル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) ファイルの保存先ディレクトリ構成： ``` @@ -166,38 +156,24 @@ FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したも ### Flux.1 Dev - - - Comfy Cloud でこのワークフローを実行 - - - JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Dev FP8」を検索 - - - 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Dev fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 + +

Comfy Cloud で実行

+ + +[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 対応する `Load Checkpoint` ノードが `flux1-dev-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 ### Flux.1 Schnell - - - Comfy Cloud でこのワークフローを実行 - - - JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Schnell FP8」を検索 - - - 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Schnell fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 +[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 対応する `Load Checkpoint` ノードが `flux1-schnell-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 diff --git a/ja/tutorials/flux/flux-1-uso.mdx b/ja/tutorials/flux/flux-1-uso.mdx index 81c03b2c7..9e8790c9a 100644 --- a/ja/tutorials/flux/flux-1-uso.mdx +++ b/ja/tutorials/flux/flux-1-uso.mdx @@ -31,14 +31,18 @@ USO は以下の3つの主要なアプローチをサポートします： ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - - - ワークフロー JSON をダウンロードし、ComfyUI にドラッグ＆ドロップしてください - - - Comfy Cloud でこのワークフローを実行 - - + +

JSON ワークフローをダウンロード

+ + + +

Comfy Cloud で実行

+ 以下の画像を入力画像として使用します。 @@ -48,18 +52,18 @@ USO は以下の3つの主要なアプローチをサポートします： **checkpoints** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) **loras** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **model_patches** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **clip_visions** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) 上記すべてのモデルをダウンロードし、以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/flux/flux-2-dev.mdx b/ja/tutorials/flux/flux-2-dev.mdx index 2667de564..653f78180 100644 --- a/ja/tutorials/flux/flux-2-dev.mdx +++ b/ja/tutorials/flux/flux-2-dev.mdx @@ -65,15 +65,15 @@ FLUX.2 Dev を用いた基本的なテキストから画像への生成ワーク **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) **モデルの保存先** diff --git a/ja/tutorials/flux/flux-2-klein.mdx b/ja/tutorials/flux/flux-2-klein.mdx index c1a373245..c12f18487 100644 --- a/ja/tutorials/flux/flux-2-klein.mdx +++ b/ja/tutorials/flux/flux-2-klein.mdx @@ -45,19 +45,19 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 ## Flux.2 Klein 4B モデルのダウンロード - + 4B モデル用のテキストエンコーダーです。 - + 拡散モデル（4B Base 版）。 - + 拡散モデル（4B 蒸留版）。 - + 4B モデル用のVAEです。 @@ -103,11 +103,11 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 拡散モデル（9B 蒸留版）。 - + 9B モデル用のテキストエンコーダーです。 - + 9B モデル用のVAEです。 diff --git a/ja/tutorials/flux/flux1-krea-dev.mdx b/ja/tutorials/flux/flux1-krea-dev.mdx index 7ba34a270..827438849 100644 --- a/ja/tutorials/flux/flux1-krea-dev.mdx +++ b/ja/tutorials/flux/flux1-krea-dev.mdx @@ -31,14 +31,13 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下の画像または JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 ![Flux Krea Dev ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - - - Comfy Cloud でこのワークフローを実行 - - - JSON をダウンロードするか、テンプレート ライブラリで「Flux.1 Krea Dev」を検索してください - - + +

JSON ワークフローをダウンロード

+ + + +

Comfy Cloud で実行

+ #### 2. 手動によるモデルのインストール @@ -51,7 +50,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' より高品質な出力を求め、かつ十分な VRAM をお持ちの場合、オリジナルの重みファイルも試すことができます： -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) の利用規約に同意する必要があります。 @@ -60,12 +59,12 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以前に Flux 関連のワークフローをご利用済みの場合、以下のモデルは既に存在するため、再ダウンロードの必要はありません。 **テキストエンコーダー** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true): VRAM が 32GB を超える環境で推奨 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors): 低 VRAM 環境向け +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) — VRAM が 32GB を超える環境で推奨 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) — 低 VRAM 環境向け **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) ファイルの保存先： ``` diff --git a/ja/tutorials/image/anima/anima.mdx b/ja/tutorials/image/anima/anima.mdx index e10df10db..e78fb3dca 100644 --- a/ja/tutorials/image/anima/anima.mdx +++ b/ja/tutorials/image/anima/anima.mdx @@ -82,15 +82,15 @@ Anima は 2 つのワークフローを提供しています——標準的な すべてのモデルファイルは Hugging Face の [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima) から入手できます。 - + Anima Base v1 用拡散モデル（2B）。 - + 両方のワークフローで共有されるテキストエンコーダー（Qwen-3 0.6B）。 - + 両方のワークフローで共有される VAE。 @@ -111,7 +111,7 @@ Anima は 2 つのワークフローを提供しています——標準的な Preview ワークフローを使用する場合は、代わりに以下のプレビュー拡散モデルをダウンロードしてください： - + Anima Preview 用拡散モデル（2B）。 diff --git a/ja/tutorials/image/boogu/boogu-image-0.1.mdx b/ja/tutorials/image/boogu/boogu-image-0.1.mdx index b15572a40..7b75c26ed 100644 --- a/ja/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ja/tutorials/image/boogu/boogu-image-0.1.mdx @@ -50,19 +50,19 @@ Boogu-Image-0.1-Turbo ワークフローは、拡散、テキストエンコー ### Boogu-Image-0.1-Turbo モデルのダウンロード - + Boogu-Image-0.1-Turbo 用の拡散モデル。 - + Boogu-Image-0.1-Turbo 用のテキストエンコーダー。 - + Boogu-Image-0.1-Turbo 用の VAE。 - + Boogu-Image-0.1-Turbo 用の LoRA モジュール (rank-128)。 @@ -100,15 +100,15 @@ Boogu-Image-0.1-Turbo ワークフローは、拡散、テキストエンコー ### Boogu-Image-0.1-Edit モデルのダウンロード - + Boogu-Image-0.1-Edit 用の拡散モデル。 - + Boogu-Image-0.1-Edit 用のテキストエンコーダー。 - + Boogu-Image-0.1-Edit 用の VAE。 diff --git a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index b4d4f2486..7b1f7a47c 100644 --- a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -43,17 +43,17 @@ Hugging Face: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos **Diffusion モデル** -- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_t2i.safetensors) +- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors) その他の重みファイルについては、[Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) からダウンロードしてください。 **テキストエンコーダー** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) ファイルの保存場所 diff --git a/ja/tutorials/image/ernie-image/ernie-image.mdx b/ja/tutorials/image/ernie-image/ernie-image.mdx index 023097979..b02fa0f45 100644 --- a/ja/tutorials/image/ernie-image/ernie-image.mdx +++ b/ja/tutorials/image/ernie-image/ernie-image.mdx @@ -55,19 +55,19 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' リパッケージされたすべてのモデルファイルは、Hugging Face の [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) で入手できます。 - + ERNIE-Image 用拡散モデル。 - + ERNIE-Image 用テキストエンコーダー。 - + ERNIE-Image 用プロンプトエンハンサーテキストエンコーダー。 - + ERNIE-Image 用 VAE。 @@ -99,19 +99,19 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### ERNIE-Image-Turbo モデルのダウンロード - + ERNIE-Image-Turbo 用拡散モデル。 - + ERNIE-Image-Turbo 用テキストエンコーダー。 - + ERNIE-Image-Turbo 用プロンプトエンハンサーテキストエンコーダー。 - + ERNIE-Image-Turbo 用 VAE。 diff --git a/ja/tutorials/image/hidream/hidream-e1.mdx b/ja/tutorials/image/hidream/hidream-e1.mdx index 3707adfc3..47798c64a 100644 --- a/ja/tutorials/image/hidream/hidream-e1.mdx +++ b/ja/tutorials/image/hidream/hidream-e1.mdx @@ -40,8 +40,8 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン **Diffusion モデル** 両方のモデルを同時にダウンロードする必要はありません。E1.1 は E1 をベースとした改良版であり、実際のテスト結果から、品質およびパフォーマンスの両面で E1 を大幅に上回ることが確認されています。 -- [hidream_e1_1_bf16.safetensors（推奨）](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors（推奨）](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **テキストエンコーダー**： @@ -74,15 +74,6 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン ## HiDream E1.1 の ComfyUI ネイティブ ワークフローの例 - - - Comfy Cloud で開く - - - JSON をダウンロード、またはテンプレートライブラリで "HiDream E1.1" を検索 - - - E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メガピクセル解像度** をサポートしています。ワークフローでは `Scale Image to Total Pixels` ノードを用いて、入力画像を自動的に 100万ピクセルにスケーリングします。 @@ -127,14 +118,9 @@ E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メ ## HiDream E1 の ComfyUI ネイティブ ワークフローの例 - - - Comfy Cloud で開く - - - JSON をダウンロード、またはテンプレートライブラリで "HiDream E1 Full" を検索 - - + +

Comfy Cloud で実行

+ E1 は 2025年4月28日にリリースされたモデルで、**768×768 の固定解像度のみ** をサポートします。 diff --git a/ja/tutorials/image/hidream/hidream-i1.mdx b/ja/tutorials/image/hidream/hidream-i1.mdx index ed3432e68..2e7c76364 100644 --- a/ja/tutorials/image/hidream/hidream-i1.mdx +++ b/ja/tutorials/image/hidream/hidream-i1.mdx @@ -101,21 +101,16 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Full バージョンのワークフロー - - - 設定不要ですぐに Comfy Cloud で実行 - - - ワークフローの JSON ファイルをダウンロード - - + +

Comfy Cloud で実行

+ #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード @@ -144,21 +139,16 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Dev バージョンのワークフロー - - - 設定不要ですぐに Comfy Cloud で実行 - - - ワークフローの JSON ファイルをダウンロード - - + +

Comfy Cloud で実行

+ #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード @@ -187,21 +177,16 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Fast バージョンのワークフロー - - - 設定不要ですぐに Comfy Cloud で実行 - - - ワークフローの JSON ファイルをダウンロード - - + +

Comfy Cloud で実行

+ #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード diff --git a/ja/tutorials/image/hidream/hidream-o1.mdx b/ja/tutorials/image/hidream/hidream-o1.mdx index 875b0d90c..6295428f3 100644 --- a/ja/tutorials/image/hidream/hidream-o1.mdx +++ b/ja/tutorials/image/hidream/hidream-o1.mdx @@ -51,19 +51,19 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` **チェックポイント** — 再パッケージおよび量子化済み。すべてのバージョンで最悪の外れ値は bf16 で保持され、未使用の deepstack 層は削除されています： -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量子化版 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 フル精度版（最大ファイル） +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量子化版 +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 フル精度版（最大ファイル） **テキストエンコーダ（プロンプト補強）** — 全バージョン共通： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA（オプション）** — Dev 蒸留は LoRA として Full モデルにも適用でき、蒸留強度を調整できます（[Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — フルランク -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — プルーニング版 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 代替 Checkpoint ベースの蒸留 +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — フルランク +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — プルーニング版 +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 代替 Checkpoint ベースの蒸留 ``` 📂 ComfyUI/ @@ -107,13 +107,13 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` **チェックポイント（Dev）** — 再パッケージおよび量子化済み。すべてのバージョンで最悪の外れ値は bf16 で保持され、未使用の deepstack 層は削除されています： -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量子化版 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 フル精度版（最大ファイル） +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量子化版 +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 フル精度版（最大ファイル） **テキストエンコーダ（プロンプト補強）** — 全バージョン共通： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/ja/tutorials/image/ideogram/ideogram-v4.mdx b/ja/tutorials/image/ideogram/ideogram-v4.mdx index 6f2a5ec82..2c44d5ea6 100644 --- a/ja/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ja/tutorials/image/ideogram/ideogram-v4.mdx @@ -46,23 +46,23 @@ Ideogram 4.0 は、Ideogram がオープンソースモデルとして公開し Hugging Face の [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) ですべての再パッケージ化されたモデルファイルを見つけることができます。 - + Ideogram 4.0 拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 - + Ideogram 4.0 条件なし拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 - + Ideogram 4.0 テキストエンコーダー（~8 GB）。models/text_encoders/ に配置 - + Ideogram 4.0 テキストエンコーダー（~2 GB）。models/text_encoders/ に配置 - + Ideogram 4.0 VAE（~335 MB）。models/vae/ に配置 diff --git a/ja/tutorials/image/krea/krea-2.mdx b/ja/tutorials/image/krea/krea-2.mdx index 88aa59318..32a6f8bd9 100644 --- a/ja/tutorials/image/krea/krea-2.mdx +++ b/ja/tutorials/image/krea/krea-2.mdx @@ -108,13 +108,13 @@ Krea 2向けのスタイルLoRAコレクションもKreaから公開されてい ローカルで使用するには、[Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2) からComfyUI最適化モデルファイルをダウンロードしてください。 - + krea2_turbo_fp8_scaled.safetensors: Turbo FP8 (推奨、ほとんどのユーザー向け) - + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B テキストエンコーダー - + qwen_image_vae.safetensors diff --git a/ja/tutorials/image/lens/lens.mdx b/ja/tutorials/image/lens/lens.mdx index 90dea6b56..edd441dec 100644 --- a/ja/tutorials/image/lens/lens.mdx +++ b/ja/tutorials/image/lens/lens.mdx @@ -94,19 +94,19 @@ Lens Turbo は蒸留版で、より少ないサンプリングステップで画 すべてのモデルファイルは Hugging Face の [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens) にあります。 - + Lens 用拡散モデル (BF16) - + Lens Turbo 用拡散モデル (BF16) - + Lens と Lens Turbo で共通のテキストエンコーダー (GPT-OSS-20B) - + Lens と Lens Turbo で共通の VAE (FLUX.2) diff --git a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index d607b2bc5..02b2594ef 100644 --- a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -30,14 +30,13 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## NewBie-image 文生成画像ワークフロー -| -| -| JSONをダウンロードするか、テンプレートライブラリで「NewBie-image」を検索してください -| -| -| クラウドで開く -| -| + +

JSON ワークフローファイルをダウンロード

+ + + +

ComfyUI Cloud で実行

+ @@ -45,16 +44,16 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/omnigen/omnigen2.mdx b/ja/tutorials/image/omnigen/omnigen2.mdx index b0c305f56..f9c0a2d72 100644 --- a/ja/tutorials/image/omnigen/omnigen2.mdx +++ b/ja/tutorials/image/omnigen/omnigen2.mdx @@ -1,6 +1,6 @@ --- title: "ComfyUI OmniGen2 ネイティブワークフローの例" -description: "ComfyUI OmniGen2 ネイティブワークフローの例 - 文字から画像生成、画像編集、および複数画像の合成を統合したモデル。" +description: "ComfyUI OmniGen2 ネイティブワークフローの例 — 文字から画像生成、画像編集、および複数画像の合成を統合したモデル。" sidebarTitle: "OmniGen2" translationSourceHash: b061ff8c translationFrom: tutorials/image/omnigen/omnigen2.mdx @@ -42,13 +42,13 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 本記事では複数のワークフローを取り扱うため、対応するモデルファイルおよびインストール先は以下の通りです。各ワークフロー内にも、該当するモデルファイルのダウンロード情報が記載されています。 **拡散モデル（Diffusion Models）** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) **テキストエンコーダー（Text Encoders）** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) ファイル保存先： ``` @@ -66,11 +66,9 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 ### 1. ワークフローファイルのダウンロード - - - Comfy Cloud で実行 - - + +

Comfy Cloud で実行

+ ![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -96,11 +94,9 @@ OmniGen2 は豊富な画像編集機能を備えており、画像へのテキ ### 1. ワークフローファイルのダウンロード - - - Comfy Cloud で実行 - - + +

Comfy Cloud で実行

+ ![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/ja/tutorials/image/ovis/ovis-image.mdx b/ja/tutorials/image/ovis/ovis-image.mdx index 78f8b80e1..ecbad6885 100644 --- a/ja/tutorials/image/ovis/ovis-image.mdx +++ b/ja/tutorials/image/ovis/ovis-image.mdx @@ -22,14 +22,13 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Ovis-Image のテキストから画像を生成するワークフロー - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで「Ovis image」を検索してください - - + +

JSONワークフローファイルをダウンロード

+ + + +

ComfyUI Cloud 上で実行

+ @@ -37,15 +36,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders（テキストエンコーダ）** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models（拡散モデル）** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/pixeldit/pixeldit.mdx b/ja/tutorials/image/pixeldit/pixeldit.mdx index 01869d6de..4cc67f103 100644 --- a/ja/tutorials/image/pixeldit/pixeldit.mdx +++ b/ja/tutorials/image/pixeldit/pixeldit.mdx @@ -62,11 +62,11 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' PixelDiT はテキストエンコーダーと拡散モデルの 2 つのモデルファイルを使用します。 - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT テキストエンコーダー - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 拡散モデル diff --git a/ja/tutorials/image/qwen/qwen-image-2512.mdx b/ja/tutorials/image/qwen/qwen-image-2512.mdx index 1c5c6f755..a144f0632 100644 --- a/ja/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ja/tutorials/image/qwen/qwen-image-2512.mdx @@ -43,14 +43,9 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - - + Comfy Cloud で実行 - - - JSON をダウンロードするか、テンプレートライブラリで "Qwen-Image-2512" を検索してください - - + ### 1. ワークフローファイル @@ -60,25 +55,28 @@ ComfyUI を更新した後、テンプレートからワークフローファイ - **Text to Image (Qwen-Image 2512)**: 標準的な50ステップ生成 - **Text to Image (Qwen-Image 2512 4steps)**: Lightning LoRA を用いた高速4ステップ生成 + +

JSON ワークフローをダウンロード

+ ### 2. モデルのダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（任意：4ステップ Lightning 加速用）** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **拡散モデル** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（ほとんどのユーザーに推奨） -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（VRAM が十分に確保でき、より高品質な出力を求めている場合） +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（ほとんどのユーザーに推奨） +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（VRAM が十分に確保でき、より高品質な出力を求めている場合） **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx index 7d1768c29..946f5c718 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -33,32 +33,31 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ComfyUI を更新した後、テンプレートからワークフローファイルを取得できます。あるいは、以下のワークフローを ComfyUI にドラッグ＆ドロップして読み込むこともできます。 - - - ComfyUI Cloud 上で実行 - - - JSON 形式ワークフローをダウンロード - - + +

JSON 形式ワークフローをダウンロード

+ + + +

ComfyUI Cloud 上で実行

+ ### 2. モデルのダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（任意：4 ステップ Lightning 加速用）** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **拡散モデル** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image-edit.mdx b/ja/tutorials/image/qwen/qwen-image-edit.mdx index bd7f93969..01cbdb893 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit.mdx @@ -47,14 +47,13 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ComfyUI を更新後、テンプレートからワークフローファイルを取得するか、下記のワークフローを ComfyUI へドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - - - JSON形式ワークフローをダウンロードするか、テンプレートライブラリで"image_qwen_image_edit"を検索してください - - - ComfyUI Cloud 上でこのワークフローを実行（ゼロセットアップ） - - + +

JSON形式ワークフローをダウンロード

+ + + +

ComfyUI Cloud 上で実行

+ 以下の画像を入力画像としてダウンロードしてください。 ![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -65,19 +64,19 @@ ComfyUI を更新後、テンプレートからワークフローファイルを **Diffusion モデル** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **テキストエンコーダ** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) モデルの保存場所 diff --git a/ja/tutorials/image/qwen/qwen-image-layered.mdx b/ja/tutorials/image/qwen/qwen-image-layered.mdx index 28749f22e..b2c05499a 100644 --- a/ja/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ja/tutorials/image/qwen/qwen-image-layered.mdx @@ -30,15 +30,13 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Qwen-Image-Layered ワークフロー -| -| -| JSON ワークフローファイルをダウンロード -| -| -| -| ComfyUI Cloud で実行 -| -| + +

JSON ワークフローファイルをダウンロード

+ + + +

ComfyUI Cloud で実行

+ @@ -46,15 +44,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) **モデルの保存場所** @@ -89,4 +87,4 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### プロンプト（任意） -テキストプロンプトは、入力画像全体の内容を記述することを目的としています。たとえば、前景オブジェクトの後ろに隠れている文字など、部分的に遮蔽されている要素も含めて指定可能です。ただし、個々のレイヤーの意味的内容を明示的に制御するためのものではありません。 +テキストプロンプトは、入力画像全体の内容を記述することを目的としています——たとえば、前景オブジェクトの後ろに隠れている文字など、部分的に遮蔽されている要素も含めて指定可能です。ただし、個々のレイヤーの意味的内容を明示的に制御するためのものではありません。 diff --git a/ja/tutorials/image/qwen/qwen-image.mdx b/ja/tutorials/image/qwen/qwen-image.mdx index bb4f04855..7fb6475cc 100644 --- a/ja/tutorials/image/qwen/qwen-image.mdx +++ b/ja/tutorials/image/qwen/qwen-image.mdx @@ -58,12 +58,9 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - - - - - - + + Comfy Cloudで実行 + 本ドキュメントに添付されたワークフローでは、以下の3種類の異なるモデルが使用されています： 1. Qwen-Imageオリジナルモデル（fp8_e4m3fn） @@ -85,11 +82,14 @@ GPU：RTX4090D（24GB） ComfyUIを更新後、テンプレートからワークフローファイルを検索するか、以下のワークフローをComfyUIにドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像へ変換するワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - + +

Qwen-Image公式モデル用ワークフロー（JSON形式）をダウンロード

+ 蒸留版 - - + +

蒸留モデル用ワークフロー（JSON形式）をダウンロード

+ ### 2. モデルのダウンロード @@ -103,12 +103,12 @@ ComfyUIを更新後、テンプレートからワークフローファイルを **拡散モデル** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill（蒸留版） -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 蒸留版のオリジナル作者は、CFG値1.0で15ステップでの使用を推奨しています。 @@ -117,15 +117,15 @@ Qwen_image_distill（蒸留版） **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** @@ -164,19 +164,18 @@ Qwen_image_distill（蒸留版） これはControlNetモデルであるため、通常のControlNetとして使用できます。 - - - - - - + + Comfy Cloudで実行 + ### 1. ワークフローおよび入力画像 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - + +

JSON形式ワークフローをダウンロード

+ 以下の画像を入力としてダウンロードしてください ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -184,7 +183,7 @@ Qwen_image_distill（蒸留版） 1. InstantX ControlNet -[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)をダウンロードし、`ComfyUI/models/controlnet/`フォルダーに保存してください +[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)をダウンロードし、`ComfyUI/models/controlnet/`フォルダーに保存してください 2. **Lotus Depthモデル** @@ -192,11 +191,11 @@ Qwen_image_distill（蒸留版） **拡散モデル** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) **VAEモデル** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) または任意のSD1.5互換VAE +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) または任意のSD1.5互換VAE ``` ComfyUI/ @@ -220,12 +219,9 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNetsモデルパッチワークフロー - - - - - - + + Comfy Cloudで実行 + このモデルは実際にはControlNetではなく、Canny、Depth、Inpaintの3種類の異なる制御モードをサポートする「モデルパッチ」です。 @@ -238,7 +234,9 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして対応するワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - + +

JSON形式ワークフローをダウンロード

+ 以下の画像を入力としてダウンロードしてください： @@ -248,9 +246,9 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http その他のモデルはQwen-Image基本ワークフローと同一です。以下のモデルのみをダウンロードし、`ComfyUI/models/model_patches`フォルダーに保存してください。 -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. ワークフローの使用方法 @@ -292,12 +290,9 @@ Inpaintモデルでは、[マスクエディター](/ja/interface/maskeditor)を ## Qwen Image Union ControlNet LoRAワークフロー - - - - - - + + Comfy Cloudで実行 + オリジナルモデルのURL：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org再ホストURL：[qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors)：Canny、Depth、Pose、Lineart、Softedge、Normal、Openposeをサポートする画像構造制御用LoRA @@ -306,7 +301,9 @@ Comfy Org再ホストURL：[qwen_image_union_diffsynth_lora.safetensors](https:/ 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - + +

JSON形式ワークフローをダウンロード

+ 以下の画像を入力としてダウンロードしてください diff --git a/ja/tutorials/image/z-image/z-image-turbo.mdx b/ja/tutorials/image/z-image/z-image-turbo.mdx index e77bfd842..fe7d24df0 100644 --- a/ja/tutorials/image/z-image/z-image-turbo.mdx +++ b/ja/tutorials/image/z-image/z-image-turbo.mdx @@ -46,15 +46,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### Z-Image-Turbo モデルのダウンロード - + Z-Image-Turbo 専用のテキストエンコーダーです。 - + Z-Image-Turbo 専用の拡散モデルです。 - + Z-Image-Turbo 専用のVAE（変分オートエンコーダー）です。 @@ -81,7 +81,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### ControlNet 用の追加モデル - + Z-Image-Turbo 専用のControlNetモデルパッチです。 diff --git a/ja/tutorials/image/z-image/z-image.mdx b/ja/tutorials/image/z-image/z-image.mdx index a25bc4c77..733137ed1 100644 --- a/ja/tutorials/image/z-image/z-image.mdx +++ b/ja/tutorials/image/z-image/z-image.mdx @@ -37,15 +37,15 @@ Z-Image（Base）は、コミュニティ主導のファインチューニング ## Z-Image モデルのダウンロード - + Z-Image 用テキストエンコーダー。 - + Z-Image 用拡散モデル。 - + Z-Image 用 VAE。 diff --git a/ja/tutorials/llm/gemma4/gemma4.mdx b/ja/tutorials/llm/gemma4/gemma4.mdx index f48575928..0234a4b0e 100644 --- a/ja/tutorials/llm/gemma4/gemma4.mdx +++ b/ja/tutorials/llm/gemma4/gemma4.mdx @@ -74,11 +74,11 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' Gemma 4 モデルは ComfyUI ではテキストエンコーダー（text encoder）として読み込まれます。該当するモデルファイルをダウンロードし、正しいディレクトリに配置してください： - + 高速・軽量、コンシューマー GPU に最適。 - + バランスの取れた性能。ワークフローのデフォルトモデル。 diff --git a/ja/tutorials/llm/qwen/qwen3.mdx b/ja/tutorials/llm/qwen/qwen3.mdx index 874bbebbb..0a2e148a1 100644 --- a/ja/tutorials/llm/qwen/qwen3.mdx +++ b/ja/tutorials/llm/qwen/qwen3.mdx @@ -71,15 +71,15 @@ Qwen 3.0 は、ComfyUI ワークフロー内で構造化テキスト生成とイ Qwen 3.0 モデルは ComfyUI でテキストエンコーダーとして読み込まれます。モデルファイルは Qwen3.5 と共有されています。ハードウェアに合わせて適切なバージョンを選択してください： - + 軽量版、約 4.5 GB。低 VRAM 環境や高速ダウンロードに最適。 - + サイズと品質のバランス型。ほとんどのコンシューマー GPU に推奨。 - + 最大版、約 19 GB。より高品質な出力、より多くの VRAM が必要。 diff --git a/ja/tutorials/llm/qwen/qwen3_5.mdx b/ja/tutorials/llm/qwen/qwen3_5.mdx index 12f65e4c5..0a8733d93 100644 --- a/ja/tutorials/llm/qwen/qwen3_5.mdx +++ b/ja/tutorials/llm/qwen/qwen3_5.mdx @@ -73,15 +73,15 @@ Qwen3.5 は、視覚的理解とテキスト生成の組み合わせが ComfyUI Qwen3.5 モデルは ComfyUI でテキストエンコーダーとして読み込まれます。ハードウェアに合わせて適切なバージョンを選択してください： - + 軽量版、約 4.5 GB。低 VRAM 環境や高速ダウンロードに最適。 - + サイズと品質のバランス型。ほとんどのコンシューマー GPU に推奨。 - + 最大版、約 19 GB。より高品質な出力、より多くの VRAM が必要。 diff --git a/ja/tutorials/partner-nodes/google/gemini.mdx b/ja/tutorials/partner-nodes/google/gemini.mdx index ab669a618..f1a6c2ec1 100644 --- a/ja/tutorials/partner-nodes/google/gemini.mdx +++ b/ja/tutorials/partner-nodes/google/gemini.mdx @@ -22,11 +22,13 @@ Google Gemini は、Google が開発した強力な AI モデルであり、対 以下の JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - - - JSON 形式のワークフローファイルをダウンロード - - + +

JSON 形式のワークフローファイルをダウンロード

+ ### 2. ワークフロー実行の手順 diff --git a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx index c2217ede4..ef05651b6 100644 --- a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -51,8 +51,9 @@ Kling 2.6 Motion Control は、快手（Kuaishou）社が開発した専用の ## Kling 2.6 Motion Control のワークフロー - - + +

JSON形式のワークフローファイルをダウンロード

+ ## 入力要件 diff --git a/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 5c3863523..951239804 100644 --- a/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -54,11 +54,9 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_text_to_video.mp4" > - - - JSON 形式のワークフロー ファイルをダウンロード - - + +

JSON 形式のワークフロー ファイルをダウンロード

+ ### 2. ワークフロー実行手順 @@ -82,11 +80,9 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_image_to_video.mp4" > - - - JSON 形式のワークフロー ファイルをダウンロード - - + +

JSON 形式のワークフロー ファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください。 @@ -116,11 +112,9 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video.mp4" > - - - JSON 形式のワークフロー ファイルをダウンロード - - + +

JSON 形式のワークフロー ファイルをダウンロード

+ 以下の動画を入力動画としてダウンロードしてください： diff --git a/ja/tutorials/partner-nodes/openai/chat.mdx b/ja/tutorials/partner-nodes/openai/chat.mdx index 12400d75f..2d195caf5 100644 --- a/ja/tutorials/partner-nodes/openai/chat.mdx +++ b/ja/tutorials/partner-nodes/openai/chat.mdx @@ -22,9 +22,13 @@ OpenAI は生成AIに特化した企業であり、強力な対話機能を提 以下の JSON ファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 -| -| JSON 形式のワークフローファイルをダウンロード -| + +

JSON 形式のワークフローファイルをダウンロード

+ ### 2. ワークフロー実行手順（ステップごと） diff --git a/ja/tutorials/partner-nodes/rodin/model-generation.mdx b/ja/tutorials/partner-nodes/rodin/model-generation.mdx index 17009c0bb..99704e747 100644 --- a/ja/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ja/tutorials/partner-nodes/rodin/model-generation.mdx @@ -32,9 +32,13 @@ ComfyUI は現在、Rodin のモデル生成 API をネイティブ統合して 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - - 単一視点モデル生成 (JSON形式) - + +

JSON 形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください。 @@ -62,9 +66,13 @@ ComfyUI は現在、Rodin のモデル生成 API をネイティブ統合して 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - - 複数視点モデル生成 (JSON形式) - + +

JSON 形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください。 diff --git a/ja/tutorials/partner-nodes/runway/video-generation.mdx b/ja/tutorials/partner-nodes/runway/video-generation.mdx index 5befee90e..e85367f8e 100644 --- a/ja/tutorials/partner-nodes/runway/video-generation.mdx +++ b/ja/tutorials/partner-nodes/runway/video-generation.mdx @@ -37,7 +37,9 @@ Runway は、生成AIに特化した企業であり、強力な動画生成機 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen3a_turbo_image_to_video/runway_image_to_video_gen3a_turbo.mp4" > - + +

JSON形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください。 @@ -65,7 +67,9 @@ Runway は、生成AIに特化した企業であり、強力な動画生成機 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen4_turbo_image_to_video/runway_gen4_turo_image_to_video.mp4" > - + +

JSON形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください。 @@ -93,7 +97,9 @@ Runway は、生成AIに特化した企業であり、強力な動画生成機 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/first_last_frame_to_video/runway_first_last_frame.mp4" > - + +

JSON形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください。 diff --git a/ja/tutorials/partner-nodes/tripo/model-generation.mdx b/ja/tutorials/partner-nodes/tripo/model-generation.mdx index b71b44788..c9a3bb055 100644 --- a/ja/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ja/tutorials/partner-nodes/tripo/model-generation.mdx @@ -35,11 +35,9 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - - - テキストからモデル生成ワークフロー - - + +

JSON形式ワークフローファイルをダウンロード

+ ### 2. ワークフロー実行手順（ステップ・バイ・ステップ） @@ -60,11 +58,9 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - - - 画像からモデル生成ワークフロー - - + +

JSON形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください： @@ -89,11 +85,9 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - - - 複数視点からのモデル生成ワークフロー - - + +

JSON形式ワークフローファイルをダウンロード

+ 以下の画像を入力画像としてダウンロードしてください： diff --git a/ja/tutorials/utility/depth-anything-3.mdx b/ja/tutorials/utility/depth-anything-3.mdx index 36a32d873..d49627577 100644 --- a/ja/tutorials/utility/depth-anything-3.mdx +++ b/ja/tutorials/utility/depth-anything-3.mdx @@ -39,10 +39,10 @@ ComfyUI は Depth Anything 3 ノードをネイティブサポートしていま Depth Anything 3 チェックポイントをダウンロードし、対応する ComfyUI フォルダに保存します： -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — 軽量で高速な推論 -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — バランスの取れた性能 -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 単眼深度に最適（空検出対応） -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — メートル単位の物理深度（空検出対応） +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — 軽量で高速な推論 +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — バランスの取れた性能 +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 単眼深度に最適（空検出対応） +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — メートル単位の物理深度（空検出対応） ``` ComfyUI/ diff --git a/ja/tutorials/utility/face-detection/mediapipe.mdx b/ja/tutorials/utility/face-detection/mediapipe.mdx index 0743c3501..0044222f6 100644 --- a/ja/tutorials/utility/face-detection/mediapipe.mdx +++ b/ja/tutorials/utility/face-detection/mediapipe.mdx @@ -55,7 +55,7 @@ ComfyUI を最新バージョンにアップデートし、`Workflow` → `Brows MediaPipe Face Detection モデルは [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe) でホストされています。 -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) 以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/utility/moge.mdx b/ja/tutorials/utility/moge.mdx index 075dc59f1..e3e7397fa 100644 --- a/ja/tutorials/utility/moge.mdx +++ b/ja/tutorials/utility/moge.mdx @@ -50,8 +50,8 @@ ComfyUI は MoGe ノードをネイティブサポートしています。始め MoGe チェックポイントをダウンロードし、ComfyUI の該当フォルダに保存します： -- **MoGe-2（推奨）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1（ベースライン）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2（推奨）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1（ベースライン）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/utility/pose-detection-sdpose.mdx b/ja/tutorials/utility/pose-detection-sdpose.mdx index ba69183f9..a63be80a2 100644 --- a/ja/tutorials/utility/pose-detection-sdpose.mdx +++ b/ja/tutorials/utility/pose-detection-sdpose.mdx @@ -87,11 +87,11 @@ ComfyUIを最新バージョンにアップデートし、`Workflow` → `Browse SDPoseとRT-DETRv4のモデルチェックポイントは、[Comfy-Org SDPose モデルリポジトリ](https://huggingface.co/Comfy-Org/SDPose) で公開されています。 **checkpoints** (SDPoseモデル): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models** (RT-DETRv4検出器): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (推奨) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (完全精度、サイズ大) +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (推奨) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (完全精度、サイズ大) 以下のディレクトリ構成に配置してください: diff --git a/ja/tutorials/utility/remove-background-birefnet.mdx b/ja/tutorials/utility/remove-background-birefnet.mdx index 7f1295b3a..c0f4cc3ff 100644 --- a/ja/tutorials/utility/remove-background-birefnet.mdx +++ b/ja/tutorials/utility/remove-background-birefnet.mdx @@ -48,7 +48,7 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` BiRefNet モデルは [Comfy-Org BiRefNet モデルリポジトリ](https://huggingface.co/Comfy-Org/BiRefNet) でホストされています。 -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) 以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/utility/video-segment-sam3.mdx b/ja/tutorials/utility/video-segment-sam3.mdx index 3295028ed..6ac27d505 100644 --- a/ja/tutorials/utility/video-segment-sam3.mdx +++ b/ja/tutorials/utility/video-segment-sam3.mdx @@ -68,7 +68,7 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` SAM 3.1 モデルは [Comfy-Org SAM 3.1 モデルリポジトリ](https://huggingface.co/Comfy-Org/sam3.1) でホストされています。 -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) 以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/utility/void-video-inpainting.mdx b/ja/tutorials/utility/void-video-inpainting.mdx index a512579a3..57a3967e5 100644 --- a/ja/tutorials/utility/void-video-inpainting.mdx +++ b/ja/tutorials/utility/void-video-inpainting.mdx @@ -65,24 +65,24 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` **拡散モデル** — 中核となる2パス修復モデル： -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — 精錬パス、時間的安定性に優れる -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — 一次パス +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 精錬パス、時間的安定性に優れる +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 一次パス **VAE：** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) **オプティカルフロー：** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 セグメンテーション：** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **テキストエンコーダ：** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ diff --git a/ja/tutorials/video/bytedance/bernini-r.mdx b/ja/tutorials/video/bytedance/bernini-r.mdx index 6485b68d0..13d6b0383 100644 --- a/ja/tutorials/video/bytedance/bernini-r.mdx +++ b/ja/tutorials/video/bytedance/bernini-r.mdx @@ -49,16 +49,16 @@ ComfyUI は Bernini-R ノードをネイティブサポートしています。 必要なモデルウェイトをダウンロードし、対応する ComfyUI フォルダに保存します： **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 32d050845..5871cea65 100644 --- a/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -14,27 +14,17 @@ Cosmos-Predict2 は、NVIDIA によって開発された次世代の物理世界 Cosmos-Predict2 は、テキストから画像（Text2Image）や動画から世界へ（Video2World）など、さまざまな生成方法をサポートしており、産業シミュレーション、自動運転、都市計画、科学研究などの分野で広く使用されています。 これは、インテリジェントビジョンと物理世界の深い統合を促進するための重要な基礎ツールです。 - - - Cosmos-Predict2 ソースコードとドキュメント - - - Cosmos-Predict2 モデルコレクション - - +GitHub:[Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) +huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) このガイドでは、ComfyUI での **Video2World** 生成の完了までの手順を説明します。 テキストから画像のセクションについては、以下の部分を参照してください。 - - - Cosmos-Predict2 を使用したテキストから画像の生成 - - - 強力な GPU で Comfy Cloud 上で Cosmos-Predict2 ワークフローを実行 - - + + Cosmos-Predict2 を使用したテキストから画像の生成 + +{/* ## Cosmos Predict2 Video2World ワークフロー @@ -51,14 +41,9 @@ Cosmos-Predict2 は、テキストから画像（Text2Image）や動画から世 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - - - JSON 形式ワークフローファイルをダウンロード - - - Comfy Cloud でこのワークフローを実行（モデルプリインストール済み） - - + +

Json 形式ワークフローファイルをダウンロード

+ 入力として以下の画像をダウンロードしてください。 @@ -70,23 +55,17 @@ Cosmos-Predict2 は、テキストから画像（Text2Image）や動画から世 **Diffusion model** - - cosmos_predict2_2B_video2world_480p_16fps.safetensors - +- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) 他の重みについては、[Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) にアクセスしてダウンロードしてください **Text encoder** - - oldt5_xxl_fp8_e4m3fn_scaled.safetensors - +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** - - wan_2.1_vae.safetensors - +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) ファイル保存場所 ``` @@ -113,4 +92,4 @@ Cosmos-Predict2 は、テキストから画像（Text2Image）や動画から世 6. (オプション) `ClipTextEncode` ノードでプロンプトを変更できます 7. (オプション) `CosmosPredict2ImageToVideoLatent` ノードでサイズとフレーム数を変更してください 8. `Run` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行してください -9. 生成が完了すると、動画は自動的に `ComfyUI/output/` ディレクトリに保存されます。`save video` ノードでプレビューすることもできます \ No newline at end of file +9. 生成が完了すると、動画は自動的に `ComfyUI/output/` ディレクトリに保存されます。`save video` ノードでプレビューすることもできます */} \ No newline at end of file diff --git a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 7c6b936e5..5ca22c6eb 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -29,17 +29,17 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## モデルリンク **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **diffusion_models** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **vae** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) モデルの保存場所 diff --git a/ja/tutorials/video/hunyuan/hunyuan-video.mdx b/ja/tutorials/video/hunyuan/hunyuan-video.mdx index 77788802f..61bee272e 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video.mdx @@ -46,9 +46,9 @@ Hunyuan Video シリーズは [Tencent](https://huggingface.co/tencent) によ 以下のモデルは、テキストから動画および画像から動画の両方のワークフローで使用されます。ダウンロードして、指定されたディレクトリに保存してください： -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) +- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) 保存場所： @@ -73,7 +73,7 @@ Hunyuan Text-to-Video は 2024 年 12 月にオープンソース化され、中 ### 2. モデルの手動インストール -[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models` フォルダに保存してください。 +[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models` フォルダに保存してください。 これらのモデルファイルがすべて正しい場所に存在することを確認してください： @@ -126,7 +126,7 @@ Hunyuan Image-to-Video モデルは 2025 年 3 月 6 日にオープンソース ### v1 および v2 バージョンで共通のモデル 以下のファイルをダウンロードし、`ComfyUI/models/clip_vision` ディレクトリに保存してください： -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) +- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) ### V1"concat"画像から動画ワークフロー @@ -140,7 +140,7 @@ Hunyuan Image-to-Video モデルは 2025 年 3 月 6 日にオープンソース #### 2. 関連モデルの手動インストール -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) これらのモデルファイルがすべて正しい場所に存在することを確認してください： @@ -185,7 +185,7 @@ v2 ワークフローは本質的に v1 ワークフローと同じです。**re #### 2. 関連モデルの手動インストール -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) これらのモデルファイルがすべて正しい場所に存在することを確認してください： diff --git a/ja/tutorials/video/kandinsky/kandinsky-5.mdx b/ja/tutorials/video/kandinsky/kandinsky-5.mdx index a6eb738f9..bb14be791 100644 --- a/ja/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/ja/tutorials/video/kandinsky/kandinsky-5.mdx @@ -51,39 +51,21 @@ Kandinsky 5.0 は、Flow Matching を備えた潜在拡散パイプラインを ComfyUI を最新バージョンに更新し、メニュー `ワークフロー` -> `テンプレートを表示` -> `動画` から "Kandinsky 5.0 T2V" を見つけてワークフローを読み込んでください。 - - - T2V ワークフローをダウンロードしてローカルで使用 - - - Comfy Cloud で開く - - + +

JSON ワークフローファイルをダウンロード

+ ### 2. モデルの手動ダウンロード **テキストエンコーダー** - - - - Qwen2.5-VL 7B テキストエンコーダー (FP8) - - - CLIP-L テキストエンコーダー - - +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) **拡散モデル** - - - Kandinsky 5.0 T2V Lite SFT 拡散モデル (5s) - +- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) **VAE** - - - HunyuanVideo 3D VAE - +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) ``` ComfyUI/ @@ -103,39 +85,21 @@ ComfyUI/ ComfyUI を最新バージョンに更新し、メニュー `ワークフロー` -> `テンプレートを表示` -> `動画` から "Kandinsky 5.0 I2V" を見つけてワークフローを読み込んでください。 - - - I2V ワークフローをダウンロードしてローカルで使用 - - - Comfy Cloud で開く - - + +

JSON ワークフローファイルをダウンロード

+ ### 2. モデルの手動ダウンロード **テキストエンコーダー** - - - - Qwen2.5-VL 7B テキストエンコーダー (FP8) - - - CLIP-L テキストエンコーダー - - +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) **拡散モデル** - - - Kandinsky 5.0 I2V Lite 拡散モデル (5s) - +- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) **VAE** - - - HunyuanVideo 3D VAE - +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/ltxv.mdx b/ja/tutorials/video/ltxv.mdx index 74c84afe6..8d047bc2e 100644 --- a/ja/tutorials/video/ltxv.mdx +++ b/ja/tutorials/video/ltxv.mdx @@ -26,17 +26,9 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 最初の [フレーム画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png) を使用して動画を制御できます。 - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "LTX-Video" を検索 - - - このワークフローのサンプル入力画像を入手 - - + +

Comfy Cloud で実行

+ LTX-Video 画像から動画ワークフロー @@ -46,15 +38,6 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## テキストから動画 - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "LTX-Video" を検索 - - - LTX-Video テキストから動画ワークフロー @@ -65,13 +48,8 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 以下のモデルをダウンロードし、下記に指定された場所に配置してください： - - ダウンロードして ComfyUI/models/checkpoints/ に配置 - - - - ダウンロードして ComfyUI/models/text_encoders/ に配置 - +- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true) ``` ├── checkpoints/ diff --git a/ja/tutorials/video/wan/fun-camera.mdx b/ja/tutorials/video/wan/fun-camera.mdx index d4c4eecaa..adfb6689a 100644 --- a/ja/tutorials/video/wan/fun-camera.mdx +++ b/ja/tutorials/video/wan/fun-camera.mdx @@ -13,6 +13,7 @@ translationBlockHashes: "Performance Reference": 32425486 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Wan2.1 Fun Camera について @@ -35,49 +36,21 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下に示すすべてのモデルは、[Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) から入手できます。 -### Diffusion モデル - -1.3B または 14B のいずれかを選択: - - - - Wan2.1 Fun Camera 1.3B 拡散モデル - - - Wan2.1 Fun Camera 14B 拡散モデル - - +**Diffusion モデル**（1.3B または 14B のいずれかを選択）: +- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) +- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) 以前に Wan2.1 関連のモデルをご利用になったことがある場合、以下のモデルは既にご所有である可能性があります。万が一不足している場合は、それぞれダウンロードしてください。 -### Text Encoders - -いずれか1つを選択: - - - - フル精度テキストエンコーダ - - - FP8 量子化テキストエンコーダ（低 VRAM 推奨） - - - -### VAE +**Text Encoders**（いずれか1つを選択）: +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - - Wan2.1 VAE モデル - - +**VAE**: +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -### CLIP Vision - - - - CLIP ビジョンエンコーダ - - +**CLIP Vision**: +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) ファイルの保存場所: @@ -97,16 +70,9 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 1.3B ネイティブワークフロー例 -### 1. ワークフローのダウンロード +### 1. ワークフロー関連ファイルのダウンロード - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "Wan 2.1 Fun Camera 1.3B" を検索 - - +#### 1.1 ワークフローファイル 以下の動画をダウンロードし、ComfyUI にドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます： @@ -116,19 +82,21 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > + +

JSON ワークフローファイルをダウンロード

+ + 14B バージョンをご利用になりたい場合は、単にモデルファイルを 14B バージョンに置き換えてください。ただし、VRAM の要件にご注意ください。 -### 2. 入力素材のダウンロード +#### 1.2 入力画像のダウンロード + +以下の画像をダウンロードし、これを開始フレームとして使用します： - - - 以下の画像をダウンロードし、1.3B ワークフローの開始フレームとして使用します - - +![入力参照画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) -### 3. ワークフローをステップ・バイ・ステップで完了させる +### 2. ワークフローをステップ・バイ・ステップで完了させる ![Wan2.1 Fun Camera ワークフロー手順](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -145,30 +113,18 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 14B ワークフローおよび入力画像 -### 1. ワークフローのダウンロード - - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "Wan 2.1 Fun Camera 14B" を検索 - - - -### 2. 入力素材のダウンロード + +

JSON ワークフローファイルをダウンロード

+ - - - 以下の画像をダウンロードし、14B ワークフローの開始フレームとして使用します - - +**入力画像** +![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) ## パフォーマンスの参考値 diff --git a/ja/tutorials/video/wan/fun-control.mdx b/ja/tutorials/video/wan/fun-control.mdx index fe5402c18..c1001c086 100644 --- a/ja/tutorials/video/wan/fun-control.mdx +++ b/ja/tutorials/video/wan/fun-control.mdx @@ -55,18 +55,18 @@ ComfyUI は現在、Wan2.1 Fun Control モデルを**ネイティブサポート 対応するリンクをクリックしてダウンロードしてください。以前に Wan 関連のワークフローを使用したことがある場合は、**Diffusion models** のみをダウンロードする必要があります。 **Diffusion models** - 1.3B または 14B を選択。14B バージョンはファイルサイズが大きく（32GB）、VRAM 要件も高くなります： -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true)：ダウンロード後に `Wan2.1-Fun-14B-Control.safetensors` にリネームしてください **Text encoders** - 以下のモデルのいずれかを選択（fp16 精度はサイズが大きく、パフォーマンス要件が高くなります）： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存場所： ``` diff --git a/ja/tutorials/video/wan/fun-inp.mdx b/ja/tutorials/video/wan/fun-inp.mdx index 181f518ce..ccae9db7b 100644 --- a/ja/tutorials/video/wan/fun-inp.mdx +++ b/ja/tutorials/video/wan/fun-inp.mdx @@ -53,18 +53,18 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 以下のモデルは [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) と [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) で見つかります。 **Diffusion models** - 1.3B または 14B を選択してください。14B バージョンはファイルサイズが大きく (32GB)、VRAM 要件も高くなります： -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true): ダウンロード後、`Wan2.1-Fun-14B-InP.safetensors` にリネームしてください +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): ダウンロード後、`Wan2.1-Fun-14B-InP.safetensors` にリネームしてください **Text encoders** - 以下のモデルのいずれかを選択してください（fp16 精度はサイズが大きく、パフォーマンス要件も高くなります）： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存場所： ``` diff --git a/ja/tutorials/video/wan/vace.mdx b/ja/tutorials/video/wan/vace.mdx index e7127c367..67687320f 100644 --- a/ja/tutorials/video/wan/vace.mdx +++ b/ja/tutorials/video/wan/vace.mdx @@ -61,18 +61,18 @@ VACE 14B は、アリババ Tongyi Wanxiang チームが公開したオープン ### モデルのダウンロード **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 以前に Wan Video 関連のワークフローをご利用になったことがある場合、以下のモデルファイルはすでにダウンロード済みです。 **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **Text encoders** からいずれか 1 つのバージョンを選択してダウンロードしてください： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) ファイルの保存先： ``` diff --git a/ja/tutorials/video/wan/wan-ati.mdx b/ja/tutorials/video/wan/wan-ati.mdx index 87176883f..d239223b3 100644 --- a/ja/tutorials/video/wan/wan-ati.mdx +++ b/ja/tutorials/video/wan/wan-ati.mdx @@ -46,17 +46,17 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ワークフローからモデルファイルを正常にダウンロードできていない場合、以下のリンクから手動でダウンロードしてみてください。 **Diffusionモデル** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **テキストエンコーダー**（以下のいずれか1つを選択） -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) ファイル保存先 ``` diff --git a/ja/tutorials/video/wan/wan-causal-forcing.mdx b/ja/tutorials/video/wan/wan-causal-forcing.mdx index 66fbf617d..4b1420be4 100644 --- a/ja/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ja/tutorials/video/wan/wan-causal-forcing.mdx @@ -83,10 +83,10 @@ Wan2.1 I2V モデルと必要なファイルをダウンロードします。対 ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B チェックポイント - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B チェックポイント（最小 8GB VRAM） @@ -94,10 +94,10 @@ Wan2.1 I2V モデルと必要なファイルをダウンロードします。対 ### CLIP と VAE - + google-bert/bert-base-uncased — CLIP テキストエンコーダー - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/ja/tutorials/video/wan/wan-dancer.mdx b/ja/tutorials/video/wan/wan-dancer.mdx index 8a6fb2fdc..6ea08b40f 100644 --- a/ja/tutorials/video/wan/wan-dancer.mdx +++ b/ja/tutorials/video/wan/wan-dancer.mdx @@ -61,20 +61,20 @@ ComfyUI を最新バージョンに更新し、ワークフローファイルを ### 3. 手動でのモデルのダウンロード **拡散モデル** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **テキストエンコーダ** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP ビジョン** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan-flf.mdx b/ja/tutorials/video/wan/wan-flf.mdx index 184c61a11..7e4a45626 100644 --- a/ja/tutorials/video/wan/wan-flf.mdx +++ b/ja/tutorials/video/wan/wan-flf.mdx @@ -56,7 +56,7 @@ Wan FLF2V（First-Last Frame Video Generation：始終フレーム動画生成 本ガイドで使用するすべてのモデルは、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)から入手できます。 **diffusion_models**：ご使用のハードウェア環境に応じて、以下のいずれかのバージョンを選択してください。 -- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8: [wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -64,14 +64,14 @@ Wan FLF2V（First-Last Frame Video Generation：始終フレーム動画生成 **Text encoders**：以下のいずれか1つのバージョンをダウンロードしてください。 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存先 diff --git a/ja/tutorials/video/wan/wan-move.mdx b/ja/tutorials/video/wan/wan-move.mdx index 8b950e59d..02d87136c 100644 --- a/ja/tutorials/video/wan/wan-move.mdx +++ b/ja/tutorials/video/wan/wan-move.mdx @@ -28,37 +28,37 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Wan-Move 画像から動画へのワークフロー - - ワークフローをダウンロード - + +

JSON ワークフローファイルをダウンロード

+ - - クラウドで開く - + +

ComfyUI Cloud で実行

+ ## モデルリンク - - **text_encoders** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors - +**text_encoders** - - **clip_vision** -- clip_vision_h.safetensors - +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - **loras** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors - +**clip_vision** - - **diffusion_models** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors - +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) - - **vae** -- wan_2.1_vae.safetensors - +**loras** + +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) + +**diffusion_models** + +- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) + +**vae** + +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **モデル保存場所** diff --git a/ja/tutorials/video/wan/wan-video.mdx b/ja/tutorials/video/wan/wan-video.mdx index a4cc007ed..73df834b7 100644 --- a/ja/tutorials/video/wan/wan-video.mdx +++ b/ja/tutorials/video/wan/wan-video.mdx @@ -41,14 +41,14 @@ Wan2.1 Video シリーズは、アリババ社が 2025 年 2 月に [Apache 2.0 このガイドで言及されるすべてのモデルは、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files) から入手できます。以下は、このガイドのサンプルで使用する共通のモデルであり、事前にダウンロードしておくことを推奨します： **Text encoders** からいずれか 1 つのバージョンを選択してダウンロードしてください： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイルの保存先ディレクトリ構成： ``` @@ -70,7 +70,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル ## Wan2.1 テキスト→動画（T2V）ワークフロー -ワークフローを開始する前に、[wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +ワークフローを開始する前に、[wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 > 他の T2V 精度バージョンが必要な場合は、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) からダウンロードしてください。 @@ -107,7 +107,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル ![Wan2.1 画像→動画ワークフロー（14B、480P）の入力画像サンプル](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) #### 2. モデルのダウンロード -[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 #### 3. ワークフローをステップごとに実行 @@ -135,7 +135,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル #### 2. モデルのダウンロード -[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 #### 3. ワークフローをステップごとに実行 diff --git a/ja/tutorials/video/wan/wan2-2-animate.mdx b/ja/tutorials/video/wan/wan2-2-animate.mdx index 410410b6e..314d5e51b 100644 --- a/ja/tutorials/video/wan/wan2-2-animate.mdx +++ b/ja/tutorials/video/wan/wan2-2-animate.mdx @@ -55,14 +55,13 @@ Wan-Animate は、WAN チームが開発した人物アニメーションおよ 以下のワークフローファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして読み込んでください。 - - - Comfy Cloud で実行 - - - JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Animate" を検索 - - + +

JSON ワークフローをダウンロード

+ + + +

Comfy Cloud で実行

+ 以下の素材を入力としてダウンロードしてください： @@ -78,40 +77,21 @@ Wan-Animate は、WAN チームが開発した人物アニメーションおよ ### 2. モデルのダウンロードリンク -**Diffusion Models** - - - - Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: Kijai のリポジトリから提供されるスケーリング FP8 モデル - - - wan2.2_animate_14B_bf16.safetensors: 元の bf16 モデル重み - - - -**CLIP Vision** - - - clip_vision_h.safetensors: CLIP Vision エンコーダ - - -**LoRAs** - - - lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4 ステップ高速化 LoRA - +**diffusion_models** +- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) — Kijai のリポジトリから提供されるモデル +- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) — 元のモデル重み -**VAE** +**clip_visions** +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) - - wan_2.1_vae.safetensors: エンコードおよびデコード用の Wan2.1 VAE - +**loras** +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) — 4ステップ高速化対応の LoRA -**Text Encoders** +**vae** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - umt5_xxl_fp8_e4m3fn_scaled.safetensors: スケーリング FP8 テキストエンコーダ - +**text_encoders** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-fun-camera.mdx b/ja/tutorials/video/wan/wan2-2-fun-camera.mdx index 8ba4abf19..2636c306a 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -46,60 +46,31 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Fun Camera" を検索 - - + +

JSON ワークフローをダウンロード

+ 以下の画像をダウンロードし、入力として使用します。 - - - 動画生成の開始フレーム。この画像をダウンロードして使用するか、ご自身の画像に置き換えてください。 - - +![入力開始画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) ### 2. モデルのダウンロードリンク 以下のモデルは、[Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 **Diffusion モデル** - - - - Wan2.2 Fun Camera 用高ノイズ拡散モデル - - - Wan2.2 Fun Camera 用低ノイズ拡散モデル - - +- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA（オプション：高速化用）** - - - - 高ノイズモデル用 4ステップ加速 LoRA - - - 低ノイズモデル用 4ステップ加速 LoRA - - +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - エンコード/デコード用 Wan2.1 VAE - - -**テキストエンコーダー** - - - FP8 スケーリング版テキストエンコーダー - +**テキストエンコーダー** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ファイル保存先のディレクトリ構成： @@ -141,4 +112,4 @@ ComfyUI/ - **Width/Height（幅／高さ）**：動画の解像度を設定 - **Length（長さ）**：動画のフレーム数を設定（デフォルトは81フレーム） - **Speed（速度）**：動画の再生速度を設定（デフォルトは1.0） -8. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合は Cmd）+ Enter` を押して動画生成を実行してください。 +8. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合は Cmd）+ Enter` を押して動画生成を実行してください。 \ No newline at end of file diff --git a/ja/tutorials/video/wan/wan2-2-fun-control.mdx b/ja/tutorials/video/wan/wan2-2-fun-control.mdx index 9bf54688c..87fd9ec0c 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-control.mdx @@ -55,27 +55,27 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### 1. ワークフローと素材のダウンロード -ComfyUI を最新バージョンに更新し、ワークフローファイルをダウンロードして ComfyUI にドラッグするか、テンプレートライブラリの `Workflow` → `Browse Templates` → `Video` から "Wan2.2 Fun Control" を見つけてください。 +以下の動画または JSON ファイルをダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Fun Control" を検索 - - + + + +

JSON ワークフローをダウンロード

+ 入力素材として以下の画像および動画をダウンロードしてください。 - - - 動画生成の開始フレームです。この画像をダウンロードして使用するか、ご自身の画像に置き換えてください。 - - - 前処理済みのポーズ制御動画です。この動画をダウンロードして使用するか、ご自身の動画に置き換えてください。 - - +![入力開始画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) + + > ここでは前処理済みの動画を使用しています。 @@ -83,39 +83,19 @@ ComfyUI を最新バージョンに更新し、ワークフローファイルを 以下のモデルは [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) で見つかります -**Diffusion Models** - - - - wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors：高ノイズ拡散モデル - - - wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors：低ノイズ拡散モデル - - +**Diffusion Model** +- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) -**Wan2.2-Lightning LoRA（オプション、加速用）** - - - - wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors：高ノイズ 4 ステップ加速 LoRA - - - wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors：低ノイズ 4 ステップ加速 LoRA - - +**Wan2.2-Lightning LoRA (オプション、加速用)** +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - wan_2.1_vae.safetensors：エンコード/デコード用 Wan2.1 VAE - - -**Text Encoder** - - - umt5_xxl_fp8_e4m3fn_scaled.safetensors：スケーリング FP8 テキストエンコーダ - +**Text Encoder** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx index 8fb36d964..b00d7fa5d 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -61,14 +61,13 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem または、ComfyUI を最新版に更新した上で、以下のリンクからワークフローファイルをダウンロードし、ComfyUI の画面にドラッグ＆ドロップして読み込んでください。 - - - JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 Fun Inp" を検索してください - - - Comfy Cloud で開く - - + +

JSON形式ワークフローをダウンロード

+ + + +

Comfy Cloud で実行

+ 以下の画像を開始フレームおよび終了フレームの素材としてご使用ください。 @@ -78,38 +77,18 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem ### 2. モデルの準備 **Diffusion モデル** - - - - wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: 首尾フレーム修復用の高ノイズ拡散モデル - - - wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: 首尾フレーム修復用の低ノイズ拡散モデル - - +- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) **Lightning LoRA（任意：高速化用）** - - - - wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 高ノイズモデル用の 4 ステップ高速化 LoRA - - - wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 低ノイズモデル用の 4 ステップ高速化 LoRA - - +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - wan_2.1_vae.safetensors: エンコードおよびデコード用の Wan2.1 VAE - - -**テキストエンコーダー** - - - umt5_xxl_fp8_e4m3fn_scaled.safetensors: スケーリングされた FP8 テキストエンコーダー - +**テキストエンコーダー** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-s2v.mdx b/ja/tutorials/video/wan/wan2-2-s2v.mdx index 179e2aa60..872e49120 100644 --- a/ja/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ja/tutorials/video/wan/wan2-2-s2v.mdx @@ -35,58 +35,38 @@ Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - - - Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで "Wan2.2 S2V" を検索 - - + +

JSON ワークフローをダウンロード

+ + + +

Comfy Cloud で実行

+ 以下の画像および音声ファイルを入力としてダウンロードしてください: +![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) + - - - デフォルトの入力画像をダウンロードするか、ご自身の画像をお使いください。 - - - デフォルトの入力音声をダウンロードするか、ご自身の音声をお使いください。 - - + +

入力音声をダウンロード

+ ### 2. モデルのダウンロードリンク すべてのモデルは、[当社のリポジトリ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 -**diffusion_models** - - - - FP8 scaled diffusion model。ComfyUI/models/diffusion_models/ に配置 - - - BF16 diffusion model。ComfyUI/models/diffusion_models/ に配置 - - - -**audio_encoders** - - - Audio encoder model。ComfyUI/models/audio_encoders/ に配置 - - -**vae** +**diffusion_models** +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) - - Wan2.1 VAE model。ComfyUI/models/vae/ に配置 - +**audio_encoders** +- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) -**text_encoders** +**vae** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - FP8 scaled text encoder。ComfyUI/models/text_encoders/ に配置 - +**text_encoders** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -114,14 +94,8 @@ ComfyUI/ 両方のモデルは、[こちらのページ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) から入手可能です： - - - FP8 scaled diffusion model - - - BF16 diffusion model - - +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) 本テンプレートでは `wan2.2_s2v_14B_fp8_scaled.safetensors` を使用しており、VRAM 使用量が少ないのが特徴です。ただし、品質劣化を抑えるために `wan2.2_s2v_14B_bf16.safetensors` を試すことも可能です。 diff --git a/ja/tutorials/video/wan/wan2_2.mdx b/ja/tutorials/video/wan/wan2_2.mdx index 2d439d858..7805a5892 100644 --- a/ja/tutorials/video/wan/wan2_2.mdx +++ b/ja/tutorials/video/wan/wan2_2.mdx @@ -102,25 +102,24 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - - - JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 5B」を検索 - - - Comfy Cloud で開く - - + +

JSON ワークフローファイルをダウンロード

+ + + +

Run on Comfy Cloud

+ ### 2. モデルの手動ダウンロード **Diffusion Model** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) +- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan2.2_vae.safetensors) +- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -158,26 +157,25 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - - - JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 14B T2V」を検索 - - - Comfy Cloud で開く - - + +

JSON ワークフローファイルをダウンロード

+ + + +

Run on Comfy Cloud

+ ### 2. モデルの手動ダウンロード **Diffusion Model** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -216,14 +214,13 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - - - JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 14B I2V」を検索 - - - Comfy Cloud で開く - - + +

JSON ワークフローファイルをダウンロード

+ + + +

Run on Comfy Cloud

+ 以下の画像を入力として使用できます： ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -231,14 +228,14 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows ### 2. モデルの手動ダウンロード **Diffusion Model** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) +- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -276,14 +273,13 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - - - JSON ワークフローをダウンロード、またはテンプレートライブラリで「Wan2.2 14B FLF2V」を検索 - - - Comfy Cloud で開く - - + +

JSON ワークフローをダウンロード

+ + + +

Run on Comfy Cloud

+ 以下の画像を入力素材としてダウンロードしてください： diff --git a/ja/tutorials/video/zai/scail2.mdx b/ja/tutorials/video/zai/scail2.mdx index 0ad537457..0e91e1098 100644 --- a/ja/tutorials/video/zai/scail2.mdx +++ b/ja/tutorials/video/zai/scail2.mdx @@ -105,23 +105,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### 必要なモデル **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders**（いずれか） -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### ファイル保存場所 diff --git a/ko/tutorials/3d/hunyuan3D-2.mdx b/ko/tutorials/3d/hunyuan3D-2.mdx index 1841d79c9..a91bcb675 100644 --- a/ko/tutorials/3d/hunyuan3D-2.mdx +++ b/ko/tutorials/3d/hunyuan3D-2.mdx @@ -57,14 +57,9 @@ Hunyuan3D-2mv 워크플로우에서는 다중뷰 이미지를 사용해 3D 모 - - - Comfy Cloud에서 이 워크플로우를 즉시 실행 - - - 워크플로우 JSON 파일 다운로드 - - + +

Comfy Cloud에서 실행

+ ### 1. 워크플로우 @@ -86,7 +81,7 @@ Hunyuan3D-2mv 워크플로우에서는 다중뷰 이미지를 사용해 3D 모 아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv.safetensors`로 이름을 변경할 수 있습니다. ``` ComfyUI/ @@ -109,14 +104,9 @@ ComfyUI/ Hunyuan3D-2mv-turbo 워크플로우에서는 Hunyuan3D-2mv-turbo 모델을 사용해 3D 모델을 생성합니다. 이 모델은 Hunyuan3D-2mv의 단계 증류 버전으로, 더 빠른 3D 모델 생성을 가능하게 합니다. 이번 버전의 워크플로우에서는 `cfg`를 1.0으로 설정하고, `flux guidance` 노드를 추가해 `증류된 cfg` 생성을 제어합니다. - - - Comfy Cloud에서 이 워크플로우를 즉시 실행 - - - 워크플로우 JSON 파일 다운로드 - - + +

Comfy Cloud에서 실행

+ ### 1. 워크플로우 @@ -135,7 +125,7 @@ Hunyuan3D-2mv-turbo 워크플로우에서는 Hunyuan3D-2mv-turbo 모델을 사 아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv-turbo.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv-turbo.safetensors`로 이름을 변경할 수 있습니다. ``` ComfyUI/ @@ -156,14 +146,9 @@ ComfyUI/ Hunyuan3D-2 워크플로우에서는 Hunyuan3D-2 모델을 사용해 3D 모델을 생성합니다. 이 모델은 다중뷰 모델이 아닙니다. 이번 워크플로우에서는 `Hunyuan3Dv2ConditioningMultiView` 노드 대신 `Hunyuan3Dv2Conditioning` 노드를 사용합니다. - - - Comfy Cloud에서 이 워크플로우를 즉시 실행 - - - 워크플로우 JSON 파일 다운로드 - - + +

Comfy Cloud에서 실행

+ ### 1. 워크플로우 @@ -178,7 +163,7 @@ Hunyuan3D-2 워크플로우에서는 Hunyuan3D-2 모델을 사용해 3D 모델 아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2.safetensors`로 이름을 변경할 수 있습니다. ``` ComfyUI/ diff --git a/ko/tutorials/3d/triposplat.mdx b/ko/tutorials/3d/triposplat.mdx index d8b14a750..18e78fdb3 100644 --- a/ko/tutorials/3d/triposplat.mdx +++ b/ko/tutorials/3d/triposplat.mdx @@ -106,23 +106,23 @@ TripoSplat은 단일 RGB 이미지를 입력으로 받아 3D 가우시안 프리 TripoSplat 모델과 필요한 파일을 다운로드하세요. 해당 `models/` 하위 디렉토리에 배치하세요. - + triposplat_fp16.safetensors — TripoSplat 확산 모델 체크포인트 - + triposplat_vae_decoder_fp16.safetensors — VAE 디코더 - + flux2-vae.safetensors — Flux.2 VAE, 잠재적 인코딩용 - + dino_v3_vit_h.safetensors — CLIP 비전 인코더 (DINOv2) - + birefnet.safetensors — 전처리용 배경 제거 모델 diff --git a/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx index d0fd07c1d..4dfcd158a 100644 --- a/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -45,7 +45,7 @@ AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공 ### AIO 모델 다운로드 - + 올인원 체크포인트 파일 (대부분의 사용자에게 권장). @@ -74,19 +74,19 @@ AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공 ### 분할 모델 다운로드 - + 디퓨전 모델. - + 텍스트 인코더 (0.6B). - + 텍스트 인코더 (1.7B). - + VAE 모델. diff --git a/ko/tutorials/audio/ace-step/ace-step-v1.mdx b/ko/tutorials/audio/ace-step/ace-step-v1.mdx index b8646e8e4..9e80a4df1 100644 --- a/ko/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ko/tutorials/audio/ace-step/ace-step-v1.mdx @@ -33,15 +33,9 @@ ACE-Step은 중국 팀 StepFun과 ACE Studio가 공동 개발한 오픈소스 아래 버튼을 클릭해 해당 워크플로 파일을 다운로드하세요. 이를 ComfyUI로 드래그해 워크플로 정보를 로드하세요. 워크플로에는 모델 다운로드 정보도 포함되어 있습니다. - - - Json 형식 워크플로 파일 다운로드 - - + +

Json 형식 워크플로 파일 다운로드

+ 또는 [ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors)를 수동으로 다운로드해 `ComfyUI/models/checkpoints` 폴더에 저장할 수도 있습니다. @@ -64,27 +58,15 @@ ACE-Step은 중국 팀 StepFun과 ACE Studio가 공동 개발한 오픈소스 아래 버튼을 클릭해 해당 워크플로우 파일을 다운로드하세요. 이를 ComfyUI로 드래그해 워크플로우 정보를 로드하세요. - - - Json 형식 워크플로우 파일 다운로드 - - + +

Json 형식 워크플로우 파일 다운로드

+ 다음 오디오 파일을 입력 오디오로 다운로드하세요: - - - 입력용 예시 오디오 파일 다운로드 - - + +

입력용 예시 오디오 파일 다운로드

+ ### 2. 워크플로우 단계별 완료하기 diff --git a/ko/tutorials/audio/stable-audio/stable-audio-1.mdx b/ko/tutorials/audio/stable-audio/stable-audio-1.mdx index b27712c45..a732bad3d 100644 --- a/ko/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ko/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -41,7 +41,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### 체크포인트 - + 2.3GB. models/checkpoints/ 폴더에 배치하세요. @@ -56,7 +56,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### 텍스트 인코더 - + 프롬프트 조건부 설정을 위한 텍스트 인코더. models/text_encoders/ 폴더에 배치하세요. diff --git a/ko/tutorials/audio/stable-audio/stable-audio-3.mdx b/ko/tutorials/audio/stable-audio/stable-audio-3.mdx index fd318f6f1..bc5d2baa4 100644 --- a/ko/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ko/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -77,11 +77,11 @@ Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화 ### 체크포인트 - + Medium 워크플로우용. models/checkpoints/에 배치하세요 - + Medium Base 워크플로우용. models/checkpoints/에 배치하세요 @@ -97,11 +97,11 @@ Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화 ### 텍스트 인코더 - + 모든 Stable Audio 3 워크플로우에 필수. models/text_encoders/에 배치하세요 - + Medium 워크플로우용 (Qwen 리프롬프트). models/text_encoders/에 배치하세요 diff --git a/ko/tutorials/basic/inpaint.mdx b/ko/tutorials/basic/inpaint.mdx index 51c611b17..d542c9610 100644 --- a/ko/tutorials/basic/inpaint.mdx +++ b/ko/tutorials/basic/inpaint.mdx @@ -34,7 +34,7 @@ AI 이미지 생성 과정에서 우리는 종종 전체적인 이미지는 만 #### 1. 모델 설치 -[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) 파일을 다운로드하여 `ComfyUI/models/checkpoints` 폴더에 넣으세요: +[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) 파일을 다운로드하여 `ComfyUI/models/checkpoints` 폴더에 넣으세요: #### 2. 인페인팅 자산 @@ -68,7 +68,7 @@ AI 이미지 생성 과정에서 우리는 종종 전체적인 이미지는 만 ![SD1.5 인페인팅 결과](/images/tutorial/basic/inpaint/inpaint_sd1.5_pruned_emaonly.png) -[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) 모델로 생성한 결과가 더 나은 인페인팅 효과와 자연스러운 전환을 보여줍니다. 이는 해당 모델이 인페인팅에 특화되어 있어 생성 영역을 더 잘 제어할 수 있고, 결과적으로 인페인팅 효과가 개선되기 때문입니다. +[512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) 모델로 생성한 결과가 더 나은 인페인팅 효과와 자연스러운 전환을 보여줍니다. 이는 해당 모델이 인페인팅에 특화되어 있어 생성 영역을 더 잘 제어할 수 있고, 결과적으로 인페인팅 효과가 개선되기 때문입니다. 앞서 사용했던 비유를 기억하시나요? 서로 다른 모델은 각각 다른 능력을 가진 예술가와 같으며, 각 예술가는 자신의 한계를 가지고 있습니다. 적합한 모델을 선택하면 더 나은 생성 결과를 얻을 수 있습니다. diff --git a/ko/tutorials/basic/outpaint.mdx b/ko/tutorials/basic/outpaint.mdx index 8211bd46e..fb246ba21 100644 --- a/ko/tutorials/basic/outpaint.mdx +++ b/ko/tutorials/basic/outpaint.mdx @@ -34,7 +34,7 @@ AI 이미지 생성 과정에서 종종 기존 이미지의 구도는 좋지만 #### 1. 모델 설치 다음 모델 파일을 다운로드하여 `ComfyUI/models/checkpoints` 디렉토리에 저장하세요: -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) #### 2. 입력 이미지 diff --git a/ko/tutorials/controlnet/controlnet.mdx b/ko/tutorials/controlnet/controlnet.mdx index a1ab868dc..ca0a1ca24 100644 --- a/ko/tutorials/controlnet/controlnet.mdx +++ b/ko/tutorials/controlnet/controlnet.mdx @@ -73,8 +73,8 @@ ControlNet의 등장으로 우리는 추가적인 조건을 도입해 이미지 - [dreamCreationVirtual3DECommerce_v10.safetensors](https://civitai.com/api/download/models/731340?type=Model&format=SafeTensor&size=full&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/depth-controlnet.mdx b/ko/tutorials/controlnet/depth-controlnet.mdx index d9241f202..eded209da 100644 --- a/ko/tutorials/controlnet/depth-controlnet.mdx +++ b/ko/tutorials/controlnet/depth-controlnet.mdx @@ -54,7 +54,7 @@ Depth ControlNet은 깊이 맵 정보를 이해하고 활용하도록 특별히 - [architecturerealmix_v11.safetensors](https://civitai.com/api/download/models/431755?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) +- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/depth-t2i-adapter.mdx b/ko/tutorials/controlnet/depth-t2i-adapter.mdx index ffb32bc4d..413009f36 100644 --- a/ko/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/ko/tutorials/controlnet/depth-t2i-adapter.mdx @@ -76,7 +76,7 @@ ComfyUI에서 T2I 어댑터를 사용하는 방법은 [ControlNet](/ko/tutorials - [interiordesignsuperm_v2.safetensors](https://civitai.com/api/download/models/93152?type=Model&format=SafeTensor&size=full&fp=fp16) -- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_depth_sd15v2.pth?download=true) +- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/resolve/main/models/t2iadapter_depth_sd15v2.pth?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/mixing-controlnets.mdx b/ko/tutorials/controlnet/mixing-controlnets.mdx index 375237151..2063e94ab 100644 --- a/ko/tutorials/controlnet/mixing-controlnets.mdx +++ b/ko/tutorials/controlnet/mixing-controlnets.mdx @@ -54,9 +54,9 @@ AI 이미지 생성에서 단일 제어 조건은 종종 복잡한 장면의 요 - [awpainting_v14.safetensors](https://civitai.com/api/download/models/624939?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx b/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx index e7701c918..16a13ffcf 100644 --- a/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -47,10 +47,10 @@ AI 이미지 생성에서 OpenPose로 생성된 골격 구조 맵은 ControlNet 네트워크 환경상 해당 모델의 자동 다운로드가 성공적으로 이루어지지 않는다면, 아래 모델을 수동으로 다운로드해 지정된 디렉토리에 배치해주세요: -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) - [majicmixRealistic_v7.safetensors](https://civitai.com/api/download/models/176425?type=Model&format=SafeTensor&size=pruned&fp=fp16) - [japaneseStyleRealistic_v20.safetensors](https://civitai.com/api/download/models/85426?type=Model&format=SafeTensor&size=pruned&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/ko/tutorials/flux/flux-1-controlnet.mdx b/ko/tutorials/flux/flux-1-controlnet.mdx index 13217fd92..42ab13e6c 100644 --- a/ko/tutorials/flux/flux-1-controlnet.mdx +++ b/ko/tutorials/flux/flux-1-controlnet.mdx @@ -49,14 +49,9 @@ Depth 버전은 깊이 맵 추출 기법을 통해 원본 이미지의 공간적 ## FLUX.1-Canny-dev 전체 버전 워크플로우 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Canny" 검색 - - - Comfy Cloud에서 열기 - - + +

Comfy Cloud에서 실행

+ ### 1. 워크플로우 및 자산 @@ -78,10 +73,10 @@ Depth 버전은 깊이 맵 추출 기법을 통해 원본 이미지의 공간적 완전한 모델 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true) (해당 리포지토리의 이용 약관에 동의했는지 확인해 주세요) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true) (해당 리포지토리의 이용 약관에 동의했는지 확인해 주세요) 파일 저장 위치: ``` @@ -121,14 +116,9 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora 워크플로우 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Depth LoRA" 검색 - - - Comfy Cloud에서 열기 - - + +

Comfy Cloud에서 실행

+ LoRA 버전 워크플로우는 완전한 버전을 기반으로 LoRA 모델을 추가한 것입니다. [Flux 워크플로우의 전체 버전](/ko/tutorials/flux/flux-1-text-to-image)과 비교해, 해당 LoRA 모델을 로드하고 사용하는 노드가 추가되었습니다. @@ -149,11 +139,11 @@ LoRA 버전 워크플로우는 완전한 버전을 기반으로 LoRA 모델을 완전한 모델 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/flux/flux-1-fill-dev.mdx b/ko/tutorials/flux/flux-1-fill-dev.mdx index 9a6398e02..36a6de4e8 100644 --- a/ko/tutorials/flux/flux-1-fill-dev.mdx +++ b/ko/tutorials/flux/flux-1-fill-dev.mdx @@ -40,10 +40,10 @@ Flux.1 fill dev의 주요 특징: ![Flux 약정](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 완성된 모델 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) 파일 저장 위치: ``` @@ -62,14 +62,13 @@ ComfyUI/ ### 1. 인페인팅 워크플로우 및 자산 - - - Download JSON or search "flux_fill_inpaint" in Template Library - - - Open in Comfy Cloud - - + +

워크플로우 이미지 다운로드

+ + + +

Comfy Cloud에서 실행

+ 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![ComfyUI Flux.1 인페인팅](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -98,15 +97,6 @@ ComfyUI/ ### 1. 아웃페인팅 워크플로우 및 자산 - - - Download JSON or search "flux_fill_outpaint" in Template Library - - - Open in Comfy Cloud - - - 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![ComfyUI Flux.1 아웃페인팅](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/ko/tutorials/flux/flux-1-kontext-dev.mdx b/ko/tutorials/flux/flux-1-kontext-dev.mdx index 7435bcd27..b74cc5546 100644 --- a/ko/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ko/tutorials/flux/flux-1-kontext-dev.mdx @@ -37,9 +37,9 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 ### 버전 정보 -- **[FLUX.1 Kontext [pro]** — 상용 버전, 빠른 반복 편집에 중점 -- **FLUX.1 Kontext [max]** — 더 강력한 프롬프트 적합성을 갖춘 실험용 버전 -- **FLUX.1 Kontext [dev]** — 오픈소스 버전 (본 튜토리얼에서 사용), 120억 파라미터, 주로 연구용 +- **[FLUX.1 Kontext [pro]** - 상용 버전, 빠른 반복 편집에 중점 +- **FLUX.1 Kontext [max]** - 더 강력한 프롬프트 적합성을 갖춘 실험용 버전 +- **FLUX.1 Kontext [dev]** - 오픈소스 버전 (본 튜토리얼에서 사용), 120억 파라미터, 주로 연구용 현재 ComfyUI에서는 이 모든 버전을 사용할 수 있으며, [Pro 및 Max 버전](/ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext)은 파트너 노드를 통해 호출할 수 있고, Dev 오픈소스 버전은 본 가이드의 지침을 참고하세요. @@ -51,7 +51,7 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 **확산 모델** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) 원본 가중치를 사용하고 싶다면 블랙 포레스트 랩스의 관련 리포지토리를 방문해 원본 모델 가중치를 받아 사용할 수 있습니다. @@ -62,7 +62,7 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 **텍스트 인코더** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) 또는 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) 또는 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) 모델 저장 위치 @@ -80,14 +80,9 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 ## Flux.1 Kontext Dev 워크플로우 - - - JSON을 다운로드하거나 템플릿 라이브러리에서 "Flux Kontext Dev"를 검색하세요 - - - Comfy Cloud에서 열기 - - + +

Comfy Cloud에서 실행하기

+ 이 워크플로우는 `Load Image(from output)` 노드를 사용해 편집할 이미지를 불러오므로, 여러 차례의 편집을 위해 편집된 이미지에 더욱 편리하게 접근할 수 있습니다. diff --git a/ko/tutorials/flux/flux-1-text-to-image.mdx b/ko/tutorials/flux/flux-1-text-to-image.mdx index 2cd435ecc..88611043c 100644 --- a/ko/tutorials/flux/flux-1-text-to-image.mdx +++ b/ko/tutorials/flux/flux-1-text-to-image.mdx @@ -50,30 +50,25 @@ Flux는 뛰어난 이미지 품질과 유연성으로 고화질의 다양한 이 #### 1. 워크플로우 파일 - - - Comfy Cloud에서 이 워크플로우 실행 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Dev" 검색 - - - 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Dev 원본 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) + +

Comfy Cloud에서 실행

+ + #### 2. 수동 모델 설치 - `flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) 계약에 동의해야 합니다. -- VRAM이 부족한 경우, [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true)를 사용해 `t5xxl_fp16.safetensors` 파일을 대체해보세요. +- VRAM이 부족한 경우, [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true)를 사용해 `t5xxl_fp16.safetensors` 파일을 대체해보세요. 다음 모델 파일을 다운로드하세요: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) 저장 위치: ``` @@ -108,19 +103,14 @@ Flux의 뛰어난 프롬프트 추종 능력 덕분에 부정적인 프롬프트 #### 1. 워크플로우 파일 - - - Comfy Cloud에서 이 워크플로우 실행 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Schnell" 검색 - - - 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Schnell 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) + +

Comfy Cloud에서 실행

+ + #### 2. 수동 모델 설치 @@ -130,10 +120,10 @@ Flux의 뛰어난 프롬프트 추종 능력 덕분에 부정적인 프롬프트 완전한 모델 파일 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) 파일 저장 위치: ``` @@ -166,38 +156,24 @@ fp8 버전은 원본 Flux.1 fp16 버전의 양자화된 버전입니다. ### Flux.1 Dev - - - Comfy Cloud에서 이 워크플로우 실행 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Dev FP8" 검색 - - - 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Dev fp8 체크포인트 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. + +

Comfy Cloud에서 실행

+ + +[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. 해당 `Load Checkpoint` 노드가 `flux1-dev-fp8.safetensors`를 로드하도록 설정했는지 확인하고, 워크플로우를 실행해보세요. ### Flux.1 Schnell - - - Comfy Cloud에서 이 워크플로우 실행 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Schnell FP8" 검색 - - - 아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. ![Flux Schnell fp8 체크포인트 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. +[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. 해당 `Load Checkpoint` 노드가 `flux1-schnell-fp8.safetensors`를 로드하도록 설정했는지 확인하고, 워크플로우를 실행해보세요. diff --git a/ko/tutorials/flux/flux-1-uso.mdx b/ko/tutorials/flux/flux-1-uso.mdx index ddb733350..93f60cba9 100644 --- a/ko/tutorials/flux/flux-1-uso.mdx +++ b/ko/tutorials/flux/flux-1-uso.mdx @@ -31,14 +31,18 @@ USO는 세 가지 주요 방식을 지원합니다: ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - - - 워크플로우 JSON을 다운로드하고 ComfyUI로 드래그하세요 - - - Comfy Cloud에서 이 워크플로우 실행 - - + +

JSON 워크플로우 다운로드

+ + + +

Comfy Cloud에서 실행

+ 아래 이미지를 입력 이미지로 사용하세요. @@ -49,19 +53,19 @@ USO는 세 가지 주요 방식을 지원합니다: **체크포인트** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) **로라** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **모델 패치** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **클립 비전** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) 모든 모델을 다운로드해 다음 디렉토리에 배치하세요: diff --git a/ko/tutorials/flux/flux-2-dev.mdx b/ko/tutorials/flux/flux-2-dev.mdx index 99e2d074e..617406994 100644 --- a/ko/tutorials/flux/flux-2-dev.mdx +++ b/ko/tutorials/flux/flux-2-dev.mdx @@ -65,15 +65,15 @@ FLUX.2 Dev를 사용하여 단일 이미지를 생성하는 기본적인 텍스 **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/flux/flux-2-klein.mdx b/ko/tutorials/flux/flux-2-klein.mdx index fff51e9d2..69d31c6db 100644 --- a/ko/tutorials/flux/flux-2-klein.mdx +++ b/ko/tutorials/flux/flux-2-klein.mdx @@ -45,19 +45,19 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 ## Flux.2 Klein 4B 모델 다운로드 - + 4B 모델용 텍스트 인코더입니다. - + 확산 모델(4B 베이스). - + 확산 모델(4B 정제). - + 4B 모델용 VAE입니다. @@ -103,11 +103,11 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 확산 모델(9B 정제). - + 9B 모델용 텍스트 인코더입니다. - + 9B 모델용 VAE입니다. diff --git a/ko/tutorials/flux/flux1-krea-dev.mdx b/ko/tutorials/flux/flux1-krea-dev.mdx index 8a63c8b7c..f1cba3b61 100644 --- a/ko/tutorials/flux/flux1-krea-dev.mdx +++ b/ko/tutorials/flux/flux1-krea-dev.mdx @@ -31,14 +31,13 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 아래 이미지 또는 JSON 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![Flux Krea Dev 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - - - Comfy Cloud에서 이 워크플로우 실행 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Krea Dev" 검색 - - + +

JSON 워크플로우 다운로드

+ + + +

Comfy Cloud에서 실행하기

+ #### 2. 수동 모델 설치 @@ -49,7 +48,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 더 높은 품질을 원하고 VRAM이 충분하다면 원본 모델 가중치를 사용해 보실 수 있습니다. -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) `flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 약정에 동의해야 합니다. @@ -58,12 +57,12 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 이전에 Flux 관련 워크플로우를 사용한 적이 있다면, 다음 모델들은 동일하므로 다시 다운로드할 필요가 없습니다. **텍스트 인코더** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/image/anima/anima.mdx b/ko/tutorials/image/anima/anima.mdx index 84ec6b714..038e74d36 100644 --- a/ko/tutorials/image/anima/anima.mdx +++ b/ko/tutorials/image/anima/anima.mdx @@ -84,15 +84,15 @@ Anima는 두 가지 워크플로를 제공합니다. 일반 사용을 위한 베 모델 파일은 Hugging Face의 [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima)에서 확인할 수 있습니다. - + Anima Base v1용 확산 모델(20억). - + 두 워크플로 공유 텍스트 인코더(Qwen-3 0.6B). - + 두 워크플로 공유 VAE. @@ -113,7 +113,7 @@ Anima는 두 가지 워크플로를 제공합니다. 일반 사용을 위한 베 미리보기 워크플로를 사용한다면, 대신 미리보기 확산 모델을 다운로드하세요: - + Anima Preview용 확산 모델(20억). diff --git a/ko/tutorials/image/boogu/boogu-image-0.1.mdx b/ko/tutorials/image/boogu/boogu-image-0.1.mdx index 043e48c95..ae1e7aa94 100644 --- a/ko/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ko/tutorials/image/boogu/boogu-image-0.1.mdx @@ -50,19 +50,19 @@ Boogu-Image-0.1-Turbo 워크플로는 확산, 텍스트 인코딩 및 VAE 디코 ### Boogu-Image-0.1-Turbo 모델 다운로드 - + Boogu-Image-0.1-Turbo용 확산 모델. - + Boogu-Image-0.1-Turbo용 텍스트 인코더. - + Boogu-Image-0.1-Turbo용 VAE. - + Boogu-Image-0.1-Turbo용 LoRA 모듈 (rank-128). @@ -100,15 +100,15 @@ Boogu-Image-0.1-Turbo 워크플로는 확산, 텍스트 인코딩 및 VAE 디코 ### Boogu-Image-0.1-Edit 모델 다운로드 - + Boogu-Image-0.1-Edit용 확산 모델. - + Boogu-Image-0.1-Edit용 텍스트 인코더. - + Boogu-Image-0.1-Edit용 VAE. diff --git a/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index 6f0fad7c3..e72a01eb8 100644 --- a/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -44,17 +44,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **디퓨전 모델** -- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_t2i.safetensors) +- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors) 기타 가중치는 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged)에서 다운로드해 주세요. **텍스트 인코더** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) 파일 저장 위치 diff --git a/ko/tutorials/image/ernie-image/ernie-image.mdx b/ko/tutorials/image/ernie-image/ernie-image.mdx index 079f9e47e..8c72a09e3 100644 --- a/ko/tutorials/image/ernie-image/ernie-image.mdx +++ b/ko/tutorials/image/ernie-image/ernie-image.mdx @@ -52,19 +52,19 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" 모든 재포장된 모델 파일은 Hugging Face의 [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image)에서 확인할 수 있습니다. - + ERNIE-Image용 디퓨전 모델입니다. - + ERNIE-Image용 텍스트 인코더입니다. - + ERNIE-Image용 프롬프트 향상기 텍스트 인코더입니다. - + ERNIE-Image용 VAE입니다. @@ -96,19 +96,19 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### ERNIE-Image-Turbo 모델 다운로드 - + ERNIE-Image-Turbo용 diffusion 모델입니다. - + ERNIE-Image-Turbo용 텍스트 인코더입니다. - + ERNIE-Image-Turbo용 프롬프트 향상기 텍스트 인코더입니다. - + ERNIE-Image-Turbo용 VAE입니다. diff --git a/ko/tutorials/image/hidream/hidream-e1.mdx b/ko/tutorials/image/hidream/hidream-e1.mdx index a326ec40f..86a1e2e38 100644 --- a/ko/tutorials/image/hidream/hidream-e1.mdx +++ b/ko/tutorials/image/hidream/hidream-e1.mdx @@ -42,8 +42,8 @@ HiDream-E1은 HiDream-ai가 공식적으로 오픈소스로 배포한 대규모 **Diffusion 모델** 두 모델을 모두 다운로드할 필요는 없습니다. E1.1은 E1을 기반으로 한 반복 버전이므로, 우리의 테스트 결과에 따르면 E1보다 품질과 성능이 크게 향상되었습니다. -- [hidream_e1_1_bf16.safetensors (권장)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors (권장)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **텍스트 인코더**: @@ -76,15 +76,6 @@ HiDream-E1은 HiDream-ai가 공식적으로 오픈소스로 배포한 대규모 ## HiDream E1.1 ComfyUI 네이티브 워크플로우 예시 - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "HiDream E1.1" 검색 - - - E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전은 동적 1메가픽셀 해상도를 지원하며, 워크플로우에서는 `Scale Image to Total Pixels` 노드를 사용해 입력 이미지를 동적으로 100만 픽셀로 조정합니다. @@ -128,14 +119,9 @@ E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전 ## HiDream E1 ComfyUI 네이티브 워크플로우 예시 - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "HiDream E1 Full" 검색 - - + +

Comfy Cloud에서 실행

+ E1은 2025년 4월 28일에 출시된 모델입니다. 이 모델은 768*768 해상도만 지원합니다. diff --git a/ko/tutorials/image/hidream/hidream-i1.mdx b/ko/tutorials/image/hidream/hidream-i1.mdx index 1042b02c6..2b27ec88d 100644 --- a/ko/tutorials/image/hidream/hidream-i1.mdx +++ b/ko/tutorials/image/hidream/hidream-i1.mdx @@ -98,21 +98,16 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 풀버전 워크플로우 - - - 설정 없이 Comfy Cloud에서 이 워크플로우 실행 - - - 워크플로우 JSON 파일 다운로드 - - + +

Comfy Cloud에서 실행

+ #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 @@ -141,20 +136,15 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Dev 버전 워크플로우 - - - 설정 없이 Comfy Cloud에서 이 워크플로우 실행 - - - 워크플로우 JSON 파일 다운로드 - - + +

Comfy Cloud에서 실행

+ #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. @@ -182,20 +172,15 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Fast 버전 워크플로우 - - - 설정 없이 Comfy Cloud에서 이 워크플로우 실행 - - - 워크플로우 JSON 파일 다운로드 - - + +

Comfy Cloud에서 실행

+ #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. diff --git a/ko/tutorials/image/hidream/hidream-o1.mdx b/ko/tutorials/image/hidream/hidream-o1.mdx index cfa00d520..4741dcbbe 100644 --- a/ko/tutorials/image/hidream/hidream-o1.mdx +++ b/ko/tutorials/image/hidream/hidream-o1.mdx @@ -50,19 +50,19 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 **체크포인트** — 재패키징되고 양자화되었습니다. 모든 모델은 최악의 이상치에 대해 bf16을 사용하며, 사용되지 않는 딥스택 레이어는 제거되었습니다: -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 양자화된 변형 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 양자화된 변형 +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) **텍스트 인코더**(프롬프트 강화) — 모든 버전에서 공유됩니다: -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA(선택사항)** — Dev 디스틸레이션은 풀 모델에도 LoRA로 적용할 수 있으며, 디스틸레이션 강도를 조정할 수 있습니다([Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 제공): -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 전체 랭크 -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 프룬드 변형 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 대안적인 체크포인트 기반 디스틸레이션 +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 전체 랭크 +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 프룬드 변형 +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 대안적인 체크포인트 기반 디스틸레이션 ``` 📂 ComfyUI/ @@ -105,13 +105,13 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 **체크포인트(Dev)** — 재패키징되고 양자화되었습니다. 모든 모델은 최악의 이상치에 대해 bf16을 사용하며, 사용되지 않는 딥스택 레이어는 제거되었습니다: -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 양자화된 변형 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 양자화된 변형 +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) **텍스트 인코더**(프롬프트 강화) — 모든 버전에서 공유됩니다: -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/ko/tutorials/image/ideogram/ideogram-v4.mdx b/ko/tutorials/image/ideogram/ideogram-v4.mdx index 717de0dbd..9ff8cc452 100644 --- a/ko/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ko/tutorials/image/ideogram/ideogram-v4.mdx @@ -46,23 +46,23 @@ Ideogram 4.0은 Ideogram에서 출시한 최신 텍스트 기반 이미지 생 Hugging Face의 [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4)에서 모든 재포장된 모델 파일을 확인할 수 있습니다. - + Ideogram 4.0용 디퓨전 모델 (~13.8 GB). models/diffusion_models/에 저장하세요 - + Ideogram 4.0용 비조건 디퓨전 모델 (~13.8 GB). models/diffusion_models/에 저장하세요 - + Ideogram 4.0용 텍스트 인코더 (~8 GB). models/text_encoders/에 저장하세요 - + Ideogram 4.0용 텍스트 인코더 (~2 GB). models/text_encoders/에 저장하세요 - + Ideogram 4.0용 VAE (~335 MB). models/vae/에 저장하세요 diff --git a/ko/tutorials/image/krea/krea-2.mdx b/ko/tutorials/image/krea/krea-2.mdx index bf2cf4a7b..78aaa2437 100644 --- a/ko/tutorials/image/krea/krea-2.mdx +++ b/ko/tutorials/image/krea/krea-2.mdx @@ -107,13 +107,13 @@ Krea는 또한 Krea 2용 스타일 LoRA 컬렉션을 출시했습니다. **Custo 로컬에서 사용하려면 [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2)에서 ComfyUI 최적화 모델 파일을 다운로드하세요. - + krea2_turbo_fp8_scaled.safetensors: Turbo FP8 (대부분의 사용자에게 권장) - + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B 텍스트 인코더 - + qwen_image_vae.safetensors diff --git a/ko/tutorials/image/lens/lens.mdx b/ko/tutorials/image/lens/lens.mdx index 7d28b9d1c..8ceae12be 100644 --- a/ko/tutorials/image/lens/lens.mdx +++ b/ko/tutorials/image/lens/lens.mdx @@ -94,19 +94,19 @@ Lens Turbo는 추출된 변형으로, 더 적은 샘플링 단계로 이미지 모든 모델 파일은 Hugging Face의 [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens)에서 찾을 수 있습니다. - + Lens용 디퓨전 모델(BF16). - + Lens Turbo용 디퓨전 모델(BF16). - + Lens와 Lens Turbo가 공유하는 텍스트 인코더(GPT-OSS-20B). - + Lens와 Lens Turbo가 공유하는 VAE(FLUX.2). diff --git a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 2c53a6f10..798943559 100644 --- a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -30,14 +30,13 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## NewBie-image 텍스트 기반 이미지 생성 워크플로 -| -| -| JSON을 다운로드하거나 템플릿 라이브러리에서 'NewBie-image' 검색 -| -| -| 클라우드에서 열기 -| -| + +

JSON 워크플로 파일 다운로드

+ + + +

ComfyUI 클라우드에서 실행하기

+ @@ -45,16 +44,16 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/omnigen/omnigen2.mdx b/ko/tutorials/image/omnigen/omnigen2.mdx index dcc09b6e3..78ed966cf 100644 --- a/ko/tutorials/image/omnigen/omnigen2.mdx +++ b/ko/tutorials/image/omnigen/omnigen2.mdx @@ -42,13 +42,13 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 본 문서에는 다양한 워크플로우가 포함되어 있으므로, 해당 모델 파일과 설치 위치는 다음과 같습니다. 모델 파일의 다운로드 정보는 각 워크플로우에도 포함되어 있습니다. **확산 모델** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) 파일 저장 위치: ``` @@ -66,11 +66,9 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 ### 1. 워크플로우 파일 다운로드 - - - Comfy Cloud에서 실행 - - + +

Comfy Cloud에서 실행

+ ![텍스트 기반 이미지 생성 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -96,11 +94,9 @@ OmniGen2는 풍부한 이미지 편집 기능을 갖추고 있으며, 이미지 ### 1. 워크플로우 파일 다운로드 - - - Comfy Cloud에서 실행 - - + +

Comfy Cloud에서 실행

+ ![이미지 편집 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/ko/tutorials/image/ovis/ovis-image.mdx b/ko/tutorials/image/ovis/ovis-image.mdx index 5456cf94b..6586e2039 100644 --- a/ko/tutorials/image/ovis/ovis-image.mdx +++ b/ko/tutorials/image/ovis/ovis-image.mdx @@ -22,14 +22,13 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Ovis-Image 텍스트 기반 이미지 생성 워크플로우 - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Ovis image" 검색 - - + +

JSON 워크플로우 파일 다운로드

+ + + +

ComfyUI 클라우드에서 실행하기

+ @@ -37,15 +36,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/pixeldit/pixeldit.mdx b/ko/tutorials/image/pixeldit/pixeldit.mdx index acda9a892..bdd451a41 100644 --- a/ko/tutorials/image/pixeldit/pixeldit.mdx +++ b/ko/tutorials/image/pixeldit/pixeldit.mdx @@ -62,11 +62,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" PixelDiT는 두 개의 모델 파일을 사용합니다: 텍스트 인코더와 확산 모델입니다. - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 텍스트 인코더 - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 확산 모델 diff --git a/ko/tutorials/image/qwen/qwen-image-2512.mdx b/ko/tutorials/image/qwen/qwen-image-2512.mdx index 04da73481..e8b9a4dc4 100644 --- a/ko/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ko/tutorials/image/qwen/qwen-image-2512.mdx @@ -43,14 +43,9 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - - + Comfy Cloud에서 실행하기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Qwen-Image-2512" 검색 - - + ### 1. 워크플로우 파일 @@ -60,25 +55,28 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 - **텍스트 기반 이미지 생성 (Qwen-Image 2512)**: 표준 50단계 생성 - **텍스트 기반 이미지 생성 (Qwen-Image 2512 4단계)**: Lightning LoRA를 사용한 가속화된 4단계 생성 + +

JSON 워크플로우 다운로드

+ ### 2. 모델 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (선택사항 - 4단계 Lightning 가속화용)** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **디퓨전 모델** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (대부분의 사용자에게 권장됨) -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (충분한 VRAM을 보유하고 더 높은 품질을 원하는 경우) +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (대부분의 사용자에게 권장됨) +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (충분한 VRAM을 보유하고 더 높은 품질을 원하는 경우) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx index 3061dce3e..8f3152cb1 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -33,32 +33,31 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그하여 불러올 수 있습니다. - - - ComfyUI 클라우드에서 실행 - - - JSON 워크플로우 다운로드 - - + +

JSON 워크플로우 다운로드

+ + + +

ComfyUI 클라우드에서 실행

+ ### 2. 모델 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (선택사항 - 4단계 라이트닝 가속화용)** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **디퓨전 모델** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image-edit.mdx b/ko/tutorials/image/qwen/qwen-image-edit.mdx index 132119b81..23f22ef51 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit.mdx @@ -47,14 +47,13 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. ![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - - - JSON 워크플로우를 다운로드하거나 템플릿 라이브러리에서 "image_qwen_image_edit" 검색 - - - ComfyUI 클라우드에서 이 워크플로우 실행 (제로 설정) - - + +

JSON 워크플로우 다운로드

+ + + +

ComfyUI 클라우드에서 실행

+ 아래 이미지를 입력으로 다운로드하세요 ![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -65,19 +64,19 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 **디퓨전 모델** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) 모델 저장 위치 diff --git a/ko/tutorials/image/qwen/qwen-image-layered.mdx b/ko/tutorials/image/qwen/qwen-image-layered.mdx index 8da4e5dca..51aae2791 100644 --- a/ko/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ko/tutorials/image/qwen/qwen-image-layered.mdx @@ -30,15 +30,13 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 워크플로우 -| -| -| JSON 워크플로우 파일 다운로드 -| -| -| -| ComfyUI 클라우드에서 실행하기 -| -| + +

JSON 워크플로우 파일 다운로드

+ + + +

ComfyUI 클라우드에서 실행하기

+ @@ -46,15 +44,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image.mdx b/ko/tutorials/image/qwen/qwen-image.mdx index c06554198..1a4c7c333 100644 --- a/ko/tutorials/image/qwen/qwen-image.mdx +++ b/ko/tutorials/image/qwen/qwen-image.mdx @@ -58,12 +58,9 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - - - - - - + + Comfy Cloud에서 실행하기 + 이 문서에 첨부된 워크플로우에는 세 가지 다른 모델이 사용됩니다: 1. Qwen-Image 원본 모델 fp8_e4m3fn @@ -85,11 +82,14 @@ GPU: RTX4090D 24GB ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. ![Qwen-image Text-to-Image 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - + +

Qwen-Image 공식 모델용 워크플로우 다운로드

+ 증류 버전 - - + +

증류 모델용 워크플로우 다운로드

+ ### 2. 모델 다운로드 @@ -103,12 +103,12 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 **디퓨전 모델** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 증류 버전의 원저자는 cfg 1.0에서 15단계 사용을 권장합니다. @@ -117,15 +117,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** @@ -164,19 +164,18 @@ Qwen_image_distill 이것은 ControlNet 모델이므로 일반 ControlNet처럼 사용할 수 있습니다. - - - - - - + + Comfy Cloud에서 실행하기 + ### 1. 워크플로우 및 입력 이미지 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - + +

JSON 형식 워크플로우 다운로드

+ 아래 이미지를 입력으로 다운로드하세요. ![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -185,7 +184,7 @@ Qwen_image_distill 1. InstantX Controlnet -[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)를 다운로드해 `ComfyUI/models/controlnet/` 폴더에 저장하세요. +[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)를 다운로드해 `ComfyUI/models/controlnet/` 폴더에 저장하세요. 2. **Lotus Depth 모델** @@ -193,11 +192,11 @@ Qwen_image_distill **디퓨전 모델** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) **VAE 모델** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) 또는 어떤 SD1.5 VAE도 가능합니다. +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) 또는 어떤 SD1.5 VAE도 가능합니다. ``` ComfyUI/ @@ -221,12 +220,9 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets 모델 패치 워크플로우 - - - - - - + + Comfy Cloud에서 실행하기 + 이 모델은 실제로 ControlNet이 아니라, 캐니, 딥스, 인페인트 등 세 가지 다른 제어 모드를 지원하는 모델 패치입니다. @@ -239,7 +235,9 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( 아래 이미지를 다운로드해 ComfyUI로 드래그해 해당 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - + +

JSON 형식 워크플로우 다운로드

+ 아래 이미지를 입력으로 다운로드하세요: @@ -249,9 +247,9 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( 다른 모델은 Qwen-Image 기본 워크플로우와 동일합니다. 아래 모델만 다운로드해 `ComfyUI/models/model_patches` 폴더에 저장하면 됩니다. -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. 워크플로우 사용 지침 @@ -294,12 +292,9 @@ ControlNet 관련 워크플로우를 처음 사용한다면, 제어 이미지는 ## Qwen Image Union ControlNet LoRA 워크플로우 - - - - - - + + Comfy Cloud에서 실행하기 + 원본 모델 주소: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 캐니, 딥스, 포즈, 라인아트, 소프트엣지, 노말, 오픈포즈 지원 이미지 구조 제어 LoRA @@ -308,7 +303,9 @@ Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](htt 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - + +

JSON 형식 워크플로우 다운로드

+ 아래 이미지를 입력으로 다운로드하세요. diff --git a/ko/tutorials/image/z-image/z-image-turbo.mdx b/ko/tutorials/image/z-image/z-image-turbo.mdx index 220f1c488..f67002698 100644 --- a/ko/tutorials/image/z-image/z-image-turbo.mdx +++ b/ko/tutorials/image/z-image/z-image-turbo.mdx @@ -46,15 +46,15 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### Z-Image-Turbo 모델 다운로드 - + Z-Image-Turbo용 텍스트 인코더입니다. - + Z-Image-Turbo용 디퓨전 모델입니다. - + Z-Image-Turbo용 VAE입니다. @@ -81,7 +81,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### ControlNet용 추가 모델 - + Z-Image-Turbo용 ControlNet 모델 패치입니다. diff --git a/ko/tutorials/image/z-image/z-image.mdx b/ko/tutorials/image/z-image/z-image.mdx index 34a2ec450..127f2ee20 100644 --- a/ko/tutorials/image/z-image/z-image.mdx +++ b/ko/tutorials/image/z-image/z-image.mdx @@ -37,15 +37,15 @@ Z-Image (Base)는 커뮤니티 주도의 미세조정 및 맞춤형 개발을 ## Z-Image 모델 다운로드 - + Z-Image용 텍스트 인코더 - + Z-Image용 확산 모델 - + Z-Image용 VAE diff --git a/ko/tutorials/llm/gemma4/gemma4.mdx b/ko/tutorials/llm/gemma4/gemma4.mdx index 41acd9e86..81eff1335 100644 --- a/ko/tutorials/llm/gemma4/gemma4.mdx +++ b/ko/tutorials/llm/gemma4/gemma4.mdx @@ -74,11 +74,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" Gemma 4 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 관련 모델 파일을 다운로드하여 올바른 디렉토리에 저장하세요: - + 빠르고 가벼움. 소비자용 GPU에 권장됩니다. - + 균형 잡힌 성능. 워크플로의 기본 모델입니다. diff --git a/ko/tutorials/llm/qwen/qwen3.mdx b/ko/tutorials/llm/qwen/qwen3.mdx index 89eb6630f..f80db86ee 100644 --- a/ko/tutorials/llm/qwen/qwen3.mdx +++ b/ko/tutorials/llm/qwen/qwen3.mdx @@ -71,15 +71,15 @@ Qwen 3.0은 ComfyUI 워크플로우 내에서 구조화된 텍스트 생성과 Qwen 3.0 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 모델 파일은 Qwen3.5와 공유되며, 하드웨어에 가장 적합한 변형을 다운로드하세요: - + 경량형, 약 4.5GB. 낮은 VRAM 환경과 빠른 다운로드에 최적입니다. - + 균형 잡힌 크기와 품질. 대부분의 소비자용 GPU에 권장됩니다. - + 가장 큰 변형, 약 19GB. 더 높은 품질의 출력을 제공하며, 더 많은 VRAM이 필요합니다. diff --git a/ko/tutorials/llm/qwen/qwen3_5.mdx b/ko/tutorials/llm/qwen/qwen3_5.mdx index 84d0dde5e..098918a6c 100644 --- a/ko/tutorials/llm/qwen/qwen3_5.mdx +++ b/ko/tutorials/llm/qwen/qwen3_5.mdx @@ -73,15 +73,15 @@ Qwen3.5는 시각적 이해와 텍스트 생성을 결합해 ComfyUI 워크플 Qwen3.5 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 하드웨어에 가장 적합한 변형을 선택하세요: - + 경량형, 약 4.5GB. 낮은 VRAM 환경과 빠른 다운로드에 최적입니다. - + 균형 잡힌 크기와 품질. 대부분의 소비자용 GPU에 권장됩니다. - + 가장 큰 변형, 약 19GB. 더 높은 품질의 출력을 제공하며, 더 많은 VRAM을 요구합니다. diff --git a/ko/tutorials/partner-nodes/google/gemini.mdx b/ko/tutorials/partner-nodes/google/gemini.mdx index 72dc0b71b..a3e14d274 100644 --- a/ko/tutorials/partner-nodes/google/gemini.mdx +++ b/ko/tutorials/partner-nodes/google/gemini.mdx @@ -22,11 +22,13 @@ Google Gemini는 구글이 개발한 강력한 AI 모델로, 대화 및 텍스 아래 Json 파일을 다운로드한 후, ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. - - - Json 형식 워크플로우 파일 다운로드 - - + +

Json 형식 워크플로우 파일 다운로드

+ ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx index b288a04fa..af7f71b93 100644 --- a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -51,8 +51,9 @@ Kling 2.6 모션 컨트롤은 Kuaishou가 개발한 특수 다중모달 모델 ## Kling 2.6 모션 컨트롤 작업 흐름 - - + +

워크플로우 파일을 JSON 형식으로 다운로드하세요

+ ## 입력 요구사항 diff --git a/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 902c51f3a..36c342eb2 100644 --- a/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -54,11 +54,9 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_text_to_video.mp4" > - - - 워크플로우 파일(JSON 형식) 다운로드하기 - - + +

워크플로우 파일(JSON 형식) 다운로드하기

+ ### 2. 워크플로우 실행 단계 따라하기 @@ -82,11 +80,9 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_image_to_video.mp4" > - - - 워크플로우 파일(JSON 형식) 다운로드하기 - - + +

워크플로우 파일(JSON 형식) 다운로드하기

+ 아래 이미지를 입력 이미지로 다운로드하세요 @@ -116,11 +112,9 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video.mp4" > - - - 워크플로우 파일(JSON 형식) 다운로드하기 - - + +

워크플로우 파일(JSON 형식) 다운로드하기

+ 아래 비디오를 입력 비디오로 다운로드하세요: diff --git a/ko/tutorials/partner-nodes/openai/chat.mdx b/ko/tutorials/partner-nodes/openai/chat.mdx index 62be63ccc..e431f97a4 100644 --- a/ko/tutorials/partner-nodes/openai/chat.mdx +++ b/ko/tutorials/partner-nodes/openai/chat.mdx @@ -22,9 +22,13 @@ OpenAI는 생성형 AI에 중점을 둔 회사로, 강력한 대화 기능을 아래 Json 파일을 다운로드한 후, ComfyUI로 드래그하여 해당 워크플로우를 불러오세요. -| -| Json 형식 워크플로우 파일 다운로드 -| + +

Json 형식 워크플로우 파일 다운로드

+ ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/rodin/model-generation.mdx b/ko/tutorials/partner-nodes/rodin/model-generation.mdx index c0d95fe25..ad386d281 100644 --- a/ko/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ko/tutorials/partner-nodes/rodin/model-generation.mdx @@ -32,9 +32,13 @@ ComfyUI는 이제 해당 Rodin 모델 생성 API를 기본적으로 통합하여 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - - 단일뷰 모델 생성 (Json 형식) - + +

Json 형식 워크플로우 파일 다운로드

+ 입력 이미지로 아래 이미지를 다운로드하세요. @@ -62,9 +66,13 @@ ComfyUI는 이제 해당 Rodin 모델 생성 API를 기본적으로 통합하여 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - - 다중뷰 모델 생성 (Json 형식) - + +

Json 형식 워크플로우 파일 다운로드

+ 입력 이미지로 아래 이미지를 다운로드하세요. diff --git a/ko/tutorials/partner-nodes/runway/video-generation.mdx b/ko/tutorials/partner-nodes/runway/video-generation.mdx index 479ae13aa..aab91e761 100644 --- a/ko/tutorials/partner-nodes/runway/video-generation.mdx +++ b/ko/tutorials/partner-nodes/runway/video-generation.mdx @@ -37,7 +37,9 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen3a_turbo_image_to_video/runway_image_to_video_gen3a_turbo.mp4" > - + +

Json 형식 워크플로우 파일 다운로드

+ 아래 이미지를 입력 이미지로 다운로드하세요. @@ -65,7 +67,9 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen4_turbo_image_to_video/runway_gen4_turo_image_to_video.mp4" > - + +

Json 형식 워크플로우 파일 다운로드

+ 아래 이미지를 입력 이미지로 다운로드하세요. @@ -95,7 +99,9 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/first_last_frame_to_video/runway_first_last_frame.mp4" > - + +

Json 형식 워크플로우 파일 다운로드

+ 아래 이미지를 입력 이미지로 다운로드하세요. diff --git a/ko/tutorials/partner-nodes/tripo/model-generation.mdx b/ko/tutorials/partner-nodes/tripo/model-generation.mdx index ad2ec6454..f51ee9ef9 100644 --- a/ko/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ko/tutorials/partner-nodes/tripo/model-generation.mdx @@ -35,11 +35,9 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - - - 텍스트 기반 3D 모델 생성 워크플로우 - - + +

Json 형식 워크플로우 파일 다운로드

+ ### 2. 워크플로우를 단계별로 완료하세요 @@ -60,11 +58,9 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - - - 이미지 기반 3D 모델 생성 워크플로우 - - + +

Json 형식 워크플로우 파일 다운로드

+ 아래 이미지를 입력 이미지로 다운로드하세요. @@ -90,11 +86,9 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - - - 멀티뷰 모델 생성 워크플로우 - - + +

Json 형식 워크플로우 파일 다운로드

+ 아래 이미지를 입력 이미지로 다운로드하세요. diff --git a/ko/tutorials/utility/depth-anything-3.mdx b/ko/tutorials/utility/depth-anything-3.mdx index ebc59b59b..6dddf7005 100644 --- a/ko/tutorials/utility/depth-anything-3.mdx +++ b/ko/tutorials/utility/depth-anything-3.mdx @@ -39,10 +39,10 @@ ComfyUI는 이제 Depth Anything 3 노드를 기본 지원합니다. 시작하 Depth Anything 3 체크포인트를 다운로드하여 해당 ComfyUI 폴더에 저장합니다: -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — 가볍고 빠른 추론 -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — 균형 잡힌 성능 -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 단일 뷰 깊이에 최적 (하늘 감지 포함) -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 미터 단위의 물리적 깊이 (하늘 감지 포함) +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — 가볍고 빠른 추론 +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — 균형 잡힌 성능 +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 단일 뷰 깊이에 최적 (하늘 감지 포함) +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 미터 단위의 물리적 깊이 (하늘 감지 포함) ``` ComfyUI/ diff --git a/ko/tutorials/utility/face-detection/mediapipe.mdx b/ko/tutorials/utility/face-detection/mediapipe.mdx index e80bdec03..539561b3c 100644 --- a/ko/tutorials/utility/face-detection/mediapipe.mdx +++ b/ko/tutorials/utility/face-detection/mediapipe.mdx @@ -55,7 +55,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` → `템 MediaPipe 얼굴 감지 모델은 [Comfy-Org MediaPipe 모델 저장소](https://huggingface.co/Comfy-Org/mediapipe)에 호스팅되어 있습니다. -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) 다음과 같은 디렉토리 구조에 배치하세요: diff --git a/ko/tutorials/utility/moge.mdx b/ko/tutorials/utility/moge.mdx index 9867dd52b..af99d09df 100644 --- a/ko/tutorials/utility/moge.mdx +++ b/ko/tutorials/utility/moge.mdx @@ -50,8 +50,8 @@ ComfyUI는 이제 MoGe 노드를 기본적으로 지원합니다. 시작하기 MoGe 체크포인트를 다운로드하고 해당 ComfyUI 폴더에 저장하세요: -- **MoGe-2 (권장)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1 (기준)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2 (권장)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1 (기준)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/utility/pose-detection-sdpose.mdx b/ko/tutorials/utility/pose-detection-sdpose.mdx index 8ad8fdd07..e4f0b2dbb 100644 --- a/ko/tutorials/utility/pose-detection-sdpose.mdx +++ b/ko/tutorials/utility/pose-detection-sdpose.mdx @@ -87,11 +87,11 @@ ComfyUI를 최신 버전으로 업데이트한 다음, `Workflow` -> `Browse Tem SDPose 및 RT-DETRv4 모델 체크포인트는 [Comfy-Org SDPose 모델 저장소](https://huggingface.co/Comfy-Org/SDPose)에 호스팅되어 있습니다. **체크포인트** (SDPose 모델): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models** (RT-DETRv4 감지기): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (권장) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (전체 정밀도, 용량이 큼) +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (권장) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (전체 정밀도, 용량이 큼) 다음과 같은 디렉토리 구조에 배치하세요: diff --git a/ko/tutorials/utility/remove-background-birefnet.mdx b/ko/tutorials/utility/remove-background-birefnet.mdx index 1e4acf988..b23e07d8b 100644 --- a/ko/tutorials/utility/remove-background-birefnet.mdx +++ b/ko/tutorials/utility/remove-background-birefnet.mdx @@ -48,7 +48,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 BiRefNet 모델은 [Comfy-Org BiRefNet 모델 저장소](https://huggingface.co/Comfy-Org/BiRefNet)에 호스팅되어 있습니다. -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) 다음과 같은 디렉토리 구조에 배치하세요: diff --git a/ko/tutorials/utility/video-segment-sam3.mdx b/ko/tutorials/utility/video-segment-sam3.mdx index 8efcee351..8dfa7da68 100644 --- a/ko/tutorials/utility/video-segment-sam3.mdx +++ b/ko/tutorials/utility/video-segment-sam3.mdx @@ -68,7 +68,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 SAM 3.1 모델은 [Comfy-Org SAM 3.1 모델 저장소](https://huggingface.co/Comfy-Org/sam3.1)에 호스팅되어 있습니다. -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) 다음과 같은 디렉터리 구조에 배치하세요: diff --git a/ko/tutorials/utility/void-video-inpainting.mdx b/ko/tutorials/utility/void-video-inpainting.mdx index f26c7f60d..e8b63bf01 100644 --- a/ko/tutorials/utility/void-video-inpainting.mdx +++ b/ko/tutorials/utility/void-video-inpainting.mdx @@ -65,24 +65,24 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 **확산 모델** — 핵심 2회 통과 인페인팅 모델: -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — 정밀화 통과, 더 나은 시간적 안정성 -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — 기본 통과 +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 정밀화 통과, 더 나은 시간적 안정성 +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 기본 통과 **VAE:** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) **광학 흐름:** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 체크포인트** — 세그먼테이션용: -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **텍스트 인코더:** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ diff --git a/ko/tutorials/video/bytedance/bernini-r.mdx b/ko/tutorials/video/bytedance/bernini-r.mdx index 82597a548..ba520d810 100644 --- a/ko/tutorials/video/bytedance/bernini-r.mdx +++ b/ko/tutorials/video/bytedance/bernini-r.mdx @@ -49,16 +49,16 @@ ComfyUI는 이제 Bernini-R 노드를 기본 지원합니다. 시작하기 전 필요한 모델 가중치를 다운로드하여 해당 ComfyUI 폴더에 저장합니다: **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 0d25fe5dd..ce6479416 100644 --- a/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -15,27 +15,17 @@ Cosmos-Predict2는 텍스트 기반 이미지 생성(Text2Image)과 비디오- 산업 시뮬레이션, 자율주행, 도시 계획, 과학 연구 등 여러 분야에서 널리 사용되고 있습니다. 이는 지능형 비전과 물리 세계의 깊은 통합을 촉진하는 핵심 기반 도구입니다. - - - Cosmos-Predict2 소스 코드 및 문서 - - - Cosmos-Predict2 모음집 - - +GitHub: [Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) +huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) 이 가이드는 ComfyUI에서 **Video2World** 생성을 완료하는 과정을 안내합니다. 텍스트 기반 이미지 생성 섹션은 다음을 참고하세요: - - - Cosmos-Predict2를 이용한 텍스트 기반 이미지 생성 - - - 강력한 GPU로 Comfy Cloud에서 Cosmos-Predict2 워크플로우 실행 - - + + Cosmos-Predict2를 이용한 텍스트 기반 이미지 생성 + +{/* ## Cosmos Predict2 Video2World 워크플로우 @@ -52,14 +42,9 @@ Cosmos-Predict2는 텍스트 기반 이미지 생성(Text2Image)과 비디오- src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - - - JSON 형식 워크플로우 파일 다운로드 - - - Comfy Cloud에서 이 워크플로우 실행 (모델 사전 설치됨) - - + +

Json 형식 워크플로우 파일 다운로드

+ 다음 이미지를 입력으로 다운로드하세요: @@ -71,23 +56,17 @@ Cosmos-Predict2는 텍스트 기반 이미지 생성(Text2Image)과 비디오- **확산 모델** - - cosmos_predict2_2B_video2world_480p_16fps.safetensors - +- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) 기타 가중치는 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged)에서 다운로드하세요. **텍스트 인코더** - - oldt5_xxl_fp8_e4m3fn_scaled.safetensors - +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** - - wan_2.1_vae.safetensors - +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) 파일 저장 위치 ``` @@ -114,4 +93,4 @@ Cosmos-Predict2는 텍스트 기반 이미지 생성(Text2Image)과 비디오- 6. (선택사항) `ClipTextEncode` 노드에서 프롬프트를 수정할 수 있습니다. 7. (선택사항) `CosmosPredict2ImageToVideoLatent` 노드에서 크기와 프레임 수를 조정합니다. 8. `Run` 버튼을 클릭하거나 `Ctrl(cmd) + Enter` 단축키를 눌러 워크플로우를 실행합니다. -9. 생성이 완료되면 비디오가 자동으로 `ComfyUI/output/` 디렉토리에 저장되며, `save video` 노드에서 미리볼 수 있습니다. \ No newline at end of file +9. 생성이 완료되면 비디오가 자동으로 `ComfyUI/output/` 디렉토리에 저장되며, `save video` 노드에서 미리볼 수 있습니다. */} \ No newline at end of file diff --git a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 8dd2bba11..414f7b0d2 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -29,17 +29,17 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 모델 링크 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **디퓨전 모델** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **VAE** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) 모델 저장 위치 diff --git a/ko/tutorials/video/hunyuan/hunyuan-video.mdx b/ko/tutorials/video/hunyuan/hunyuan-video.mdx index cb53c37e7..16902e496 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video.mdx @@ -46,9 +46,9 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델들은 텍스트-투-비디오와 이미지-투-비디오 워크플로우 모두에 사용됩니다. 아래 모델들을 다운로드하여 지정된 디렉토리에 저장해주세요: -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) +- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) 저장 위치: @@ -73,7 +73,7 @@ ComfyUI/ ### 2. 수동 모델 설치 -[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models` 폴더에 저장하세요. +[hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models` 폴더에 저장하세요. 다음 모델 파일들이 올바른 위치에 있는지 확인하세요: @@ -126,7 +126,7 @@ ComfyUI/ ### v1과 v2 버전 공통 모델 다음 파일을 다운로드해 `ComfyUI/models/clip_vision` 디렉토리에 저장하세요: -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) +- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) ### v1 "concat" 이미지-투-비디오 워크플로우 @@ -140,7 +140,7 @@ ComfyUI/ #### 2. 관련 모델 수동 설치 -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) 다음 모델 파일들이 올바른 위치에 있는지 확인하세요: @@ -185,7 +185,7 @@ v2 워크플로우는 기본적으로 v1 워크플로우와 동일합니다. 다 #### 2. 관련 모델 수동 설치 -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) 다음 모델 파일들이 올바른 위치에 있는지 확인하세요: diff --git a/ko/tutorials/video/kandinsky/kandinsky-5.mdx b/ko/tutorials/video/kandinsky/kandinsky-5.mdx index 4730e9e55..f75218eb3 100644 --- a/ko/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/ko/tutorials/video/kandinsky/kandinsky-5.mdx @@ -51,39 +51,21 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ComfyUI를 최신 버전으로 업데이트해 주시고, 메뉴 `워크플로우` -> `템플릿 탐색` -> `비디오`를 통해 "칸딘스키 5.0 T2V"를 찾아 워크플로우를 로드해 주세요. - - - T2V 워크플로우를 다운로드하여 로컬에서 사용 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로우 파일 다운로드

+ ### 2. 모델 수동 다운로드 **텍스트 인코더** - - - - Qwen2.5-VL 7B 텍스트 인코더 (FP8) - - - CLIP-L 텍스트 인코더 - - +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) **확산 모델** - - - Kandinsky 5.0 T2V Lite SFT 확산 모델 (5s) - +- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) **VAE** - - - HunyuanVideo 3D VAE - +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) ``` ComfyUI/ @@ -103,39 +85,21 @@ ComfyUI/ ComfyUI를 최신 버전으로 업데이트해 주시고, 메뉴 `워크플로우` -> `템플릿 탐색` -> `비디오`를 통해 "칸딘스키 5.0 I2V"를 찾아 워크플로우를 로드해 주세요. - - - I2V 워크플로우를 다운로드하여 로컬에서 사용 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로우 파일 다운로드

+ ### 2. 모델 수동 다운로드 **텍스트 인코더** - - - - Qwen2.5-VL 7B 텍스트 인코더 (FP8) - - - CLIP-L 텍스트 인코더 - - +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) **확산 모델** - - - Kandinsky 5.0 I2V Lite 확산 모델 (5s) - +- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) **VAE** - - - HunyuanVideo 3D VAE - +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/ltxv.mdx b/ko/tutorials/video/ltxv.mdx index 6f361c5d4..67608d2eb 100644 --- a/ko/tutorials/video/ltxv.mdx +++ b/ko/tutorials/video/ltxv.mdx @@ -26,17 +26,9 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 첫 번째 [프레임 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)를 통해 비디오를 제어할 수 있습니다. - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "LTX-Video" 검색 - - - 이 워크플로우의 예제 입력 이미지 가져오기 - - + +

Comfy Cloud에서 실행하기

+ LTX-Video 이미지에서 비디오로 @@ -46,15 +38,6 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 텍스트에서 비디오로 - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "LTX-Video" 검색 - - - LTX-Video 텍스트에서 비디오로 @@ -65,13 +48,8 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델을 다운로드하여 아래 지정된 위치에 배치하세요: - - 다운로드하여 ComfyUI/models/checkpoints/에 배치 - - - - 다운로드하여 ComfyUI/models/text_encoders/에 배치 - +- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true) ``` ├── checkpoints/ diff --git a/ko/tutorials/video/wan/fun-camera.mdx b/ko/tutorials/video/wan/fun-camera.mdx index fa8d854f9..d4be5c366 100644 --- a/ko/tutorials/video/wan/fun-camera.mdx +++ b/ko/tutorials/video/wan/fun-camera.mdx @@ -13,6 +13,7 @@ translationBlockHashes: "Performance Reference": 32425486 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Wan2.1 Fun Camera 소개 @@ -35,49 +36,21 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 다음 모델들은 모두 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged)에서 찾을 수 있습니다. -### 디퓨전 모델 - -1.3B 또는 14B 중 하나를 선택하세요: - - - - Wan2.1 Fun Camera 1.3B 디퓨전 모델 - - - Wan2.1 Fun Camera 14B 디퓨전 모델 - - +**디퓨전 모델**은 1.3B 또는 14B 중 하나를 선택하세요: +- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) +- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) 이전에 Wan2.1 관련 모델을 사용한 적이 있다면 이미 다음 모델들이 있을 것입니다. 아직 없다면 다운로드해 주세요: -### 텍스트 인코더 - -하나를 선택하세요: - - - - 전체 정밀도 텍스트 인코더 - - - FP8 양자화 텍스트 인코더 (낮은 VRAM 권장) - - - -### VAE +**텍스트 인코더**는 하나를 선택하세요: +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - - Wan2.1 VAE 모델 - - +**VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -### CLIP 비전 - - - - CLIP 비전 인코더 - - +**CLIP 비전** +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) 파일 저장 위치: @@ -97,16 +70,9 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 1.3B 기본 워크플로우 예시 -### 1. 워크플로우 다운로드 +### 1. 워크플로우 관련 파일 다운로드 - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan 2.1 Fun Camera 1.3B" 검색 - - +#### 1.1 워크플로우 파일 아래 동영상을 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요: @@ -116,19 +82,21 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > + +

Json 워크플로우 파일 다운로드

+ + 14B 버전을 사용하고 싶다면 모델 파일을 14B 버전으로 교체하면 되지만, VRAM 요구 사항에 유의해 주세요. -### 2. 입력 자료 다운로드 +#### 1.2 입력 이미지 다운로드 + +아래 이미지를 다운로드해 시작 프레임으로 사용하세요: - - - 아래 이미지를 다운로드하여 1.3B 워크플로우의 시작 프레임으로 사용하세요 - - +![입력 참조 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) -### 3. 워크플로우 단계별 완료하기 +### 2. 워크플로우 단계별 완료하기 ![Wan2.1 Fun Camera 워크플로우 단계](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -145,30 +113,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 14B 워크플로우 및 입력 이미지 -### 1. 워크플로우 다운로드 - - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan 2.1 Fun Camera 14B" 검색 - - - -### 2. 입력 자료 다운로드 + +

Json 워크플로우 파일 다운로드

+ - - - 아래 이미지를 다운로드하여 14B 워크플로우의 시작 프레임으로 사용하세요 - - +**입력 이미지** +![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) ## 성능 참고사항 diff --git a/ko/tutorials/video/wan/fun-control.mdx b/ko/tutorials/video/wan/fun-control.mdx index a2467675e..9cb0cc3d9 100644 --- a/ko/tutorials/video/wan/fun-control.mdx +++ b/ko/tutorials/video/wan/fun-control.mdx @@ -52,18 +52,18 @@ ComfyUI는 현재 Wan2.1 Fun Control 모델을 **네이티브로 지원**합니 해당 링크를 클릭해 다운로드하세요. 이전에 Wan 관련 워크플로우를 사용한 적이 있다면 **Diffusion 모델**만 다운로드하면 됩니다. **Diffusion 모델** - 1.3B 또는 14B 중 선택하세요. 14B 버전은 파일 크기가 더 크고 (32GB), VRAM 요구 사항도 높습니다: -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-Control.safetensors`로 이름 변경 **텍스트 인코더** - 다음 모델 중 하나를 선택하세요 (fp16 정밀도는 파일 크기가 더 크고 성능 요구 사항도 높습니다): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/video/wan/fun-inp.mdx b/ko/tutorials/video/wan/fun-inp.mdx index 6bfd4ff40..dc9d11849 100644 --- a/ko/tutorials/video/wan/fun-inp.mdx +++ b/ko/tutorials/video/wan/fun-inp.mdx @@ -53,18 +53,18 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델들은 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 및 [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334)에서 찾을 수 있습니다. **디퓨전 모델** - 1.3B 또는 14B를 선택하세요. 14B 버전은 파일 크기가 더 크고(32GB), VRAM 요구 사항도 높습니다: -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-InP.safetensors`로 이름 변경 +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-InP.safetensors`로 이름 변경 **텍스트 인코더** - 다음 모델 중 하나를 선택하세요 (fp16 정밀도는 크기가 더 크고 성능 요구 사항도 높습니다): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/video/wan/vace.mdx b/ko/tutorials/video/wan/vace.mdx index 4b8f0aed2..7069223b3 100644 --- a/ko/tutorials/video/wan/vace.mdx +++ b/ko/tutorials/video/wan/vace.mdx @@ -59,18 +59,18 @@ VACE 14B는 알리바바 통이 완샹팀이 출시한 오픈소스 통합 비 ### 모델 다운로드 **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 이전에 Wan Video 관련 워크플로를 사용하셨다면 이미 다음 모델 파일을 다운로드하셨을 것입니다. **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **텍스트 인코더** 중 하나를 선택해 다운로드하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/wan/wan-ati.mdx b/ko/tutorials/video/wan/wan-ati.mdx index c8d1aee32..345de9b22 100644 --- a/ko/tutorials/video/wan/wan-ati.mdx +++ b/ko/tutorials/video/wan/wan-ati.mdx @@ -47,18 +47,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 워크플로우에서 모델 파일을 성공적으로 다운로드하지 못했다면, 아래 링크를 이용해 수동으로 다운로드해 보세요. **디퓨전 모델** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **텍스트 인코더** 다음 모델 중 하나를 선택하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/wan/wan-causal-forcing.mdx b/ko/tutorials/video/wan/wan-causal-forcing.mdx index c6231a10a..7f90de227 100644 --- a/ko/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ko/tutorials/video/wan/wan-causal-forcing.mdx @@ -83,10 +83,10 @@ Wan2.1 I2V 모델과 필요한 파일을 다운로드합니다. 해당 `models/` ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B 체크포인트 - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B 체크포인트 (최소 8GB VRAM) @@ -94,10 +94,10 @@ Wan2.1 I2V 모델과 필요한 파일을 다운로드합니다. 해당 `models/` ### CLIP 및 VAE - + google-bert/bert-base-uncased — CLIP 텍스트 인코더 - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/ko/tutorials/video/wan/wan-dancer.mdx b/ko/tutorials/video/wan/wan-dancer.mdx index 2f90e471a..87022fb2d 100644 --- a/ko/tutorials/video/wan/wan-dancer.mdx +++ b/ko/tutorials/video/wan/wan-dancer.mdx @@ -61,20 +61,20 @@ ComfyUI를 최신 버전으로 업데이트한 후 워크플로 파일을 다운 ### 3. 모델 수동 다운로드 **확산 모델** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **텍스트 인코더** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan-flf.mdx b/ko/tutorials/video/wan/wan-flf.mdx index 97b624eea..825284aa2 100644 --- a/ko/tutorials/video/wan/wan-flf.mdx +++ b/ko/tutorials/video/wan/wan-flf.mdx @@ -58,7 +58,7 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 이 가이드에 포함된 모든 모델은 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)에서 확인할 수 있습니다. **diffusion_models** 하드웨어 환경에 따라 버전을 선택하세요. -- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8: [wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -66,14 +66,14 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 **Text encoders**에서 버전을 하나 선택해 다운로드하세요. -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치 diff --git a/ko/tutorials/video/wan/wan-move.mdx b/ko/tutorials/video/wan/wan-move.mdx index 061f43ccb..b767247e3 100644 --- a/ko/tutorials/video/wan/wan-move.mdx +++ b/ko/tutorials/video/wan/wan-move.mdx @@ -28,37 +28,37 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Wan-Move 이미지에서 비디오 워크플로우 - - 워크플로우 다운로드 - + +

JSON 워크플로우 파일 다운로드

+ - - 클라우드에서 열기 - + +

ComfyUI 클라우드에서 실행하기

+ ## 모델 링크 - - **텍스트 인코더** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors - +**텍스트 인코더** - - **클립 비전** -- clip_vision_h.safetensors - +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - **로라** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors - +**클립 비전** - - **디퓨전 모델** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors - +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) - - **VAE** -- wan_2.1_vae.safetensors - +**로라** + +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) + +**디퓨전 모델** + +- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) + +**VAE** + +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/video/wan/wan-video.mdx b/ko/tutorials/video/wan/wan-video.mdx index 13d23d236..657814f1b 100644 --- a/ko/tutorials/video/wan/wan-video.mdx +++ b/ko/tutorials/video/wan/wan-video.mdx @@ -41,14 +41,14 @@ Wan2.1 Video 시리즈는 2025년 2월 알리바바가 [Apache 2.0 라이선스] 이 가이드에서 언급된 모든 모델은 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)에서 확인할 수 있습니다. 아래는 이 가이드의 예시에 필요한 일반적인 모델들로, 미리 다운로드해 두시면 됩니다: **텍스트 인코더**에서 하나의 버전을 선택해 다운로드하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` @@ -70,7 +70,7 @@ ComfyUI/ ## Wan2.1 텍스트 투 비디오 워크플로우 -워크플로우를 시작하기 전에 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +워크플로우를 시작하기 전에 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. > 다른 t2v 정밀도 버전이 필요하시다면 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models)를 방문해 다운로드해 주세요. @@ -108,7 +108,7 @@ ComfyUI/ #### 2. 모델 다운로드 -[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. #### 3. 워크플로우 단계별 완료 @@ -136,7 +136,7 @@ ComfyUI/ #### 2. 모델 다운로드 -[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. #### 3. 워크플로우 단계별 완료 diff --git a/ko/tutorials/video/wan/wan2-2-animate.mdx b/ko/tutorials/video/wan/wan2-2-animate.mdx index 3b7f9a838..9155a436e 100644 --- a/ko/tutorials/video/wan/wan2-2-animate.mdx +++ b/ko/tutorials/video/wan/wan2-2-animate.mdx @@ -55,14 +55,13 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 다음 워크플로우 파일을 다운로드해 ComfyUI로 끌어다 놓으면 워크플로우가 로드됩니다. - - - Comfy Cloud에서 실행 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Animate" 검색 - - + +

JSON 워크플로우 다운로드

+ + + +

Comfy Cloud에서 실행

+ 아래 자료를 입력으로 다운로드하세요: @@ -78,40 +77,21 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 ### 2. 모델 링크 -**Diffusion Models** - - - - Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: Kijai의 리포지토리에서 제공하는 스케일링 FP8 모델 - - - wan2.2_animate_14B_bf16.safetensors: 원본 bf16 모델 가중치 - - - -**CLIP Vision** - - - clip_vision_h.safetensors: CLIP Vision 인코더 - - -**LoRAs** - - - lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4단계 가속 LoRA - +**diffusion_models** +- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) Kijai의 리포지토리에서 가져온 모델입니다. +- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 원본 모델 가중치 -**VAE** +**clip_visions** +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) - - wan_2.1_vae.safetensors: 인코딩 및 디코딩용 Wan2.1 VAE - +**loras** +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 4단계 가속 Lora입니다. -**Text Encoders** +**vae** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - umt5_xxl_fp8_e4m3fn_scaled.safetensors: 스케일링 FP8 텍스트 인코더 - +**text_encoders** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-fun-camera.mdx b/ko/tutorials/video/wan/wan2-2-fun-camera.mdx index 7645aa374..ae9e27c58 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -46,60 +46,31 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Camera" 검색 - - + +

JSON 워크플로우 다운로드

+ 아래 이미지를 다운로드해 입력으로 사용하세요. - - - 동영상 생성을 위한 시작 프레임입니다. 이 이미지를 다운로드하여 사용하거나, 자신의 이미지로 교체하세요. - - +![입력 시작 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) ### 2. 모델 링크 다음 모델들은 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인할 수 있습니다: **디퓨전 모델** - - - - Wan2.2 Fun Camera 고노이즈 디퓨전 모델 - - - Wan2.2 Fun Camera 저노이즈 디퓨전 모델 - - +- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA (선택사항, 가속화용)** - - - - 고노이즈 모델용 4단계 가속 LoRA - - - 저노이즈 모델용 4단계 가속 LoRA - - +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - 인코딩/디코딩용 Wan2.1 VAE - - -**텍스트 인코더** - - - FP8 스케일링 텍스트 인코더 - +**텍스트 인코더** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) 파일 저장 위치 @@ -141,4 +112,4 @@ ComfyUI/ - **너비/높이**: 동영상 해상도 설정 - **길이**: 동영상 프레임 수 설정(기본값은 81프레임) - **속도**: 동영상 속도 설정(기본값은 1.0) -8. `Run` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 동영상 생성을 실행하세요. +8. `Run` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 동영상 생성을 실행하세요. \ No newline at end of file diff --git a/ko/tutorials/video/wan/wan2-2-fun-control.mdx b/ko/tutorials/video/wan/wan2-2-fun-control.mdx index 416a9b56e..b60dbd1cc 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-control.mdx @@ -55,27 +55,27 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ### 1. 워크플로우 및 자료 다운로드 -ComfyUI를 최신 버전으로 업데이트한 후 워크플로우 파일을 다운로드하여 ComfyUI로 드래그하거나, 템플릿 라이브러리의 `Workflow` → `Browse Templates` → `Video`에서 "Wan2.2 Fun Control"을 찾아보세요. +아래 비디오 또는 JSON 파일을 다운로드해 ComfyUI로 드래그하여 워크플로우를 로드하세요. - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Control" 검색 - - + + + +

JSON 워크플로우 다운로드

+ 다음 이미지와 비디오를 입력 자료로 다운로드해 주세요. - - - 비디오 생성을 위한 시작 프레임입니다. 이 이미지를 다운로드하여 사용하거나, 자신의 이미지로 교체하세요. - - - 전처리된 포즈 제어 비디오입니다. 이 비디오를 다운로드하여 사용하거나, 자신의 비디오로 교체하세요. - - +![입력 시작 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) + + > 여기서는 사전 처리된 비디오를 사용합니다. @@ -83,39 +83,19 @@ ComfyUI를 최신 버전으로 업데이트한 후 워크플로우 파일을 다 아래 모델들은 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인하실 수 있습니다. -**Diffusion Models** - - - - wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors：고노이즈 확산 모델 - - - wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors：저노이즈 확산 모델 - - +**디퓨전 모델** +- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA (선택사항, 가속화용)** - - - - wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors：고노이즈 4단계 가속 LoRA - - - wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors：저노이즈 4단계 가속 LoRA - - +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - wan_2.1_vae.safetensors：인코딩/디코딩용 Wan2.1 VAE - - -**Text Encoder** - - - umt5_xxl_fp8_e4m3fn_scaled.safetensors：스케일링 FP8 텍스트 인코더 - +**텍스트 인코더** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx index 492c6ef28..6a076123b 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -61,14 +61,13 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> 또는 ComfyUI를 최신 버전으로 업데이트한 후 아래 워크플로우를 다운로드해 ComfyUI에 드래그하여 로드하세요. - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Inp" 검색 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로우 다운로드

+ + + +

Comfy Cloud에서 실행

+ 다음 자료를 시작 및 끝 프레임으로 사용하세요. @@ -78,38 +77,18 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> ### 2. 모델 **디퓨전 모델** - - - - wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: 시작-끝 프레임 인페인팅용 고노이즈 확산 모델 - - - wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: 시작-끝 프레임 인페인팅용 저노이즈 확산 모델 - - +- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) **Lightning LoRA (선택사항, 가속화용)** - - - - wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 고노이즈 모델용 4단계 가속 LoRA - - - wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 저노이즈 모델용 4단계 가속 LoRA - - +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - wan_2.1_vae.safetensors: 인코딩 및 디코딩용 Wan2.1 VAE - - -**텍스트 인코더** - - - umt5_xxl_fp8_e4m3fn_scaled.safetensors: 스케일링된 FP8 텍스트 인코더 - +**텍스트 인코더** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-s2v.mdx b/ko/tutorials/video/wan/wan2-2-s2v.mdx index 41cb67f32..bd4747e65 100644 --- a/ko/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ko/tutorials/video/wan/wan2-2-s2v.mdx @@ -35,58 +35,38 @@ Wan2.2 S2V 모델: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - - - Comfy Cloud에서 열기 - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 S2V" 검색 - - + +

JSON 워크플로우 다운로드

+ + + +

Comfy Cloud에서 실행

+ 다음 이미지와 오디오를 입력으로 다운로드하세요: +![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) + - - - 기본 입력 이미지를 다운로드하거나, 자신의 이미지를 사용하세요. - - - 기본 입력 오디오를 다운로드하거나, 자신의 오디오를 사용하세요. - - + +

입력 오디오 다운로드

+ ### 2. 모델 링크 모델들은 [우리 리포지토리](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인하실 수 있습니다. -**diffusion_models** - - - - FP8 scaled diffusion model. ComfyUI/models/diffusion_models/에 배치 - - - BF16 diffusion model. ComfyUI/models/diffusion_models/에 배치 - - +**diffusion_models** +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) **audio_encoders** - - - Audio encoder model. ComfyUI/models/audio_encoders/에 배치 - +- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) **vae** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - Wan2.1 VAE model. ComfyUI/models/vae/에 배치 - - -**text_encoders** - - - FP8 scaled text encoder. ComfyUI/models/text_encoders/에 배치 - +**text_encoders** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -114,14 +94,8 @@ ComfyUI/ 두 모델 모두 [여기](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models)에서 확인하실 수 있습니다: - - - FP8 scaled diffusion model - - - BF16 diffusion model - - +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) 이 템플릿에서는 `wan2.2_s2v_14B_fp8_scaled.safetensors`를 사용하며, 이 모델은 더 적은 VRAM을 필요로 합니다. 하지만 품질 저하를 줄이기 위해 `wan2.2_s2v_14B_bf16.safetensors`를 시도해볼 수도 있습니다. diff --git a/ko/tutorials/video/wan/wan2_2.mdx b/ko/tutorials/video/wan/wan2_2.mdx index ea1a1d460..1303255f4 100644 --- a/ko/tutorials/video/wan/wan2_2.mdx +++ b/ko/tutorials/video/wan/wan2_2.mdx @@ -101,25 +101,24 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - - - JSON 워크플로우 파일을 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 5B" 검색 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로우 파일 다운로드

+ + + +

Comfy Cloud에서 실행

+ ### 2. 모델 수동 다운로드 **디퓨전 모델** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) +- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan2.2_vae.safetensors) +- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -157,26 +156,25 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - - - JSON 워크플로우 파일을 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 14B T2V" 검색 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로우 파일 다운로드

+ + + +

Comfy Cloud에서 실행

+ ### 2. 모델 수동 다운로드 **디퓨전 모델** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -215,14 +213,13 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - - - JSON 워크플로우 파일을 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 14B I2V" 검색 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로 파일 다운로드

+ + + +

Comfy Cloud에서 실행

+ 다음 이미지를 입력으로 사용할 수 있습니다: ![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -230,14 +227,14 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` ### 2. 모델 수동 다운로드 **디퓨전 모델** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) +- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -275,14 +272,13 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - - - JSON 워크플로우를 다운로드하거나 템플릿 라이브러리에서 "Wan2.2 14B FLF2V" 검색 - - - Comfy Cloud에서 열기 - - + +

JSON 워크플로 다운로드

+ + + +

Comfy Cloud에서 실행

+ 다음 이미지를 입력 자료로 다운로드하세요: diff --git a/ko/tutorials/video/zai/scail2.mdx b/ko/tutorials/video/zai/scail2.mdx index 3edc8d8b7..61c6af2ff 100644 --- a/ko/tutorials/video/zai/scail2.mdx +++ b/ko/tutorials/video/zai/scail2.mdx @@ -105,23 +105,23 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ### 필수 모델 **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders** (하나 선택) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### 파일 저장 위치 diff --git a/tutorials/flux/flux-1-kontext-dev.mdx b/tutorials/flux/flux-1-kontext-dev.mdx index 52c562850..96d35ba6e 100644 --- a/tutorials/flux/flux-1-kontext-dev.mdx +++ b/tutorials/flux/flux-1-kontext-dev.mdx @@ -29,9 +29,9 @@ While the previously released API version offers the highest fidelity and speed, ### Version Information -- **[FLUX.1 Kontext [pro]** — Commercial version, focused on rapid iterative editing -- **FLUX.1 Kontext [max]** — Experimental version with stronger prompt adherence -- **FLUX.1 Kontext [dev]** — Open source version (used in this tutorial), 12B parameters, mainly for research +- **[FLUX.1 Kontext [pro]** : Commercial version, focused on rapid iterative editing +- **FLUX.1 Kontext [max]** : Experimental version with stronger prompt adherence +- **FLUX.1 Kontext [dev]** : Open source version (used in this tutorial), 12B parameters, mainly for research Currently in ComfyUI, you can use all these versions, where [Pro and Max versions](/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) can be called through Partner Nodes, while the Dev open source version please refer to the instructions in this guide. diff --git a/tutorials/video/wan/fun-inp.mdx b/tutorials/video/wan/fun-inp.mdx index d0770f209..3f09d0d5e 100644 --- a/tutorials/video/wan/fun-inp.mdx +++ b/tutorials/video/wan/fun-inp.mdx @@ -62,7 +62,7 @@ If automatic model downloading is ineffective, please download the models manual All models involved in this guide can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) and [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334). -**Diffusion Models** — Choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: +**Diffusion Models** : Choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: @@ -73,7 +73,7 @@ All models involved in this guide can be found at [Wan_2.1_ComfyUI_repackaged](h -**Text Encoders** — Choose one of the following models (fp16 precision has a larger size and higher performance requirements): +**Text Encoders** : Choose one of the following models (fp16 precision has a larger size and higher performance requirements): diff --git a/tutorials/video/wan/wan-alpha.mdx b/tutorials/video/wan/wan-alpha.mdx index b10b24783..36a09915b 100644 --- a/tutorials/video/wan/wan-alpha.mdx +++ b/tutorials/video/wan/wan-alpha.mdx @@ -52,7 +52,7 @@ Update your ComfyUI to the latest version, then download and drag the workflow f All models mentioned can be found at [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged). -**Diffusion Models** — Choose one version: +**Diffusion Models** : Choose one version: diff --git a/tutorials/video/wan/wan-ati.mdx b/tutorials/video/wan/wan-ati.mdx index 333de9cd6..6ff3e114d 100644 --- a/tutorials/video/wan/wan-ati.mdx +++ b/tutorials/video/wan/wan-ati.mdx @@ -67,7 +67,7 @@ We will use the following image as input: All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files). -**Diffusion Model** — Choose one version: +**Diffusion Model** : Choose one version: @@ -75,7 +75,7 @@ All models involved in this guide can be found [here](https://huggingface.co/Com -**Text Encoders** — Choose one version: +**Text Encoders** : Choose one version: diff --git a/tutorials/video/wan/wan-flf.mdx b/tutorials/video/wan/wan-flf.mdx index c7305710d..eaea39267 100644 --- a/tutorials/video/wan/wan-flf.mdx +++ b/tutorials/video/wan/wan-flf.mdx @@ -65,14 +65,14 @@ Update your ComfyUI to the latest version, then download and drag the workflow f All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files). -**Diffusion Models** — Choose one version based on your hardware +**Diffusion Models** : Choose one version based on your hardware - wan2.1_flf2v_720p_14B_fp16.safetensors — Full precision, requires more VRAM. Place in ComfyUI/models/diffusion_models/ + wan2.1_flf2v_720p_14B_fp16.safetensors : Full precision, requires more VRAM. Place in ComfyUI/models/diffusion_models/ - wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors — Quantized version, lower VRAM usage. Place in ComfyUI/models/diffusion_models/ + wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors : Quantized version, lower VRAM usage. Place in ComfyUI/models/diffusion_models/ @@ -80,27 +80,27 @@ All models involved in this guide can be found [here](https://huggingface.co/Com If you have previously tried Wan Video related workflows, you may already have the following files. -**Text Encoders** — Choose one version +**Text Encoders** : Choose one version - umt5_xxl_fp16.safetensors — Full precision text encoder. Place in ComfyUI/models/text_encoders/ + umt5_xxl_fp16.safetensors : Full precision text encoder. Place in ComfyUI/models/text_encoders/ - umt5_xxl_fp8_e4m3fn_scaled.safetensors — Quantized text encoder. Place in ComfyUI/models/text_encoders/ + umt5_xxl_fp8_e4m3fn_scaled.safetensors : Quantized text encoder. Place in ComfyUI/models/text_encoders/ **VAE** - wan_2.1_vae.safetensors — Wan2.1 VAE for encoding/decoding. Place in ComfyUI/models/vae/ + wan_2.1_vae.safetensors : Wan2.1 VAE for encoding/decoding. Place in ComfyUI/models/vae/ **CLIP Vision** - clip_vision_h.safetensors — CLIP Vision encoder. Place in ComfyUI/models/clip_vision/ + clip_vision_h.safetensors : CLIP Vision encoder. Place in ComfyUI/models/clip_vision/ File Storage Location diff --git a/zh/tutorials/3d/hunyuan3D-2.mdx b/zh/tutorials/3d/hunyuan3D-2.mdx index ac0d5f809..8e581a080 100644 --- a/zh/tutorials/3d/hunyuan3D-2.mdx +++ b/zh/tutorials/3d/hunyuan3D-2.mdx @@ -58,14 +58,9 @@ Hunyuan3D-2mv 工作流中，我们将使用多视角的图片来生成3D模型 - - - 在 Comfy Cloud 上立即运行此工作流 - - - 下载工作流 JSON 文件 - - + +

Run on Comfy Cloud

+ ### 1. 工作流 @@ -88,7 +83,7 @@ Hunyuan3D-2mv 工作流中，我们将使用多视角的图片来生成3D模型 下载下面的模型，并保存到对应的 ComfyUI 文件夹 -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv.safetensors` +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv.safetensors` ``` ComfyUI/ @@ -111,14 +106,9 @@ ComfyUI/ Hunyuan3D-2mv-turbo 工作流中，我们将使用 Hunyuan3D-2mv-turbo 模型来生成3D模型，这个模型是 Hunyuan3D-2mv 的分步蒸馏（Step Distillation）版本，可以更快地生成3D模型，在这个版本的工作流中我们设置 `cfg` 为 1.0 并添加 `flux guidance` 节点来控制 `distilled cfg` 的生成。 - - - 在 Comfy Cloud 上立即运行此工作流 - - - 下载工作流 JSON 文件 - - + +

Run on Comfy Cloud

+ ### 1. 工作流 @@ -135,7 +125,7 @@ Hunyuan3D-2mv-turbo 工作流中，我们将使用 Hunyuan3D-2mv-turbo 模型来 下载下面的模型，并保存到对应的 ComfyUI 文件夹 -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv-turbo.safetensors` +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv-turbo.safetensors` ``` ComfyUI/ @@ -156,14 +146,9 @@ ComfyUI/ Hunyuan3D-2 工作流中，我们将使用 Hunyuan3D-2 模型来生成3D模型，这个模型不是一个多视角的模型，在这个工作流中，我们使用`Hunyuan3Dv2Conditioning` 节点替换掉 `Hunyuan3Dv2ConditioningMultiView` 节点。 - - - 在 Comfy Cloud 上立即运行此工作流 - - - 下载工作流 JSON 文件 - - + +

Run on Comfy Cloud

+ ### 1. 工作流 @@ -179,7 +164,7 @@ Hunyuan3D-2 工作流中，我们将使用 Hunyuan3D-2 模型来生成3D模型 下载下面的模型，并保存到对应的 ComfyUI 文件夹 -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2.safetensors` +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2.safetensors` ``` ComfyUI/ diff --git a/zh/tutorials/3d/triposplat.mdx b/zh/tutorials/3d/triposplat.mdx index 254287c1b..634c07066 100644 --- a/zh/tutorials/3d/triposplat.mdx +++ b/zh/tutorials/3d/triposplat.mdx @@ -106,23 +106,23 @@ TripoSplat 使用 **前馈架构**，接收单张 RGB 图像并直接预测一 下载 TripoSplat 模型及所需文件。放入对应的 `models/` 子目录。 - + triposplat_fp16.safetensors — TripoSplat 扩散模型检查点 - + triposplat_vae_decoder_fp16.safetensors — VAE 解码器 - + flux2-vae.safetensors — Flux.2 VAE，用于潜空间编码 - + dino_v3_vit_h.safetensors — CLIP 视觉编码器（DINOv2） - + birefnet.safetensors — 用于预处理的背景去除模型 diff --git a/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx b/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx index 0b2ef80ea..266b852e4 100644 --- a/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -45,7 +45,7 @@ AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和 ### AIO 模型下载 - + 一体化 checkpoint 文件（推荐大多数用户使用）。 @@ -74,19 +74,19 @@ AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和 ### 分离模型下载 - + 扩散模型。 - + 文本编码器 (0.6B)。 - + 文本编码器 (1.7B)。 - + VAE 模型。 diff --git a/zh/tutorials/audio/ace-step/ace-step-v1.mdx b/zh/tutorials/audio/ace-step/ace-step-v1.mdx index 4844eee60..8e87ef6a6 100644 --- a/zh/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/zh/tutorials/audio/ace-step/ace-step-v1.mdx @@ -33,15 +33,9 @@ ACE-Step 作为一个强大的音乐生成基座，提供了丰富的扩展能 点击下面的按钮下载对应的工作流文件，拖入 ComfyUI 中即可加载对应的工作流信息，对应工作流已包含模型下载信息。 - - - 下载 Json 格式工作流文件 - - + +

下载 Json 格式工作流文件

+ 你也可以手动下载[ace_step_v1_3.5b.safetensors](https://huggingface.co/Comfy-Org/ACE-Step_ComfyUI_repackaged/blob/main/all_in_one/ace_step_v1_3.5b.safetensors) 后保存到 `ComfyUI/models/checkpoints` 文件夹下 @@ -66,28 +60,16 @@ ACE-Step 作为一个强大的音乐生成基座，提供了丰富的扩展能 点击下面的按钮下载对应的工作流文件，拖入 ComfyUI 中即可加载对应的工作流信息 - - - 下载 Json 格式工作流文件 - - + +

下载 Json 格式工作流文件

+ 下载下面的音频作为输入音频 - - - 下载示例音频文件用于输入 - - + +

下载示例音频文件用于输入

+ ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/audio/stable-audio/stable-audio-1.mdx b/zh/tutorials/audio/stable-audio/stable-audio-1.mdx index e1487cdfe..29e124b64 100644 --- a/zh/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/zh/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -41,7 +41,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" ### 检查点 - + 2.3GB。放入 models/checkpoints/ @@ -56,7 +56,7 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" ### 文本编码器 - + 提示词处理的文本编码器。放入 models/text_encoders/ diff --git a/zh/tutorials/audio/stable-audio/stable-audio-3.mdx b/zh/tutorials/audio/stable-audio/stable-audio-3.mdx index 03969d679..114dc7b44 100644 --- a/zh/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/zh/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -77,11 +77,11 @@ Stable Audio 3 提供三个变体： ### 检查点 - + 用于 Medium 工作流。放入 models/checkpoints/ - + 用于 Medium Base 工作流。放入 models/checkpoints/ @@ -97,11 +97,11 @@ Stable Audio 3 提供三个变体： ### 文本编码器 - + 所有 Stable Audio 3 工作流都需要。放入 models/text_encoders/ - + Medium 工作流需要（Qwen 重新提示）。放入 models/text_encoders/ diff --git a/zh/tutorials/basic/inpaint.mdx b/zh/tutorials/basic/inpaint.mdx index 18fd9fa29..f109e95f1 100644 --- a/zh/tutorials/basic/inpaint.mdx +++ b/zh/tutorials/basic/inpaint.mdx @@ -35,7 +35,7 @@ translationBlockHashes: #### 1. 模型安装 下载下面的模型文件，并保存到`ComfyUI/models/checkpoints`目录下 -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) #### 2. 局部重绘素材 @@ -64,7 +64,7 @@ translationBlockHashes: ![ComfyUI 局部重绘工作流 - SD1.5](/images/tutorial/basic/inpaint/inpaint_sd1.5_pruned_emaonly.png) -你会发现 [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) 模型生成的结果局部重绘的效果更好过渡更自然。 +你会发现 [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) 模型生成的结果局部重绘的效果更好过渡更自然。 这因为这个模型是专为 inpainting 设计的模型，它可以帮助我们更好地控制生成区域，从而获得更好的局部重绘效果。 记得我们一直用的比喻吗？不同的模型就像能力不同的画家一样，但每个画家都有自己能力的上限，选择合适的模型可以让你的生成效果更好。 diff --git a/zh/tutorials/basic/outpaint.mdx b/zh/tutorials/basic/outpaint.mdx index 34733f562..1cff771a3 100644 --- a/zh/tutorials/basic/outpaint.mdx +++ b/zh/tutorials/basic/outpaint.mdx @@ -38,7 +38,7 @@ import InstallationModels from '/snippets/zh/tutorials/basic/installation-models #### 1. 模型安装 -- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/blob/main/512-inpainting-ema.safetensors) +- [512-inpainting-ema.safetensors](https://huggingface.co/Comfy-Org/stable_diffusion_2.1_repackaged/resolve/main/512-inpainting-ema.safetensors) #### 2. 输入图片 diff --git a/zh/tutorials/controlnet/controlnet.mdx b/zh/tutorials/controlnet/controlnet.mdx index 813295c8d..753c45c72 100644 --- a/zh/tutorials/controlnet/controlnet.mdx +++ b/zh/tutorials/controlnet/controlnet.mdx @@ -74,8 +74,8 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [dreamCreationVirtual3DECommerce_v10.safetensors](https://civitai.com/api/download/models/731340?type=Model&format=SafeTensor&size=full&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/depth-controlnet.mdx b/zh/tutorials/controlnet/depth-controlnet.mdx index 92d6456c0..df103196b 100644 --- a/zh/tutorials/controlnet/depth-controlnet.mdx +++ b/zh/tutorials/controlnet/depth-controlnet.mdx @@ -55,7 +55,7 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [architecturerealmix_v11.safetensors](https://civitai.com/api/download/models/431755?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) +- [control_v11f1p_sd15_depth_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11f1p_sd15_depth_fp16.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/depth-t2i-adapter.mdx b/zh/tutorials/controlnet/depth-t2i-adapter.mdx index ebc2603e7..6c774a563 100644 --- a/zh/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/zh/tutorials/controlnet/depth-t2i-adapter.mdx @@ -76,7 +76,7 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [interiordesignsuperm_v2.safetensors](https://civitai.com/api/download/models/93152?type=Model&format=SafeTensor&size=full&fp=fp16) -- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/blob/main/models/t2iadapter_depth_sd15v2.pth?download=true) +- [t2iadapter_depth_sd15v2.pth](https://huggingface.co/TencentARC/T2I-Adapter/resolve/main/models/t2iadapter_depth_sd15v2.pth?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/mixing-controlnets.mdx b/zh/tutorials/controlnet/mixing-controlnets.mdx index 6530fbe94..05e0f3c74 100644 --- a/zh/tutorials/controlnet/mixing-controlnets.mdx +++ b/zh/tutorials/controlnet/mixing-controlnets.mdx @@ -54,9 +54,9 @@ translationBlockHashes: - [awpainting_v14.safetensors](https://civitai.com/api/download/models/624939?type=Model&format=SafeTensor&size=full&fp=fp16) -- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [control_v11p_sd15_scribble_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_scribble_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx b/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx index 88713f882..63c1d973f 100644 --- a/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -49,10 +49,10 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 如果你网络无法顺利完成对应模型的自动下载，请尝试手动下载下面的模型，并放置到指定目录中 -- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/blob/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) +- [control_v11p_sd15_openpose_fp16.safetensors](https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/resolve/main/control_v11p_sd15_openpose_fp16.safetensors?download=true) - [majicmixRealistic_v7.safetensors](https://civitai.com/api/download/models/176425?type=Model&format=SafeTensor&size=pruned&fp=fp16) - [japaneseStyleRealistic_v20.safetensors](https://civitai.com/api/download/models/85426?type=Model&format=SafeTensor&size=pruned&fp=fp16) -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true) ``` ComfyUI/ diff --git a/zh/tutorials/flux/flux-1-controlnet.mdx b/zh/tutorials/flux/flux-1-controlnet.mdx index 2937a115a..018ff43cc 100644 --- a/zh/tutorials/flux/flux-1-controlnet.mdx +++ b/zh/tutorials/flux/flux-1-controlnet.mdx @@ -44,14 +44,9 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 ## FLUX.1-Canny-dev 完整版工作流 - - - 下载 JSON 或在模板库中搜索 "Flux.1 Canny" - - - 在 Comfy Cloud 中打开 - - + +

Run on Comfy Cloud

+ ### 1. 工作流及相关素材 请下载下面的工作流图片,并拖入 ComfyUI 以加载工作流 @@ -71,10 +66,10 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 完整模型列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true) （请确保你已经同意了对应 repo 的协议） +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true) （请确保你已经同意了对应 repo 的协议） 文件保存位置： ``` @@ -114,14 +109,9 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora 工作流 - - - 下载 JSON 或在模板库中搜索 "Flux.1 Depth LoRA" - - - 在 Comfy Cloud 中打开 - - + +

Run on Comfy Cloud

+ LoRA 版本的工作流是在完整版本的基础上，添加了 LoRA 模型,相对于[完整版本的 Flux 工作流](/zh/tutorials/flux/flux-1-text-to-image),增加了对应 LoRA 模型的加载使用节点。 @@ -142,11 +132,11 @@ LoRA 版本的工作流是在完整版本的基础上，添加了 LoRA 模型, 完整模型列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) 文件保存位置： ``` diff --git a/zh/tutorials/flux/flux-1-fill-dev.mdx b/zh/tutorials/flux/flux-1-fill-dev.mdx index ba66cd3f0..6e924c295 100644 --- a/zh/tutorials/flux/flux-1-fill-dev.mdx +++ b/zh/tutorials/flux/flux-1-fill-dev.mdx @@ -39,10 +39,10 @@ Flux.1 fill dev 的核心特点： ![Flux Agreement](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 完整模型列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) 文件保存位置： ``` @@ -61,14 +61,13 @@ ComfyUI/ ### 1. Inpainting 工作流及相关素材 - - - Download JSON or search "flux_fill_inpaint" in Template Library - - - Open in Comfy Cloud - - + +

下载工作流图片

+ + + +

在 Comfy Cloud 上运行

+ 请下载下面的图片，并拖入 ComfyUI 以加载对应的工作流 ![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) @@ -94,16 +93,7 @@ ComfyUI/ ## Flux.1 Fill dev Outpainting 工作流 -### 1. Outpainting 工作流及相关素材 - - - - Download JSON or search "flux_fill_outpaint" in Template Library - - - Open in Comfy Cloud - - +### 1. Outpainting 工作流 请下载下面的图片，并拖入 ComfyUI 以加载对应的工作流 ![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) diff --git a/zh/tutorials/flux/flux-1-kontext-dev.mdx b/zh/tutorials/flux/flux-1-kontext-dev.mdx index 93e2dd03f..7934f644d 100644 --- a/zh/tutorials/flux/flux-1-kontext-dev.mdx +++ b/zh/tutorials/flux/flux-1-kontext-dev.mdx @@ -37,9 +37,9 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 ### 版本说明 -- **[FLUX.1 Kontext [pro]** — 商业版本，专注快速迭代编辑 -- **FLUX.1 Kontext [max]** — 实验版本，更强的提示遵循能力 -- **FLUX.1 Kontext [dev]** — 开源版本（本教程使用），12B参数，主要用于研究 +- **[FLUX.1 Kontext [pro]** - 商业版本，专注快速迭代编辑 +- **FLUX.1 Kontext [max]** - 实验版本，更强的提示遵循能力 +- **FLUX.1 Kontext [dev]** - 开源版本（本教程使用），12B参数，主要用于研究 目前在 ComfyUI 中，你可以使用所有的这些版本，其中 [Pro 及 Max 版本](/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) 可以通过 API 节点来进行调用，而 Dev 版本开源版本请参考本篇指南中的说明。 @@ -52,7 +52,7 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 **Diffusion Model** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) 如果你想要使用原始权重，可以访问 Black Forest Labs 的相关仓库获取原始模型权重进行使用。 @@ -63,7 +63,7 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 **Text Encoder** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) 或 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) 或 [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) 模型保存位置 @@ -81,14 +81,9 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 ## Flux.1 Kontext Dev 工作流 - - - 下载 JSON 或在模板库中搜索 "Flux Kontext Dev" - - - 在 Comfy Cloud 中打开 - - + +

Run on Comfy Cloud

+ 这个工作流使用了 `Load Image(from output)` 节点来加载需要编辑的图像，可以让你更方便地获取到编辑后的图像，从而进行多轮次编辑 diff --git a/zh/tutorials/flux/flux-1-text-to-image.mdx b/zh/tutorials/flux/flux-1-text-to-image.mdx index d315f0305..9cf090d5a 100644 --- a/zh/tutorials/flux/flux-1-text-to-image.mdx +++ b/zh/tutorials/flux/flux-1-text-to-image.mdx @@ -50,30 +50,25 @@ Flux 以其卓越的画面质量和灵活性而闻名，能够生成高质量、 #### 1. 工作流文件 - - - 在 Comfy Cloud 上运行此工作流 - - - 下载 JSON 或在模板库中搜索"Flux.1 Dev" - - - 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Dev 原始版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) + +

在 Comfy Cloud 上运行

+ + #### 2. 手动安装模型 - `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) 的协议后才能使用浏览器进行下载。 -- 如果你的显存较低，可以尝试使用 [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) 来替换 `t5xxl_fp16.safetensors` 文件。 +- 如果你的显存较低，可以尝试使用 [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) 来替换 `t5xxl_fp16.safetensors` 文件。 请下载下面的模型文件： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) 文件保存位置： @@ -109,19 +104,14 @@ ComfyUI/ #### 1. 工作流文件 - - - 在 Comfy Cloud 上运行此工作流 - - - 下载 JSON 或在模板库中搜索"Flux.1 Schnell" - - - 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Schnell 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) + +

在 Comfy Cloud 上运行

+ + #### 2. 手动安装模型 @@ -132,10 +122,10 @@ ComfyUI/ 完整模型文件列表： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) 文件保存位置： ``` @@ -167,38 +157,24 @@ fp8 版本是对 flux1 原版 fp16 版本的量化版本，在一定程度上这 ### Flux.1 Dev fp8 Checkpoint 版工作流 - - - 在 Comfy Cloud 上运行此工作流 - - - 下载 JSON 或在模板库中搜索"Flux.1 Dev FP8" - - - 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Dev fp8 Checkpoint 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) -请下载 [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 + +

在 Comfy Cloud 上运行

+ + +请下载 [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 确保对应的 `Load Checkpoint` 节点加载了 `flux1-dev-fp8.safetensors`，即可测试运行。 ### Flux.1 Schnell fp8 Checkpoint 版工作流 - - - 在 Comfy Cloud 上运行此工作流 - - - 下载 JSON 或在模板库中搜索"Flux.1 Schnell FP8" - - - 请下载下面的图片，并拖入 ComfyUI 中加载工作流。 ![Flux Schnell fp8 Checkpoint 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -请下载[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 +请下载[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 确保对应的 `Load Checkpoint` 节点加载了 `flux1-schnell-fp8.safetensors`，即可测试运行。 diff --git a/zh/tutorials/flux/flux-1-uso.mdx b/zh/tutorials/flux/flux-1-uso.mdx index fad515cc9..599402de6 100644 --- a/zh/tutorials/flux/flux-1-uso.mdx +++ b/zh/tutorials/flux/flux-1-uso.mdx @@ -32,14 +32,18 @@ USO 支持三种主要方法： ![工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - - - 下载工作流 JSON 并将其拖拽到 ComfyUI 中 - - - 在 Comfy Cloud 上运行此工作流 - - + +

下载 JSON 工作流

+ + + +

在 Comfy Cloud 上运行

+ 使用下面的图片作为输入 @@ -49,18 +53,18 @@ USO 支持三种主要方法： **checkpoints** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) **loras** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **model_patches** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **clip_visions** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) 请下载所有模型并将它们放置在以下目录中： diff --git a/zh/tutorials/flux/flux-2-dev.mdx b/zh/tutorials/flux/flux-2-dev.mdx index 50d2120de..6d500db06 100644 --- a/zh/tutorials/flux/flux-2-dev.mdx +++ b/zh/tutorials/flux/flux-2-dev.mdx @@ -65,15 +65,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/flux/flux-2-klein.mdx b/zh/tutorials/flux/flux-2-klein.mdx index 97884c79b..c3823aee0 100644 --- a/zh/tutorials/flux/flux-2-klein.mdx +++ b/zh/tutorials/flux/flux-2-klein.mdx @@ -45,19 +45,19 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 ## Flux.2 Klein 4B 模型下载 - + 4B 模型文本编码器。 - + 扩散模型（4B Base）。 - + 扩散模型（4B 蒸馏版）。 - + 4B 模型 VAE。 @@ -103,11 +103,11 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 扩散模型（9B 蒸馏版）。 - + 9B 模型文本编码器。 - + 9B 模型 VAE。 diff --git a/zh/tutorials/flux/flux1-krea-dev.mdx b/zh/tutorials/flux/flux1-krea-dev.mdx index 4d5e62769..23f2514c3 100644 --- a/zh/tutorials/flux/flux1-krea-dev.mdx +++ b/zh/tutorials/flux/flux1-krea-dev.mdx @@ -31,14 +31,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下载下面的图片或JSON，并拖入 ComfyUI 以加载对应工作流 ![Flux Krea Dev 工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - - - 在 Comfy Cloud 上运行此工作流 - - - 下载 JSON 或在模板库中搜索 "Flux.1 Krea Dev" - - + +

下载 JSON 格式工作流

+ + + +

在 Comfy Cloud 上运行

+ #### 2. 模型链接 @@ -48,7 +47,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面这个版本是原始权重，如果你追求更高质量有足够的显存，可以尝试这个版本 -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) - `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 的协议后才能使用浏览器进行下载。 @@ -56,12 +55,12 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 如果你使用过 Flux 相关的工作流，下面的模型是相同的，不需要重复下载 **Text encoders** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) 文件保存位置： ``` diff --git a/zh/tutorials/image/anima/anima.mdx b/zh/tutorials/image/anima/anima.mdx index 5639ccb84..d2691102f 100644 --- a/zh/tutorials/image/anima/anima.mdx +++ b/zh/tutorials/image/anima/anima.mdx @@ -82,15 +82,15 @@ Anima 提供两个工作流——基础版适用于标准使用，预览版适 所有模型文件均可从 Hugging Face 的 [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima) 获取。 - + Anima Base v1 扩散模型（2B）。 - + 两个工作流共用的文本编码器（Qwen-3 0.6B）。 - + 两个工作流共用的 VAE。 @@ -111,7 +111,7 @@ Anima 提供两个工作流——基础版适用于标准使用，预览版适 如果您使用 Preview 工作流，请下载以下预览版扩散模型： - + Anima Preview 扩散模型（2B）。 diff --git a/zh/tutorials/image/boogu/boogu-image-0.1.mdx b/zh/tutorials/image/boogu/boogu-image-0.1.mdx index 6fe4dc03c..7d70429c3 100644 --- a/zh/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/zh/tutorials/image/boogu/boogu-image-0.1.mdx @@ -50,19 +50,19 @@ Boogu-Image-0.1-Turbo 工作流使用一个子图封装了扩散、文本编码 ### Boogu-Image-0.1-Turbo 模型下载 - + Boogu-Image-0.1-Turbo 的扩散模型。 - + Boogu-Image-0.1-Turbo 的文本编码器。 - + Boogu-Image-0.1-Turbo 的 VAE。 - + Boogu-Image-0.1-Turbo 的 LoRA 模块 (rank-128)。 @@ -100,15 +100,15 @@ Boogu-Image-0.1-Turbo 工作流使用一个子图封装了扩散、文本编码 ### Boogu-Image-0.1-Edit 模型下载 - + Boogu-Image-0.1-Edit 的扩散模型。 - + Boogu-Image-0.1-Edit 的文本编码器。 - + Boogu-Image-0.1-Edit 的 VAE。 diff --git a/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index 620e11fdc..ad33ee57c 100644 --- a/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -37,17 +37,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **Diffusion model** -- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_t2i.safetensors) +- [cosmos_predict2_2B_t2i.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors) 其它权重请访问 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) 进行下载 **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) 文件保存位置 diff --git a/zh/tutorials/image/ernie-image/ernie-image.mdx b/zh/tutorials/image/ernie-image/ernie-image.mdx index da10ec8c4..51f8c668a 100644 --- a/zh/tutorials/image/ernie-image/ernie-image.mdx +++ b/zh/tutorials/image/ernie-image/ernie-image.mdx @@ -55,19 +55,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 所有重新打包的模型文件均可在 Hugging Face 的 [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) 获取。 - + ERNIE-Image 扩散模型。 - + ERNIE-Image 文本编码器。 - + ERNIE-Image 提示词增强器文本编码器。 - + ERNIE-Image VAE。 @@ -99,19 +99,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### ERNIE-Image-Turbo 模型下载 - + ERNIE-Image-Turbo 扩散模型。 - + ERNIE-Image-Turbo 文本编码器。 - + ERNIE-Image-Turbo 提示词增强器文本编码器。 - + ERNIE-Image-Turbo VAE。 diff --git a/zh/tutorials/image/hidream/hidream-e1.mdx b/zh/tutorials/image/hidream/hidream-e1.mdx index b281ca41d..8a9fd0afa 100644 --- a/zh/tutorials/image/hidream/hidream-e1.mdx +++ b/zh/tutorials/image/hidream/hidream-e1.mdx @@ -40,8 +40,8 @@ HiDream-E1 是智象未来(HiDream-ai) 正式开源的交互式图像编辑大 **Diffusion Model** 你不用同时下载这两个模型，由于 E1.1 是基于 E1 的迭代版本，在实际测试中它的质量和效果较 E1 都有较大提升 -- [hidream_e1_1_bf16.safetensors(推荐)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors(推荐)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **Text Encoder**： @@ -73,15 +73,6 @@ HiDream-E1 是智象未来(HiDream-ai) 正式开源的交互式图像编辑大 ## HiDream E1.1 ComfyUI 原生工作流示例 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "HiDream E1.1" - - - E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一百万分辨率，在工作流中使用了 `Scale Image to Total Pixels` 节点来将输入图片动态调整为 1百万像素 @@ -124,14 +115,9 @@ E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一 ## HiDream E1 ComfyUI 原生 工作流示例 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "HiDream E1 Full" - - + +

Run on Comfy Cloud

+ E1 是于 2025 年 4 月 28 日发布的，这个模型只支持 768*768 的分辨率 diff --git a/zh/tutorials/image/hidream/hidream-i1.mdx b/zh/tutorials/image/hidream/hidream-i1.mdx index e72b3a8da..040270b12 100644 --- a/zh/tutorials/image/hidream/hidream-i1.mdx +++ b/zh/tutorials/image/hidream/hidream-i1.mdx @@ -97,20 +97,15 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 full 版本工作流 - - - 在 Comfy Cloud 上运行此工作流，无需任何设置 - - - 下载工作流 JSON 文件 - - + +

Run on Comfy Cloud

+ #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 @@ -139,20 +134,15 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 dev 版本工作流 - - - 在 Comfy Cloud 上运行此工作流，无需任何设置 - - - 下载工作流 JSON 文件 - - + +

Run on Comfy Cloud

+ #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 @@ -180,20 +170,15 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 fast 版本工作流 - - - 在 Comfy Cloud 上运行此工作流，无需任何设置 - - - 下载工作流 JSON 文件 - - + +

Run on Comfy Cloud

+ #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 diff --git a/zh/tutorials/image/hidream/hidream-o1.mdx b/zh/tutorials/image/hidream/hidream-o1.mdx index 58f9b6bf8..4ef55b41f 100644 --- a/zh/tutorials/image/hidream/hidream-o1.mdx +++ b/zh/tutorials/image/hidream/hidream-o1.mdx @@ -51,19 +51,19 @@ HiDream-O1-Image 基于 [MIT 协议](https://github.com/HiDream-ai/HiDream-O1-Im **Checkpoint** — 经过重新打包和量化。所有版本均将最严重的离群值保留在 bf16，并移除了未使用的 deepstack 层： -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量化变体 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 全精度版（文件最大） +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量化变体 +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 全精度版（文件最大） **文本编码器（提示词优化）** — 所有版本通用： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA（可选）** — Dev 蒸馏也可以作为 LoRA 应用到 Full 模型中，让你可以调节蒸馏强度（由 [Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 全秩版 -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 剪枝版 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 基于 checkpoint 的替代蒸馏 +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 全秩版 +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 剪枝版 +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 基于 checkpoint 的替代蒸馏 ``` 📂 ComfyUI/ @@ -107,13 +107,13 @@ HiDream-O1-Image 基于 [MIT 协议](https://github.com/HiDream-ai/HiDream-O1-Im **Checkpoint（Dev 版）** — 经过重新打包和量化。所有版本均将最严重的离群值保留在 bf16，并移除了未使用的 deepstack 层： -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量化变体 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 全精度版（文件最大） +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量化变体 +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 全精度版（文件最大） **文本编码器（提示词优化）** — 所有版本通用： -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/zh/tutorials/image/ideogram/ideogram-v4.mdx b/zh/tutorials/image/ideogram/ideogram-v4.mdx index 73a743a1d..bc4f330ed 100644 --- a/zh/tutorials/image/ideogram/ideogram-v4.mdx +++ b/zh/tutorials/image/ideogram/ideogram-v4.mdx @@ -46,23 +46,23 @@ Ideogram 4.0 是 Ideogram 最新推出的文生图模型，已作为开源模型 你可以在 Hugging Face 的 [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) 找到所有重新打包的模型文件。 - + Ideogram 4.0 扩散模型（~13.8 GB）。放入 models/diffusion_models/ - + Ideogram 4.0 无条件扩散模型（~13.8 GB）。放入 models/diffusion_models/ - + Ideogram 4.0 文本编码器（~8 GB）。放入 models/text_encoders/ - + Ideogram 4.0 文本编码器（~2 GB）。放入 models/text_encoders/ - + Ideogram 4.0 VAE（~335 MB）。放入 models/vae/ diff --git a/zh/tutorials/image/krea/krea-2.mdx b/zh/tutorials/image/krea/krea-2.mdx index ba2115ce8..d9e9ff4c5 100644 --- a/zh/tutorials/image/krea/krea-2.mdx +++ b/zh/tutorials/image/krea/krea-2.mdx @@ -107,13 +107,13 @@ Krea 还发布了一组 Krea 2 的风格 LoRA。在 **CustomCombo** 节点中选 如需本地使用，请从 [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2) 下载 ComfyUI 优化版模型文件。 - + krea2_turbo_fp8_scaled.safetensors：Turbo FP8（推荐大多数用户使用） - + qwen3vl_4b_fp8_scaled.safetensors：Qwen3VL-4B 文本编码器 - + qwen_image_vae.safetensors diff --git a/zh/tutorials/image/lens/lens.mdx b/zh/tutorials/image/lens/lens.mdx index bbb3c4780..14f673e34 100644 --- a/zh/tutorials/image/lens/lens.mdx +++ b/zh/tutorials/image/lens/lens.mdx @@ -94,19 +94,19 @@ Lens Turbo 是蒸馏版，只需较少的采样步数即可生成图像，推理 所有模型文件可在 Hugging Face 上的 [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens) 找到。 - + Lens 扩散模型 (BF16) - + Lens Turbo 扩散模型 (BF16) - + Lens 和 Lens Turbo 共用的文本编码器 (GPT-OSS-20B) - + Lens 和 Lens Turbo 共用的 VAE (FLUX.2) diff --git a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 48bd8f50a..969ac372c 100644 --- a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -30,14 +30,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## NewBie-image 文生图工作流 -| -| -| 下载 JSON 或搜索"NewBie-image"模板 -| -| -| 在云端打开 -| -| + +

下载 JSON 工作流文件

+ + + +

在 ComfyUI Cloud 上运行

+ @@ -45,16 +44,16 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) **模型存放位置** diff --git a/zh/tutorials/image/omnigen/omnigen2.mdx b/zh/tutorials/image/omnigen/omnigen2.mdx index 667f22d59..eb09cac91 100644 --- a/zh/tutorials/image/omnigen/omnigen2.mdx +++ b/zh/tutorials/image/omnigen/omnigen2.mdx @@ -42,13 +42,13 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 由于本文涉及不同工作流，对应的模型文件及安装位置如下，对应工作流中也已包含了模型文件下载信息： **Diffusion Models）** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) **Text Encoders）** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) 文件保存位置： @@ -67,11 +67,9 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 ### 1. 工作流文件下载 - - - 在 Comfy Cloud 上运行 - - + +

在 Comfy Cloud 上运行

+ ![文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -97,11 +95,9 @@ OmniGen2 有丰富的图像编辑能力，并且支持为图像添加文本 ### 1. 工作流文件下载 - - - 在 Comfy Cloud 上运行 - - + +

在 Comfy Cloud 上运行

+ ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) 下载下面的图片，我们将使用它作为输入图片。 diff --git a/zh/tutorials/image/ovis/ovis-image.mdx b/zh/tutorials/image/ovis/ovis-image.mdx index 9aacd25c0..b28c481ee 100644 --- a/zh/tutorials/image/ovis/ovis-image.mdx +++ b/zh/tutorials/image/ovis/ovis-image.mdx @@ -22,14 +22,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Ovis-Image 文生图工作流 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索"Ovis image" - - + +

下载 JSON 工作流文件

+ + + +

在 ComfyUI Cloud 上运行

+ @@ -37,15 +36,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders（文本编码器）** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models（扩散模型）** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/pixeldit/pixeldit.mdx b/zh/tutorials/image/pixeldit/pixeldit.mdx index 00b5a3ee8..9daacab20 100644 --- a/zh/tutorials/image/pixeldit/pixeldit.mdx +++ b/zh/tutorials/image/pixeldit/pixeldit.mdx @@ -62,11 +62,11 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' PixelDiT 使用两个模型文件：文本编码器和扩散模型。 - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 文本编码器 - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 扩散模型 diff --git a/zh/tutorials/image/qwen/qwen-image-2512.mdx b/zh/tutorials/image/qwen/qwen-image-2512.mdx index 0abcfebac..c7e9877ee 100644 --- a/zh/tutorials/image/qwen/qwen-image-2512.mdx +++ b/zh/tutorials/image/qwen/qwen-image-2512.mdx @@ -43,14 +43,9 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - - + 在 Comfy Cloud 上运行 - - - 下载 JSON 或搜索 "Qwen-Image-2512" 在模板库中 - - + ### 1. 工作流文件 @@ -60,25 +55,28 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **Text to Image (Qwen-Image 2512)**：标准 50 步生成 - **Text to Image (Qwen-Image 2512 4steps)**：使用 Lightning LoRA 的 4 步加速生成 + +

下载 JSON 工作流

+ ### 2. 模型下载 **文本编码器** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（可选 - 用于 4 步 Lightning 加速）** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **扩散模型** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（推荐大多数用户使用） -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（如果您有足够的显存并想要更好的质量） +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（推荐大多数用户使用） +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（如果您有足够的显存并想要更好的质量） **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx index 5d8fe82dc..345d68713 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -33,32 +33,31 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载。 - - - 在 ComfyUI Cloud 上运行 - - - 下载 JSON 格式工作流 - - + +

下载 JSON 格式工作流

+ + + +

在 ComfyUI Cloud 上运行

+ ### 2. 模型下载 **Text Encoders（文本编码器）** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（可选 - 用于 4 步 Lightning 加速）** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **Diffusion Models（扩散模型）** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/qwen/qwen-image-edit.mdx b/zh/tutorials/image/qwen/qwen-image-edit.mdx index 4ab071a5b..97a4a2e7e 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit.mdx @@ -47,14 +47,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - - - 下载 JSON 格式工作流或在模板库中搜索"image_qwen_image_edit" - - - 在 ComfyUI Cloud 上运行此工作流，零设置 - - + +

下载 JSON 格式工作流

+ + + +

在 ComfyUI Cloud 上运行

+ 下载下面的图片作为输入 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -65,19 +64,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Diffusion model** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) Model Storage Location diff --git a/zh/tutorials/image/qwen/qwen-image-layered.mdx b/zh/tutorials/image/qwen/qwen-image-layered.mdx index 2a835dc5b..49f685ab0 100644 --- a/zh/tutorials/image/qwen/qwen-image-layered.mdx +++ b/zh/tutorials/image/qwen/qwen-image-layered.mdx @@ -30,15 +30,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 工作流 -| -| -| 下载 JSON 格式工作流 -| -| -| -| 在 ComfyUI Cloud 上运行 -| -| + +

下载 JSON 格式工作流

+ + + +

在 ComfyUI Cloud 上运行

+ @@ -46,15 +44,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) **模型保存位置** @@ -89,4 +87,4 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 提示词（可选） -文本提示词用于描述输入图像的整体内容，包括可能被部分遮挡的元素（例如，你可以指定隐藏在前景物体后面的文字）。它不是用来明确控制各个图层的语义内容的。 +文本提示词用于描述输入图像的整体内容——包括可能被部分遮挡的元素（例如，你可以指定隐藏在前景物体后面的文字）。它不是用来明确控制各个图层的语义内容的。 diff --git a/zh/tutorials/image/qwen/qwen-image.mdx b/zh/tutorials/image/qwen/qwen-image.mdx index a004afd05..86cca15d3 100644 --- a/zh/tutorials/image/qwen/qwen-image.mdx +++ b/zh/tutorials/image/qwen/qwen-image.mdx @@ -61,12 +61,9 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - - - - - - + + 在 Comfy Cloud 上运行 + 在本篇文档所附工作流中使用的不同模型有三种 1. Qwen-Image 原版模型 fp8_e4m3fn @@ -87,11 +84,14 @@ GPU: RTX4090D 24GB 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - + +

下载原始版 JSON 格式工作流

+ 蒸馏版 - - + +

下载蒸馏版JSON 格式工作流

+ ### 2. 模型下载 **你可以在 ComfyOrg 仓库找到的版本** @@ -104,12 +104,12 @@ GPU: RTX4090D 24GB **Diffusion model** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 蒸馏版本原始作者建议在 15 步 cfg 1.0 @@ -118,15 +118,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) 模型保存位置 @@ -165,19 +165,18 @@ Qwen_image_distill 这是一个 ControlNet 模型 - - - - - - + + 在 Comfy Cloud 上运行 + ### 1. 工作流及输入图片 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - + +

下载 JSON 格式工作流

+ 下载下面的图片作为输入 ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -186,7 +185,7 @@ Qwen_image_distill 1. InstantX Controlnet -下载 [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) 并保存到 `ComfyUI/models/controlnet/` 文件夹下 +下载 [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) 并保存到 `ComfyUI/models/controlnet/` 文件夹下 2. **Lotus Depth model** @@ -200,11 +199,11 @@ Qwen_image_distill **Diffusion Model** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) **VAE Model** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) 或者任意的 SD1.5 的 VAE 都可以使用 +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) 或者任意的 SD1.5 的 VAE 都可以使用 ``` ComfyUI/ @@ -226,12 +225,9 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches 工作流 - - - - - - + + 在 Comfy Cloud 上运行 + 这个模型实际上并不是一个 controlnet，而是一个 Model patch， 支持 canny、depth、inpaint 三种不同的控制模式 @@ -244,7 +240,9 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 下载下面的图片拖入 ComfyUI 中以加载对应的工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - + +

下载 JSON 格式工作流

+ 下载下面的图片作为输入图片： @@ -254,9 +252,9 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 其它模型与 Qwen-Image 基础工作流一致，你只需下载下面的模型并保存到 `ComfyUI/models/model_patches` 文件夹中 -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. 工作流使用说明 @@ -298,12 +296,9 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http ## Qwen Image union ControlNet LoRA 工作流 - - - - - - + + 在 Comfy Cloud 上运行 + 原始模型地址：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 图像结构控制lora 支持 canny、depth、post、lineart、softedge、normal、openpose @@ -312,7 +307,9 @@ Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://h 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - + +

下载 JSON 格式工作流

+ 下载下面的图片作为输入图片 diff --git a/zh/tutorials/image/z-image/z-image-turbo.mdx b/zh/tutorials/image/z-image/z-image-turbo.mdx index bc39c7864..e8871670a 100644 --- a/zh/tutorials/image/z-image/z-image-turbo.mdx +++ b/zh/tutorials/image/z-image/z-image-turbo.mdx @@ -46,15 +46,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### Z-Image-Turbo 模型下载 - + Z-Image-Turbo 文本编码器。 - + Z-Image-Turbo 扩散模型。 - + Z-Image-Turbo VAE。 @@ -81,7 +81,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### ControlNet 所需的额外模型 - + Z-Image-Turbo ControlNet 模型补丁。 diff --git a/zh/tutorials/image/z-image/z-image.mdx b/zh/tutorials/image/z-image/z-image.mdx index da8e02180..6115cb4b8 100644 --- a/zh/tutorials/image/z-image/z-image.mdx +++ b/zh/tutorials/image/z-image/z-image.mdx @@ -37,15 +37,15 @@ Z-Image（Base）是非蒸馏基础模型，专为社区驱动的微调和自定 ## Z-Image 模型下载 - + Z-Image 文本编码器。 - + Z-Image 扩散模型。 - + Z-Image VAE。 diff --git a/zh/tutorials/llm/gemma4/gemma4.mdx b/zh/tutorials/llm/gemma4/gemma4.mdx index 956ac5d5f..c953ec9f4 100644 --- a/zh/tutorials/llm/gemma4/gemma4.mdx +++ b/zh/tutorials/llm/gemma4/gemma4.mdx @@ -74,11 +74,11 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' Gemma 4 模型在 ComfyUI 中以文本编码器（text encoder）形式加载。下载对应的模型文件并放入正确的目录： - + 快速轻量，推荐消费级 GPU 使用。 - + 性能均衡，工作流默认使用此模型。 diff --git a/zh/tutorials/llm/qwen/qwen3.mdx b/zh/tutorials/llm/qwen/qwen3.mdx index 854a1d1ab..8e0bfa5bc 100644 --- a/zh/tutorials/llm/qwen/qwen3.mdx +++ b/zh/tutorials/llm/qwen/qwen3.mdx @@ -71,15 +71,15 @@ Qwen 3.0 非常适合在 ComfyUI 工作流中需要结构化文本生成和智 Qwen 3.0 模型以文本编码器的形式加载到 ComfyUI 中，模型文件与 Qwen3.5 共用。根据你的硬件选择合适的版本： - + 轻量版，约 4.5 GB。适合低显存环境和快速下载。 - + 大小和质量均衡。推荐大多数消费级 GPU。 - + 最大版本，约 19 GB。输出质量更高，需要更多显存。 diff --git a/zh/tutorials/llm/qwen/qwen3_5.mdx b/zh/tutorials/llm/qwen/qwen3_5.mdx index 56cb5f3b8..1e14b68d0 100644 --- a/zh/tutorials/llm/qwen/qwen3_5.mdx +++ b/zh/tutorials/llm/qwen/qwen3_5.mdx @@ -73,15 +73,15 @@ Qwen3.5 在需要将视觉理解与文本生成结合的 ComfyUI 场景中表现 Qwen3.5 模型以文本编码器的形式加载到 ComfyUI 中。根据你的硬件选择合适的版本： - + 轻量版，约 4.5 GB。适合低显存环境和快速下载。 - + 大小和质量均衡。推荐大多数消费级 GPU。 - + 最大版本，约 19 GB。输出质量更高，需要更多显存。 diff --git a/zh/tutorials/partner-nodes/google/gemini.mdx b/zh/tutorials/partner-nodes/google/gemini.mdx index 2f619e998..92c3a0fa7 100644 --- a/zh/tutorials/partner-nodes/google/gemini.mdx +++ b/zh/tutorials/partner-nodes/google/gemini.mdx @@ -22,11 +22,13 @@ Google Gemini 是 Google 推出的一款强大的 AI 模型，支持对话、文 请下载下面的 Json 文件并拖入 ComfyUI 中加载对应工作流。 - - - 下载 Json 格式工作流文件 - - + +

下载 Json 格式工作流文件

+ ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx index 6fc8db3e1..f685102aa 100644 --- a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -51,8 +51,9 @@ Kling 2.6 Motion Control 是由快手开发的专门多模态模型，能够实 ## Kling 2.6 Motion Control 工作流 - - + +

下载 Json 格式工作流文件

+ ## 输入要求 diff --git a/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index d479aed4a..7d82f0d0e 100644 --- a/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -54,11 +54,9 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_text_to_video.mp4" > - - - 下载 Json 格式工作流文件 - - + +

下载 Json 格式工作流文件

+ ### 2. 按步骤完成工作流的运行 @@ -82,11 +80,9 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_image_to_video.mp4" > - - - 下载 Json 格式工作流文件 - - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 @@ -116,11 +112,9 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video.mp4" > - - - 下载 Json 格式工作流文件 - - + +

下载 Json 格式工作流文件

+ 下载下面的视频作为输入视频： diff --git a/zh/tutorials/partner-nodes/openai/chat.mdx b/zh/tutorials/partner-nodes/openai/chat.mdx index a5f94132c..10107c75e 100644 --- a/zh/tutorials/partner-nodes/openai/chat.mdx +++ b/zh/tutorials/partner-nodes/openai/chat.mdx @@ -22,9 +22,13 @@ OpenAI 是一家专注于生成式 AI 的科技公司，提供强大的对话功 请下载下面的 Json 文件并拖入 ComfyUI 中加载对应工作流。 -| -| 下载 Json 格式工作流文件 -| + +

下载 Json 格式工作流文件

+ ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/partner-nodes/rodin/model-generation.mdx b/zh/tutorials/partner-nodes/rodin/model-generation.mdx index 8d490add5..051d1f871 100644 --- a/zh/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/zh/tutorials/partner-nodes/rodin/model-generation.mdx @@ -32,9 +32,13 @@ Hyper3D Rodin (hyper3d.ai) 是一个专注于通过人工智能快速生成高 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - - 单视角模型生成 (Json格式) - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 @@ -62,9 +66,13 @@ Hyper3D Rodin (hyper3d.ai) 是一个专注于通过人工智能快速生成高 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - - 多视角模型生成 (Json格式) - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 diff --git a/zh/tutorials/partner-nodes/runway/video-generation.mdx b/zh/tutorials/partner-nodes/runway/video-generation.mdx index 1bfb860f8..7000f87c8 100644 --- a/zh/tutorials/partner-nodes/runway/video-generation.mdx +++ b/zh/tutorials/partner-nodes/runway/video-generation.mdx @@ -37,7 +37,9 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen3a_turbo_image_to_video/runway_image_to_video_gen3a_turbo.mp4" > - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 @@ -65,7 +67,9 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/gen4_turbo_image_to_video/runway_gen4_turo_image_to_video.mp4" > - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 @@ -95,7 +99,9 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/runway/first_last_frame_to_video/runway_first_last_frame.mp4" > - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 diff --git a/zh/tutorials/partner-nodes/tripo/model-generation.mdx b/zh/tutorials/partner-nodes/tripo/model-generation.mdx index a65091422..dca3da2b4 100644 --- a/zh/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/zh/tutorials/partner-nodes/tripo/model-generation.mdx @@ -35,11 +35,9 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - - - 文生模型工作流 - - + +

下载 Json 格式工作流文件

+ ### 2. 按步骤完成工作流的运行 @@ -60,11 +58,9 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - - - 图生模型工作流 - - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 @@ -90,11 +86,9 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - - - 多视图模型生成工作流 - - + +

下载 Json 格式工作流文件

+ 下载下面的图片作为输入图片 diff --git a/zh/tutorials/utility/depth-anything-3.mdx b/zh/tutorials/utility/depth-anything-3.mdx index 6fdbca81a..26666a4db 100644 --- a/zh/tutorials/utility/depth-anything-3.mdx +++ b/zh/tutorials/utility/depth-anything-3.mdx @@ -39,10 +39,10 @@ ComfyUI 现已原生支持 Depth Anything 3 节点。开始前请确保已更新 下载 Depth Anything 3 的模型文件并将其保存到对应的 ComfyUI 文件夹： -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — 轻量快速推理 -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — 平衡性能 -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 最佳单目深度，含天空检测 -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 物理度量深度（米级） +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — 轻量快速推理 +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — 平衡性能 +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — 最佳单目深度，含天空检测 +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — 物理度量深度（米级） ``` ComfyUI/ diff --git a/zh/tutorials/utility/face-detection/mediapipe.mdx b/zh/tutorials/utility/face-detection/mediapipe.mdx index 0f5817700..7a7289a5d 100644 --- a/zh/tutorials/utility/face-detection/mediapipe.mdx +++ b/zh/tutorials/utility/face-detection/mediapipe.mdx @@ -55,7 +55,7 @@ MediaPipe Face Detection 已原生集成到 ComfyUI（PR [#14009](https://github MediaPipe Face Detection 模型托管在 [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe) 上。 -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) 将其放置在以下目录结构中： diff --git a/zh/tutorials/utility/moge.mdx b/zh/tutorials/utility/moge.mdx index df34497e1..0c6c7223d 100644 --- a/zh/tutorials/utility/moge.mdx +++ b/zh/tutorials/utility/moge.mdx @@ -50,8 +50,8 @@ ComfyUI 现已原生支持 MoGe 节点。开始前请确保已更新到最新版 下载 MoGe 检查点并保存到相应的 ComfyUI 文件夹： -- **MoGe-2（推荐）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1（基线版本）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2（推荐）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1（基线版本）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/utility/pose-detection-sdpose.mdx b/zh/tutorials/utility/pose-detection-sdpose.mdx index 1bc6257dc..81f0c1d4d 100644 --- a/zh/tutorials/utility/pose-detection-sdpose.mdx +++ b/zh/tutorials/utility/pose-detection-sdpose.mdx @@ -87,11 +87,11 @@ SDPose + RT-DETRv4 已在 ComfyUI 中原生支持（PR [#12748](https://github.c SDPose 和 RT-DETRv4 模型文件托管在 [Comfy-Org SDPose 模型仓库](https://huggingface.co/Comfy-Org/SDPose) 中。 **checkpoints**（SDPose 模型）： -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models**（RT-DETRv4 检测器）： -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors)（推荐） -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors)（全精度，体积更大） +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors)（推荐） +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors)（全精度，体积更大） 将模型放置在以下目录结构中： diff --git a/zh/tutorials/utility/remove-background-birefnet.mdx b/zh/tutorials/utility/remove-background-birefnet.mdx index 8550f8625..9a248d3d2 100644 --- a/zh/tutorials/utility/remove-background-birefnet.mdx +++ b/zh/tutorials/utility/remove-background-birefnet.mdx @@ -48,7 +48,7 @@ BiRefNet 在 ComfyUI 中获得原生支持（PR [#12747](https://github.com/Comf BiRefNet 模型托管在 [Comfy-Org BiRefNet 模型仓库](https://huggingface.co/Comfy-Org/BiRefNet)。 -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) 放置到以下目录结构： diff --git a/zh/tutorials/utility/video-segment-sam3.mdx b/zh/tutorials/utility/video-segment-sam3.mdx index 7ff2ad8d1..1883a1d74 100644 --- a/zh/tutorials/utility/video-segment-sam3.mdx +++ b/zh/tutorials/utility/video-segment-sam3.mdx @@ -68,7 +68,7 @@ SAM 3.1 能根据文本提示在视频帧中分割并追踪物体。以上示例 SAM 3.1 模型托管在 [Comfy-Org SAM 3.1 模型仓库](https://huggingface.co/Comfy-Org/sam3.1)。 -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) 放置到以下目录结构： diff --git a/zh/tutorials/utility/void-video-inpainting.mdx b/zh/tutorials/utility/void-video-inpainting.mdx index a9a6a49d3..bab1a6c0a 100644 --- a/zh/tutorials/utility/void-video-inpainting.mdx +++ b/zh/tutorials/utility/void-video-inpainting.mdx @@ -65,24 +65,24 @@ VOID 在 ComfyUI 中获得原生支持（PR [#13403](https://github.com/Comfy-Or **扩散模型** — 核心的两阶段修复模型： -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — 精炼阶段，时间稳定性更佳 -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — 主要阶段 +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 精炼阶段，时间稳定性更佳 +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 主要阶段 **VAE：** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) **光流模型：** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 分割模型：** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **文本编码器：** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ diff --git a/zh/tutorials/video/bytedance/bernini-r.mdx b/zh/tutorials/video/bytedance/bernini-r.mdx index de5dc5c8b..8506530ac 100644 --- a/zh/tutorials/video/bytedance/bernini-r.mdx +++ b/zh/tutorials/video/bytedance/bernini-r.mdx @@ -49,16 +49,16 @@ ComfyUI 现已原生支持 Bernini-R 节点。开始前请确保已更新到最 下载所需的模型权重并将其保存到对应的 ComfyUI 文件夹： **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index ecafc9cbe..a61a79890 100644 --- a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -12,36 +12,24 @@ Cosmos-Predict2 是由 NVIDIA 推出的新一代物理世界基础模型，专 该模型具备极高的物理准确性、环境交互性和细节还原能力，能够真实模拟复杂的物理现象与动态场景。 Cosmos-Predict2 支持文本到图像（Text2Image）和视频到世界（Video2World）等多种生成方式，广泛应用于工业仿真、自动驾驶、城市规划、科学研究等领域，是推动智能视觉与物理世界深度融合的重要基础工具。 - - - Cosmos-Predict2 源代码和文档 - - - Cosmos-Predict2 模型集合 - - +GitHub:[Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) +huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) 本篇指南将引导你完成在 ComfyUI 中 **图生视频** 的工作流 对于文生图部分，请参考下面的部分 - - - 使用 Cosmos-Predict2 进行文生图 - - - 在 Comfy Cloud 上使用强大的 GPU 运行 Cosmos-Predict2 工作流 - - + + 使用 Cosmos-Predict2 的进行文生图 + +{/* ## Cosmos Predict2 Video2World 工作流 对于 2B 版本，在我们测试使用时，大约占用 16GB 的显存 -### 1.下载工作流文件 - -请下载以下视频并将其拖入 ComfyUI 以加载工作流。工作流已嵌入模型下载链接。 +#### 1.下载工作流文件 - - - 下载 JSON 格式工作流文件 - - - 在 Comfy Cloud 上运行此工作流，模型预装完毕 - - + +

下载 Json 格式工作流文件

+ + 请下载下面的图片作为输入文件： ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/input.png) -### 2.手动模型安装 - -如果模型未成功下载，您可以在此部分手动下载。 +### 2.手动模型安装 **Diffusion model** - - cosmos_predict2_2B_video2world_480p_16fps.safetensors - +- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) 其它权重请访问 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) 进行下载 **Text encoder** - - oldt5_xxl_fp8_e4m3fn_scaled.safetensors - +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** - - wan_2.1_vae.safetensors - +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + 文件保存位置 ``` @@ -111,4 +88,4 @@ Cosmos-Predict2 支持文本到图像（Text2Image）和视频到世界（Video2 6. (可选) 你可以在 `ClipTextEncode` 节点中修改提示词 7. (可选) 修改 `CosmosPredict2ImageToVideoLatent` 节点中的尺寸和帧数 8. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行视频生成 -9. 生成完成后对应的视频会自动保存到 `ComfyUI/output/` 目录下，你也可以在 `save video` 节点中预览或者调整保存位置 +9. 生成完成后对应的视频会自动保存到 `ComfyUI/output/` 目录下，你也可以在 `save video` 节点中预览或者调整保存位置 */} diff --git a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index ff50f788c..6d4a77890 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -24,17 +24,17 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **diffusion_models** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **vae** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) Model Storage Location diff --git a/zh/tutorials/video/hunyuan/hunyuan-video.mdx b/zh/tutorials/video/hunyuan/hunyuan-video.mdx index 38aa3e770..6f53ebd0a 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video.mdx @@ -47,9 +47,9 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 在文生视频和图生视频的工作流中下面的这些模型是共有的，请完成下载并保存到指定目录中 -- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors?download=true) -- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) -- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/clip_l.safetensors?download=true) +- [llava_llama3_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors?download=true) +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/vae/hunyuan_video_vae_bf16.safetensors?download=true) 保存位置： @@ -74,7 +74,7 @@ ComfyUI/ ### 2. 混元文生图模型 -请下载 [hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) 并保存至 `ComfyUI/models/diffusion_models` 文件夹中 +请下载 [hunyuan_video_t2v_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_t2v_720p_bf16.safetensors?download=true) 并保存至 `ComfyUI/models/diffusion_models` 文件夹中 确保包括共用模型文件夹有以下完整的模型文件： @@ -126,7 +126,7 @@ ComfyUI/ ### v1 及 v2 版本共用的模型 请下载下面的文件，并保存到 `ComfyUI/models/clip_vision` 目录中 -- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) +- [llava_llama3_vision.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors?download=true) ### v1 “concat” 图生视频工作流 @@ -140,7 +140,7 @@ ComfyUI/ #### 2. v1 版本模型 -- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors?download=true) 确保包括共用模型文件夹有以下完整的模型文件： @@ -186,7 +186,7 @@ v2 版本的工作流与 v1 版本的工作流基本相同，你只需要下载 #### 2. v2 版本模型 -- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) +- [hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors?download=true) 确保包括共用模型文件夹有以下完整的模型文件： diff --git a/zh/tutorials/video/kandinsky/kandinsky-5.mdx b/zh/tutorials/video/kandinsky/kandinsky-5.mdx index 03674838e..c7987f9df 100644 --- a/zh/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/zh/tutorials/video/kandinsky/kandinsky-5.mdx @@ -51,39 +51,21 @@ Kandinsky 5.0 使用带有 Flow Matching 的潜在扩散管道，具有以下特 请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `视频` 找到 "Kandinsky 5.0 T2V" 以加载工作流。 - - - 下载 T2V 工作流以本地使用 - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ ### 2. 手动下载模型 **Text Encoders** - - - - Qwen2.5-VL 7B 文本编码器 (FP8) - - - CLIP-L 文本编码器 - - +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) **Diffusion Model** - - - Kandinsky 5.0 T2V Lite SFT 扩散模型 (5s) - +- [kandinsky5lite_t2v_sft_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s/resolve/main/model/kandinsky5lite_t2v_sft_5s.safetensors) **VAE** - - - HunyuanVideo 3D VAE - +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) ``` ComfyUI/ @@ -103,39 +85,21 @@ ComfyUI/ 请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `视频` 找到 "Kandinsky 5.0 I2V" 以加载工作流。 - - - 下载 I2V 工作流以本地使用 - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ ### 2. 手动下载模型 **Text Encoders** - - - - Qwen2.5-VL 7B 文本编码器 (FP8) - - - CLIP-L 文本编码器 - - +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors) **Diffusion Model** - - - Kandinsky 5.0 I2V Lite 扩散模型 (5s) - +- [kandinsky5lite_i2v_5s.safetensors](https://huggingface.co/kandinskylab/Kandinsky-5.0-I2V-Lite-5s/resolve/main/model/kandinsky5lite_i2v_5s.safetensors) **VAE** - - - HunyuanVideo 3D VAE - +- [hunyuan_video_vae_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/resolve/main/hunyuan_video_vae_bf16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/ltxv.mdx b/zh/tutorials/video/ltxv.mdx index 890d8bfa4..a777f7cdd 100644 --- a/zh/tutorials/video/ltxv.mdx +++ b/zh/tutorials/video/ltxv.mdx @@ -26,17 +26,9 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 通过首帧图像控制视频生成：[示例首帧](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)。 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "LTX-Video" - - - 获取此工作流的示例输入图片 - - + +

在 Comfy 云上运行

+ LTX-Video 图生视频 @@ -46,15 +38,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 文生视频 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "LTX-Video" - - - LTX-Video 文生视频 @@ -65,13 +48,8 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 将以下模型下载并放到指定位置： - - 下载并放置到 ComfyUI/models/checkpoints/ - - - - 下载并放置到 ComfyUI/models/text_encoders/（如尚未下载） - +- [ltx-video-2b-v0.9.5.safetensors](https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltx-video-2b-v0.9.5.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/mochi_preview_repackaged/resolve/main/split_files/text_encoders/t5xxl_fp16.safetensors?download=true)（如尚未下载） ``` ├── checkpoints/ diff --git a/zh/tutorials/video/wan/fun-camera.mdx b/zh/tutorials/video/wan/fun-camera.mdx index e0d9387c6..88be391c5 100644 --- a/zh/tutorials/video/wan/fun-camera.mdx +++ b/zh/tutorials/video/wan/fun-camera.mdx @@ -13,6 +13,7 @@ translationBlockHashes: "Performance Reference": 32425486 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 Wan2.1 Fun Camera @@ -35,49 +36,21 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面的所有模型你可以在 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 找到 -### Diffusion Models - -选择 1.3B 或 14B： - - - - Wan2.1 Fun Camera 1.3B 扩散模型 - - - Wan2.1 Fun Camera 14B 扩散模型 - - +**Diffusion Models** 选择 1.3B 或 14B： +- [wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_1.3B_bf16.safetensors) +- [wan2.1_fun_camera_v1.1_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_camera_v1.1_14B_bf16.safetensors) 下面的模型，如果你使用过 Wan2.1 的相关模型，那么你应该已经有了下面的模型，如果没有，请下载下面的模型： -### Text Encoders - -选择其中一个： - - - - 全精度文本编码器 - - - FP8 量化文本编码器（推荐用于更低显存） - - - -### VAE +**Text Encoders** 选择其中一个： +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - - Wan2.1 VAE 模型 - - +**VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) -### CLIP Vision - - - - CLIP 视觉编码器 - - +**CLIP Vision** +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) 文件保存位置： @@ -97,16 +70,9 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 1.3B 原生工作流示例 -### 1. 下载工作流 +### 1. 工作流相关文件下载 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "Wan 2.1 Fun Camera 1.3B" - - +#### 1.1 工作流文件 下载下面的视频，并拖入 ComfyUI 中以加载对应的工作流： @@ -116,19 +82,21 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B.mp4" > + +

下载 Json 格式工作流文件

+ 如果你想使用 14B 版本，只需要将模型文件替换为 14B 版本即可，但请注意显存要求。 -### 2. 下载输入素材 +#### 1.2 输入图片下载 + + +请下载下面的图片，我们将作为起始帧： - - - 下载下面的图片作为 1.3B 工作流的起始帧 - - +![输入参考图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) -### 3. 按步骤完成工作流 +### 2. 按步骤完成工作流 ![Wan2.1 Fun Camera 工作流步骤](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -145,30 +113,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## ComfyUI Wan2.1 Fun Camera 14B 工作流及输入图片 -### 1. 下载工作流 - - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "Wan 2.1 Fun Camera 14B" - - - -### 2. 下载输入素材 + +

下载 Json 格式工作流文件

+ - - - 下载下面的图片作为 14B 工作流的起始帧 - - +**输入图片** +![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) ## 性能参考 diff --git a/zh/tutorials/video/wan/fun-control.mdx b/zh/tutorials/video/wan/fun-control.mdx index e63c29035..a45cec2a1 100644 --- a/zh/tutorials/video/wan/fun-control.mdx +++ b/zh/tutorials/video/wan/fun-control.mdx @@ -54,18 +54,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 点击对应链接进行下载，如果你之前使用过 Wan 相关的工作流，那么你仅需要下载 **Diffusino models** **Diffusion models** 选择 1.3B 或 14B, 14B 的文件体积更大（32GB）但是对于运行显存要求也较高， -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-Control.safetensors` **Text encoders** 选择下面两个模型中的一个，fp16 精度体积较大对性能要求高 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/fun-inp.mdx b/zh/tutorials/video/wan/fun-inp.mdx index dcaa210d8..d0e8741aa 100644 --- a/zh/tutorials/video/wan/fun-inp.mdx +++ b/zh/tutorials/video/wan/fun-inp.mdx @@ -54,18 +54,18 @@ Wan-Fun InP 是阿里巴巴推出的开源视频生成模型，属于 ​​Wan2 下面的模型你可以在 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 和 [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) 找到 **Diffusion models** 选择 1.3B 或 14B, 14B 的文件体积更大（32GB）但是对于运行显存要求也较高， -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-InP.safetensors` +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-InP.safetensors` **Text encoders** 选择下面两个模型中的一个，fp16 精度体积较大对性能要求高 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/vace.mdx b/zh/tutorials/video/wan/vace.mdx index 4bd6dfe2a..4d6fc967f 100644 --- a/zh/tutorials/video/wan/vace.mdx +++ b/zh/tutorials/video/wan/vace.mdx @@ -59,19 +59,19 @@ VACE 14B 是阿里通义万相团队推出的开源视频编辑统一模型。 ### 模型下载 **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 如果你之前使用过 Wan Video 相关的工作流，下面的模型文件你已经下载过了。 **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) 从**Text encoders** 选择一个版本进行下载 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/wan-ati.mdx b/zh/tutorials/video/wan/wan-ati.mdx index 3ab064d08..240fdb0fd 100644 --- a/zh/tutorials/video/wan/wan-ati.mdx +++ b/zh/tutorials/video/wan/wan-ati.mdx @@ -47,17 +47,17 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 如果你没有成功下载工作流中的模型文件，可以尝试使用下面的链接手动下载 **Diffusion Model** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **Text encoders** Chose one of following model -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) File save location diff --git a/zh/tutorials/video/wan/wan-causal-forcing.mdx b/zh/tutorials/video/wan/wan-causal-forcing.mdx index 854b7a722..62324a6a1 100644 --- a/zh/tutorials/video/wan/wan-causal-forcing.mdx +++ b/zh/tutorials/video/wan/wan-causal-forcing.mdx @@ -83,10 +83,10 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B 检查点 - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B 检查点（最低 8GB 显存） @@ -94,10 +94,10 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### CLIP 和 VAE - + google-bert/bert-base-uncased — CLIP 文本编码器 - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/zh/tutorials/video/wan/wan-dancer.mdx b/zh/tutorials/video/wan/wan-dancer.mdx index a03c899b5..21ae06fa7 100644 --- a/zh/tutorials/video/wan/wan-dancer.mdx +++ b/zh/tutorials/video/wan/wan-dancer.mdx @@ -61,20 +61,20 @@ Wan万相 Dancer 工作流接受两个输入：角色参考图像和一个音频 ### 3. 手动下载模型 **扩散模型** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **文本编码器** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP 视觉** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan-flf.mdx b/zh/tutorials/video/wan/wan-flf.mdx index 55a47107c..a624005e3 100644 --- a/zh/tutorials/video/wan/wan-flf.mdx +++ b/zh/tutorials/video/wan/wan-flf.mdx @@ -56,7 +56,7 @@ Wan FLF2V（首尾帧视频生成）是由阿里通义万相团队推出的开 **diffusion_models** 根据你的硬件情况选择一个版本进行下载，FP8 版本对显存要求低一些 -- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8:[wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -64,14 +64,14 @@ Wan FLF2V（首尾帧视频生成）是由阿里通义万相团队推出的开 从**Text encoders** 选择一个版本进行下载， -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/wan-move.mdx b/zh/tutorials/video/wan/wan-move.mdx index 793153134..93a027390 100644 --- a/zh/tutorials/video/wan/wan-move.mdx +++ b/zh/tutorials/video/wan/wan-move.mdx @@ -28,37 +28,37 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Wan-Move 图生视频工作流 - - 下载工作流 - + +

下载 JSON 工作流文件

+ - - 在云端运行 - + +

在 ComfyUI Cloud 上运行

+ ## 模型下载链接 - - **text_encoders** -- umt5_xxl_fp8_e4m3fn_scaled.safetensors - +**text_encoders** - - **clip_vision** -- clip_vision_h.safetensors - +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - **loras** -- lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors - +**clip_vision** - - **diffusion_models** -- Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors - +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) - - **vae** -- wan_2.1_vae.safetensors - +**loras** + +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) + +**diffusion_models** + +- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) + +**vae** + +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **模型存放位置** diff --git a/zh/tutorials/video/wan/wan-video.mdx b/zh/tutorials/video/wan/wan-video.mdx index 2bd273b2f..5ee12bf95 100644 --- a/zh/tutorials/video/wan/wan-video.mdx +++ b/zh/tutorials/video/wan/wan-video.mdx @@ -34,14 +34,14 @@ Wan2.1 Video 系列为阿里巴巴于 2025年2月开源的视频生成模型， 本篇指南涉及的所有模型你都可以在[这里](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)找到, 下面是本篇示例中将会使用到的共用的模型，你可以提前进行下载： 从**Text encoders** 选择一个版本进行下载， -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 @@ -64,7 +64,7 @@ ComfyUI/ ## Wan2.1 文生视频工作流 -在开始工作流前请下载 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下。 +在开始工作流前请下载 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下。 > 如果你需要其它的 t2v 精度版本，请访问[这里](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models)进行下载 @@ -101,7 +101,7 @@ ComfyUI/ ![Wan2.1 图生视频工作流 14B 480P Workflow 输入图片示例](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) #### 2. 模型下载 -请下载[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 +请下载[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 #### 3. 按步骤完成工作流的运行 @@ -129,7 +129,7 @@ ComfyUI/ #### 2. 模型下载 -请下载[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 +请下载[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 #### 3. 按步骤完成工作流的运行 diff --git a/zh/tutorials/video/wan/wan2-2-animate.mdx b/zh/tutorials/video/wan/wan2-2-animate.mdx index ce39bd583..23407f486 100644 --- a/zh/tutorials/video/wan/wan2-2-animate.mdx +++ b/zh/tutorials/video/wan/wan2-2-animate.mdx @@ -54,14 +54,13 @@ Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架 下载以下工作流文件并将其拖入 ComfyUI 以加载工作流。 - - - 在 Comfy 云上运行 - - - 下载 JSON 或在模板库中搜索 "Wan2.2 Animate" - - + +

下载工作流

+ + + +

在 Comfy 云上运行

+ 下载以下素材作为输入： @@ -77,40 +76,20 @@ Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架 ### 2. 模型链接 -**Diffusion Models** - - - - Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: 来自 Kijai 仓库的缩放 FP8 版本 - - - wan2.2_animate_14B_bf16.safetensors: 原始 bf16 模型权重 - - - -**CLIP Vision** - - - clip_vision_h.safetensors: CLIP Vision 编码器 - - -**LoRAs** - - - lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4 步加速 LoRA - - -**VAE** +**diffusion_models** +- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) 这是来自 Kijai 仓库的模型 +- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 原始模型权重 - - wan_2.1_vae.safetensors: 用于编码和解码的 Wan2.1 VAE - +**clip_visions** +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) -**Text Encoders** +**loras** +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 这是一个 4 步的加速 lora +**vae** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - umt5_xxl_fp8_e4m3fn_scaled.safetensors: 缩放 FP8 文本编码器 - +**text_encoders** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan2-2-fun-camera.mdx b/zh/tutorials/video/wan/wan2-2-fun-camera.mdx index 31fb21cf4..8702967cb 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -19,6 +19,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [🤗Wan2.2-Fun-A14B-Control-Camera](https://huggingface.co/alibaba-pai/Wan2.2-Fun-A14B-Control-Camera) - 代码仓库：[VideoX-Fun](https://github.com/aigc-apps/VideoX-Fun) + ## Wan2.2 Fun Camera Control 相机控制视频生成工作流示例 @@ -36,6 +37,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 由于使用了4 步 LoRA 对于初次使用工作流的用户体验较好， 但可能导致生成的视频动态会有损失， 我们默认启用了使用了加速 LoRA 版本，如果你需要启用另一组的工作流，框选后使用 **Ctrl+B** 即可启用 + ### 1. 工作流及素材下载 下载下面的视频或者 JSON 文件并拖入 ComfyUI 中以加载对应的工作流，工作流会提示下载模型 @@ -46,60 +48,32 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "Wan2.2 Fun Camera" - - + +

下载 JSON 格式工作流

+ 请下载下面的图片，我们将作为输入。 - - - 视频生成的起始帧。下载并使用此图片，或替换为您自己的图片。 - - +![输入起始图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/input.jpg) ### 2. 模型链接 下面的模型你可以在 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 找到 -**Diffusion Models** - - - - Wan2.2 Fun Camera 高噪声扩散模型 - - - Wan2.2 Fun Camera 低噪声扩散模型 - - +**Diffusion Model** +- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) -**Wan2.2-Lightning LoRA（可选，用于加速）** - - - - 高噪声模型用的 4 步加速 LoRA - - - 低噪声模型用的 4 步加速 LoRA - - +**Wan2.2-Lightning LoRA (可选，用于加速)** +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - 用于编码/解码的 Wan2.1 VAE - - -**Text Encoder** +**Text Encoder** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - - FP8 缩放版文本编码器 - File save location @@ -118,6 +92,7 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` + ### 3. 按步骤完成工作流 ![Wan2.2 Fun Camera Control 工作流步骤](/images/tutorial/video/wan/wan_2.2_14b_fun_camera.jpg) diff --git a/zh/tutorials/video/wan/wan2-2-fun-control.mdx b/zh/tutorials/video/wan/wan2-2-fun-control.mdx index 147242751..1447efb01 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-control.mdx @@ -58,27 +58,27 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 1. 工作流及素材下载 -更新您的 ComfyUI 到最新版本，然后下载工作流文件并拖入 ComfyUI 中，或在模板库中通过 `Workflow` → `Browse Templates` → `Video` 找到 "Wan2.2 Fun Control"。 +下载下面的视频或者 JSON 文件并拖入 ComfyUI 中以加载对应的工作流 - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 文件，或在模板库中搜索 "Wan2.2 Fun Control" - - + + + +

下载 JSON 格式工作流

+ 请下载下面的图片及视频，我们将作为输入。 - - - 视频生成的起始帧。下载并使用此图片，或替换为您自己的图片。 - - - 预处理后的姿态控制视频。下载并使用此视频，或替换为您自己的视频。 - - +![输入起始图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/input.jpg) + + > 这里我们使用了经过预处理的视频, 可以直接用于控制视频生成 @@ -86,39 +86,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面的模型你可以在 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 找到 -**Diffusion Models** - - - - wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors：高噪声扩散模型 - - - wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors：低噪声扩散模型 - - +**Diffusion Model** +- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) -**Wan2.2-Lightning LoRA（可选，用于加速）** - - - - wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors：高噪声 4 步加速 LoRA - - - wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors：低噪声 4 步加速 LoRA - - +** Wan2.2-Lightning LoRA (可选，用于加速)** +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - wan_2.1_vae.safetensors：Wan2.1 VAE，用于编码/解码 - - -**Text Encoder** - - - umt5_xxl_fp8_e4m3fn_scaled.safetensors：缩放 FP8 文本编码器 - +**Text Encoder** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) File save location diff --git a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx index 97d98b2aa..50dd30fa8 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -61,14 +61,13 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 或者，在更新 ComfyUI 至最新版本后，下载下面的工作流并拖入 ComfyUI 中加载。 - - - 下载 JSON 或在模板库中搜索 "Wan2.2 Fun Inp" - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ + + +

在 Comfy 云上运行

+ 使用下面的素材作为首尾帧 @@ -77,39 +76,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 2. 手动下载模型 -**Diffusion Models** - - - - wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: 用于首尾帧修复的高噪点扩散模型 - - - wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: 用于首尾帧修复的低噪点扩散模型 - - - -**Lightning LoRA（可选，用于加速）** +**Diffusion Model** +- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) - - - wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 用于高噪点模型的 4 步加速 LoRA - - - wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 用于低噪点模型的 4 步加速 LoRA - - +**Lightning LoRA (可选，用于加速)** +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - wan_2.1_vae.safetensors: 用于编码和解码的 Wan2.1 VAE - - -**Text Encoder** - - - umt5_xxl_fp8_e4m3fn_scaled.safetensors: 缩放后的 FP8 文本编码器 - +**Text Encoder** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan2-2-s2v.mdx b/zh/tutorials/video/wan/wan2-2-s2v.mdx index 34520f445..447201527 100644 --- a/zh/tutorials/video/wan/wan2-2-s2v.mdx +++ b/zh/tutorials/video/wan/wan2-2-s2v.mdx @@ -33,58 +33,37 @@ Wan2.2 S2V 模型仓库：[Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - - - 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索 "Wan2.2 S2V" - - + +

Download JSON Workflow

+ + + +

Run on Comfy Cloud

+ 下载下面的图片及音频作为输入: +![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - - - 下载默认输入图片，或使用你自己的图片。 - - - 下载默认输入音频，或使用你自己的音频。 - - + +

下载输入音频

+ ### 2. 模型链接 你可以在 [我们的仓库](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 中找到所有模型。 -**diffusion_models** - - - - FP8 scaled diffusion model。放置在 ComfyUI/models/diffusion_models/ - - - BF16 diffusion model。放置在 ComfyUI/models/diffusion_models/ - - +**diffusion_models** +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) **audio_encoders** - - - Audio encoder model。放置在 ComfyUI/models/audio_encoders/ - +- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) **vae** +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - - Wan2.1 VAE model。放置在 ComfyUI/models/vae/ - - -**text_encoders** - - - FP8 scaled text encoder。放置在 ComfyUI/models/text_encoders/ - +**text_encoders** +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -113,14 +92,8 @@ ComfyUI/ 你可以在 [这里](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) 找到两种模型： - - - FP8 scaled diffusion model - - - BF16 diffusion model - - +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) 本模板使用 `wan2.2_s2v_14B_fp8_scaled.safetensors`，它需要更少的显存。但你可以尝试 `wan2.2_s2v_14B_bf16.safetensors` 来减少质量损失。 diff --git a/zh/tutorials/video/wan/wan2_2.mdx b/zh/tutorials/video/wan/wan2_2.mdx index 72e167ffd..3c2588525 100644 --- a/zh/tutorials/video/wan/wan2_2.mdx +++ b/zh/tutorials/video/wan/wan2_2.mdx @@ -101,25 +101,24 @@ Wan2.2 5B 版本配合 ComfyUI 原生 offloading功能，能很好地适配 8GB src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > - - - 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 5B" - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ + + +

Run on Comfy Cloud

+ ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) +- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors) **VAE** -- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan2.2_vae.safetensors) +- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -157,26 +156,25 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > - - - 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 14B T2V" - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ + + +

Run on Comfy Cloud

+ ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -217,14 +215,13 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > - - - 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 14B I2V" - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ + + +

Run on Comfy Cloud

+ 你可以使用下面的图片作为输入 ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -232,14 +229,14 @@ ComfyUI/ ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) -- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) +- [wan2.2_i2v_high_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_high_noise_14B_fp16.safetensors) +- [wan2.2_i2v_low_noise_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_i2v_low_noise_14B_fp16.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ @@ -277,14 +274,13 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > - - - 下载 JSON 格式工作流，或在模板库中搜索 "Wan2.2 14B FLF2V" - - - 在 Comfy Cloud 中打开 - - + +

下载 JSON 格式工作流

+ + + +

Run on Comfy Cloud

+ 下载下面的素材作为输入 diff --git a/zh/tutorials/video/zai/scail2.mdx b/zh/tutorials/video/zai/scail2.mdx index 03684d603..ff1e51c4d 100644 --- a/zh/tutorials/video/zai/scail2.mdx +++ b/zh/tutorials/video/zai/scail2.mdx @@ -105,23 +105,23 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 所需模型 **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders**（选择其一） -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### 文件存放位置 From 1af7f02635642d32f210b8defce11b0f41ba4587 Mon Sep 17 00:00:00 2001 From: lin-bot23 Date: Fri, 24 Jul 2026 17:00:40 +0800 Subject: [PATCH 10/19] i18n: sync zh/ja/ko translations for tutorial Card component standardisation --- ja/tutorials/3d/hunyuan3D-2.mdx | 48 ++++-- ja/tutorials/3d/triposplat.mdx | 22 ++- ja/tutorials/flux/flux-1-controlnet.mdx | 47 +++-- ja/tutorials/flux/flux-1-fill-dev.mdx | 41 +++-- ja/tutorials/flux/flux-1-kontext-dev.mdx | 24 ++- ja/tutorials/flux/flux-1-text-to-image.mdx | 70 +++++--- ja/tutorials/flux/flux-1-uso.mdx | 21 ++- ja/tutorials/flux/flux-2-klein.mdx | 19 +- ja/tutorials/flux/flux1-krea-dev.mdx | 36 ++-- ja/tutorials/image/hidream/hidream-e1.mdx | 24 ++- ja/tutorials/image/hidream/hidream-i1.mdx | 50 ++++-- .../newbie-image/newbie-image-exp-0-1.mdx | 37 ++-- ja/tutorials/image/omnigen/omnigen2.mdx | 29 ++-- ja/tutorials/image/ovis/ovis-image.mdx | 32 ++-- ja/tutorials/image/pixeldit/pixeldit.mdx | 9 +- ja/tutorials/image/qwen/qwen-image-2512.mdx | 19 +- .../image/qwen/qwen-image-edit-2511.mdx | 34 ++-- ja/tutorials/image/qwen/qwen-image-edit.mdx | 37 ++-- .../image/qwen/qwen-image-layered.mdx | 30 ++-- ja/tutorials/image/qwen/qwen-image.mdx | 62 ++++--- ja/tutorials/image/z-image/z-image-turbo.mdx | 13 +- .../partner-nodes/anthropic/claude.mdx | 11 +- .../partner-nodes/beeble/beeble-switchx.mdx | 21 ++- .../partner-nodes/bria/background-removal.mdx | 7 +- .../bytedance/seed-audio-1-0.mdx | 9 +- .../bytedance/seedance-2-0-real-human.mdx | 5 +- .../partner-nodes/bytedance/seedance-2-0.mdx | 7 +- .../bytedance/seedream-5-lite.mdx | 7 +- .../bytedance/seedream-5-pro.mdx | 9 +- .../google/gemini-omni-flash.mdx | 5 +- ja/tutorials/partner-nodes/google/gemini.mdx | 17 +- .../partner-nodes/google/nano-banana-2.mdx | 5 +- .../grok/grok-imagine-video-1-5.mdx | 2 + .../happyhorse/happyhorse1-0.mdx | 11 +- .../happyhorse/happyhorse1-1.mdx | 9 +- .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 11 +- .../hunyuan3d/model-generation.mdx | 9 +- .../partner-nodes/ideogram/ideogram-v4.mdx | 5 +- .../kling/kling-motion-control.mdx | 22 ++- .../partner-nodes/luma/luma-uni-1.mdx | 5 +- ja/tutorials/partner-nodes/meshy/meshy-6.mdx | 11 +- ja/tutorials/partner-nodes/openai/chat.mdx | 17 +- .../partner-nodes/openai/gpt-image-2.mdx | 5 +- ja/tutorials/partner-nodes/pricing.mdx | 7 +- .../partner-nodes/recraft/recraft-v4.mdx | 7 +- .../partner-nodes/rodin/model-generation.mdx | 49 ++++-- ja/tutorials/partner-nodes/topaz/astra-2.mdx | 2 +- .../partner-nodes/tripo/model-generation.mdx | 51 ++++-- .../partner-nodes/tripo/tripo-3-1.mdx | 5 +- ja/tutorials/partner-nodes/wan/wan2-7.mdx | 11 +- ja/tutorials/utility/moge.mdx | 13 +- ja/tutorials/video/bytedance/bernini-r.mdx | 21 +-- .../cosmos/cosmos-predict2-video2world.mdx | 19 +- .../video/hunyuan/hunyuan-video-1-5.mdx | 22 ++- ja/tutorials/video/hunyuan/hunyuan-video.mdx | 25 +-- ja/tutorials/video/kandinsky/kandinsky-5.mdx | 72 ++++++-- ja/tutorials/video/ltx/ltx-2-3.mdx | 10 ++ ja/tutorials/video/ltx/ltx-2.mdx | 107 +++++++++--- ja/tutorials/video/ltxv.mdx | 31 +++- ja/tutorials/video/wan/fun-camera.mdx | 39 +++-- ja/tutorials/video/wan/fun-control.mdx | 23 +-- ja/tutorials/video/wan/fun-inp.mdx | 19 +- ja/tutorials/video/wan/vace.mdx | 34 ++-- ja/tutorials/video/wan/wan-alpha.mdx | 9 +- ja/tutorials/video/wan/wan-ati.mdx | 17 +- ja/tutorials/video/wan/wan-causal-forcing.mdx | 17 +- ja/tutorials/video/wan/wan-dancer.mdx | 21 +-- ja/tutorials/video/wan/wan-flf.mdx | 12 +- ja/tutorials/video/wan/wan-move.mdx | 59 +++++-- ja/tutorials/video/wan/wan-video.mdx | 27 +-- ja/tutorials/video/wan/wan2-2-animate.mdx | 42 +++-- ja/tutorials/video/wan/wan2-2-fun-camera.mdx | 46 +++-- ja/tutorials/video/wan/wan2-2-fun-control.mdx | 52 ++++-- ja/tutorials/video/wan/wan2-2-fun-inp.mdx | 67 ++++++-- ja/tutorials/video/wan/wan2-2-s2v.mdx | 67 ++++++-- ja/tutorials/video/wan/wan2_2.mdx | 162 +++++++++++++----- ja/tutorials/video/zai/scail2.mdx | 25 +-- ko/tutorials/3d/hunyuan3D-2.mdx | 36 ++-- ko/tutorials/3d/triposplat.mdx | 103 ++--------- ko/tutorials/flux/flux-1-controlnet.mdx | 39 +++-- ko/tutorials/flux/flux-1-fill-dev.mdx | 27 ++- ko/tutorials/flux/flux-1-kontext-dev.mdx | 20 ++- ko/tutorials/flux/flux-1-text-to-image.mdx | 75 +++++--- ko/tutorials/flux/flux-1-uso.mdx | 27 ++- ko/tutorials/flux/flux-2-klein.mdx | 19 +- ko/tutorials/flux/flux1-krea-dev.mdx | 22 ++- ko/tutorials/image/hidream/hidream-e1.mdx | 20 ++- ko/tutorials/image/hidream/hidream-i1.mdx | 44 +++-- .../newbie-image/newbie-image-exp-0-1.mdx | 23 ++- ko/tutorials/image/omnigen/omnigen2.mdx | 29 ++-- ko/tutorials/image/ovis/ovis-image.mdx | 18 +- ko/tutorials/image/pixeldit/pixeldit.mdx | 9 +- ko/tutorials/image/qwen/qwen-image-2512.mdx | 26 ++- .../image/qwen/qwen-image-edit-2511.mdx | 22 +-- ko/tutorials/image/qwen/qwen-image-edit.mdx | 25 +-- .../image/qwen/qwen-image-layered.mdx | 30 ++-- ko/tutorials/image/qwen/qwen-image.mdx | 93 +++++----- ko/tutorials/image/z-image/z-image-turbo.mdx | 13 +- .../partner-nodes/anthropic/claude.mdx | 14 +- .../partner-nodes/beeble/beeble-switchx.mdx | 27 ++- .../partner-nodes/bria/background-removal.mdx | 17 +- .../bytedance/seed-audio-1-0.mdx | 9 +- .../bytedance/seedance-2-0-real-human.mdx | 21 ++- .../partner-nodes/bytedance/seedance-2-0.mdx | 43 ++--- .../bytedance/seedream-5-lite.mdx | 31 ++-- .../bytedance/seedream-5-pro.mdx | 30 ++-- .../google/gemini-omni-flash.mdx | 11 +- ko/tutorials/partner-nodes/google/gemini.mdx | 13 +- .../partner-nodes/google/nano-banana-2.mdx | 11 +- .../grok/grok-imagine-video-1-5.mdx | 8 +- .../happyhorse/happyhorse1-0.mdx | 59 +++---- .../happyhorse/happyhorse1-1.mdx | 15 +- .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 11 +- .../hunyuan3d/model-generation.mdx | 9 +- .../partner-nodes/ideogram/ideogram-v4.mdx | 9 +- .../kling/kling-motion-control.mdx | 31 +++- .../partner-nodes/luma/luma-uni-1.mdx | 29 ++-- ko/tutorials/partner-nodes/meshy/meshy-6.mdx | 29 +++- ko/tutorials/partner-nodes/openai/chat.mdx | 13 +- .../partner-nodes/openai/gpt-image-2.mdx | 5 +- ko/tutorials/partner-nodes/pricing.mdx | 7 +- .../partner-nodes/recraft/recraft-v4.mdx | 11 +- .../partner-nodes/rodin/model-generation.mdx | 87 ++++++---- ko/tutorials/partner-nodes/topaz/astra-2.mdx | 16 +- .../partner-nodes/tripo/model-generation.mdx | 46 +++-- .../partner-nodes/tripo/tripo-3-1.mdx | 47 ++--- ko/tutorials/partner-nodes/wan/wan2-7.mdx | 59 +++---- ko/tutorials/utility/moge.mdx | 13 +- ko/tutorials/video/bytedance/bernini-r.mdx | 21 +-- .../cosmos/cosmos-predict2-video2world.mdx | 15 +- .../video/hunyuan/hunyuan-video-1-5.mdx | 22 ++- ko/tutorials/video/hunyuan/hunyuan-video.mdx | 25 +-- ko/tutorials/video/kandinsky/kandinsky-5.mdx | 45 +++-- ko/tutorials/video/ltx/ltx-2-3.mdx | 10 ++ ko/tutorials/video/ltx/ltx-2.mdx | 78 ++++----- ko/tutorials/video/ltxv.mdx | 22 ++- ko/tutorials/video/wan/fun-camera.mdx | 112 ++++++++---- ko/tutorials/video/wan/fun-control.mdx | 23 +-- ko/tutorials/video/wan/fun-inp.mdx | 19 +- ko/tutorials/video/wan/vace.mdx | 30 ++-- ko/tutorials/video/wan/wan-alpha.mdx | 9 +- ko/tutorials/video/wan/wan-ati.mdx | 17 +- ko/tutorials/video/wan/wan-causal-forcing.mdx | 17 +- ko/tutorials/video/wan/wan-dancer.mdx | 21 +-- ko/tutorials/video/wan/wan-flf.mdx | 12 +- ko/tutorials/video/wan/wan-move.mdx | 66 ++++--- ko/tutorials/video/wan/wan-video.mdx | 29 ++-- ko/tutorials/video/wan/wan2-2-animate.mdx | 28 ++- ko/tutorials/video/wan/wan2-2-fun-camera.mdx | 23 +-- ko/tutorials/video/wan/wan2-2-fun-control.mdx | 24 +-- ko/tutorials/video/wan/wan2-2-fun-inp.mdx | 29 ++-- ko/tutorials/video/wan/wan2-2-s2v.mdx | 37 ++-- ko/tutorials/video/wan/wan2_2.mdx | 75 ++++---- ko/tutorials/video/zai/scail2.mdx | 25 +-- zh/tutorials/3d/hunyuan3D-2.mdx | 48 ++++-- zh/tutorials/3d/triposplat.mdx | 15 +- zh/tutorials/flux/flux-1-controlnet.mdx | 49 ++++-- zh/tutorials/flux/flux-1-fill-dev.mdx | 43 +++-- zh/tutorials/flux/flux-1-kontext-dev.mdx | 26 ++- zh/tutorials/flux/flux-1-text-to-image.mdx | 115 +++++-------- zh/tutorials/flux/flux-1-uso.mdx | 21 ++- zh/tutorials/flux/flux-2-klein.mdx | 19 +- zh/tutorials/flux/flux1-krea-dev.mdx | 26 +-- zh/tutorials/image/hidream/hidream-e1.mdx | 24 ++- zh/tutorials/image/hidream/hidream-i1.mdx | 54 ++++-- .../newbie-image/newbie-image-exp-0-1.mdx | 29 ++-- zh/tutorials/image/omnigen/omnigen2.mdx | 29 ++-- zh/tutorials/image/ovis/ovis-image.mdx | 24 +-- zh/tutorials/image/pixeldit/pixeldit.mdx | 9 +- zh/tutorials/image/qwen/qwen-image-2512.mdx | 26 +-- .../image/qwen/qwen-image-edit-2511.mdx | 30 ++-- zh/tutorials/image/qwen/qwen-image-edit.mdx | 27 +-- .../image/qwen/qwen-image-layered.mdx | 30 ++-- zh/tutorials/image/qwen/qwen-image.mdx | 91 +++++----- zh/tutorials/image/z-image/z-image-turbo.mdx | 13 +- .../partner-nodes/anthropic/claude.mdx | 17 +- .../partner-nodes/beeble/beeble-switchx.mdx | 20 +-- .../partner-nodes/bria/background-removal.mdx | 17 +- .../bytedance/seed-audio-1-0.mdx | 9 +- .../bytedance/seedance-2-0-real-human.mdx | 53 +++++- .../partner-nodes/bytedance/seedance-2-0.mdx | 7 +- .../bytedance/seedream-5-lite.mdx | 22 ++- .../bytedance/seedream-5-pro.mdx | 34 +++- .../google/gemini-omni-flash.mdx | 11 +- zh/tutorials/partner-nodes/google/gemini.mdx | 19 +- .../partner-nodes/google/nano-banana-2.mdx | 5 +- .../grok/grok-imagine-video-1-5.mdx | 2 + .../happyhorse/happyhorse1-0.mdx | 54 ++++-- .../happyhorse/happyhorse1-1.mdx | 9 +- .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 11 +- .../hunyuan3d/model-generation.mdx | 9 +- .../partner-nodes/ideogram/ideogram-v4.mdx | 5 +- .../kling/kling-motion-control.mdx | 36 +++- .../partner-nodes/luma/luma-uni-1.mdx | 48 +++++- zh/tutorials/partner-nodes/meshy/meshy-6.mdx | 45 +++-- zh/tutorials/partner-nodes/openai/chat.mdx | 19 +- .../partner-nodes/openai/gpt-image-2.mdx | 5 +- zh/tutorials/partner-nodes/pricing.mdx | 7 +- .../partner-nodes/recraft/recraft-v4.mdx | 7 +- .../partner-nodes/rodin/model-generation.mdx | 34 ++-- zh/tutorials/partner-nodes/topaz/astra-2.mdx | 2 +- .../partner-nodes/tripo/model-generation.mdx | 39 +++-- .../partner-nodes/tripo/tripo-3-1.mdx | 42 ++++- zh/tutorials/partner-nodes/wan/wan2-7.mdx | 52 ++++-- zh/tutorials/utility/moge.mdx | 13 +- zh/tutorials/video/bytedance/bernini-r.mdx | 21 +-- .../cosmos/cosmos-predict2-video2world.mdx | 19 +- .../video/hunyuan/hunyuan-video-1-5.mdx | 22 ++- zh/tutorials/video/hunyuan/hunyuan-video.mdx | 25 +-- zh/tutorials/video/kandinsky/kandinsky-5.mdx | 72 ++++++-- zh/tutorials/video/ltx/ltx-2-3.mdx | 10 ++ zh/tutorials/video/ltx/ltx-2.mdx | 5 +- zh/tutorials/video/ltxv.mdx | 24 ++- zh/tutorials/video/wan/fun-camera.mdx | 45 +++-- zh/tutorials/video/wan/fun-control.mdx | 23 +-- zh/tutorials/video/wan/fun-inp.mdx | 19 +- zh/tutorials/video/wan/vace.mdx | 30 ++-- zh/tutorials/video/wan/wan-alpha.mdx | 9 +- zh/tutorials/video/wan/wan-ati.mdx | 17 +- zh/tutorials/video/wan/wan-causal-forcing.mdx | 17 +- zh/tutorials/video/wan/wan-dancer.mdx | 21 +-- zh/tutorials/video/wan/wan-flf.mdx | 12 +- zh/tutorials/video/wan/wan-move.mdx | 67 ++++++-- zh/tutorials/video/wan/wan-video.mdx | 25 +-- zh/tutorials/video/wan/wan2-2-animate.mdx | 68 ++++++-- zh/tutorials/video/wan/wan2-2-fun-camera.mdx | 25 +-- zh/tutorials/video/wan/wan2-2-fun-control.mdx | 28 +-- zh/tutorials/video/wan/wan2-2-fun-inp.mdx | 78 +++++++-- zh/tutorials/video/wan/wan2-2-s2v.mdx | 53 ++++-- zh/tutorials/video/wan/wan2_2.mdx | 112 +++++++----- zh/tutorials/video/zai/scail2.mdx | 25 +-- 231 files changed, 4082 insertions(+), 2580 deletions(-) diff --git a/ja/tutorials/3d/hunyuan3D-2.mdx b/ja/tutorials/3d/hunyuan3D-2.mdx index 35f163849..928e5ee5d 100644 --- a/ja/tutorials/3d/hunyuan3D-2.mdx +++ b/ja/tutorials/3d/hunyuan3D-2.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Hunyuan3D-2 例" description: "このガイドでは、ComfyUI で Hunyuan3D-2 を使用して 3D アセットを生成する方法を説明します。" sidebarTitle: "Hunyuan3D-2" -translationSourceHash: 61db1e41 +translationSourceHash: 8a90bee3 translationFrom: tutorials/3d/hunyuan3D-2.mdx translationBlockHashes: "_intro": 0078e687 - "ComfyUI Hunyuan3D-2mv Workflow Example": 1a1a8896 - "Hunyuan3D-2mv-turbo Workflow": afdb6432 - "Hunyuan3D-2 Single View Workflow": 1f454a64 + "ComfyUI Hunyuan3D-2mv Workflow Example": 62734596 + "Hunyuan3D-2mv-turbo Workflow": 7c6bb309 + "Hunyuan3D-2 Single View Workflow": 72e10fde "Community Resources": d12dc58a "Hunyuan3D 2.0 Open-Source Model Series": 363ff037 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # Hunyuan3D 2.0 はじめに @@ -57,9 +58,14 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して - -

Run on Comfy Cloud

- + + + Comfy Cloud でこのワークフローを即座に実行 + + + ワークフロー JSON ファイルをダウンロード + + ### 1. ワークフロー @@ -81,7 +87,7 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます ``` ComfyUI/ @@ -104,9 +110,14 @@ ComfyUI/ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを使用して 3D モデルを生成します。このモデルは Hunyuan3D-2mv のステップ蒸留バージョンで、より高速な 3D モデル生成を可能にします。このバージョンのワークフローでは、`cfg` を 1.0 に設定し、`flux guidance` ノードを追加して `distilled cfg` 生成を制御します。 - -

Run on Comfy Cloud

- + + + Comfy Cloud でこのワークフローを即座に実行 + + + ワークフロー JSON ファイルをダウンロード + + ### 1. ワークフロー @@ -125,7 +136,7 @@ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます ``` ComfyUI/ @@ -146,9 +157,14 @@ ComfyUI/ Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D モデルを生成します。このモデルはマルチビューモデルではありません。このワークフローでは、`Hunyuan3Dv2ConditioningMultiView` ノードの代わりに `Hunyuan3Dv2Conditioning` ノードを使用します。 - -

Run on Comfy Cloud

- + + + Comfy Cloud でこのワークフローを即座に実行 + + + ワークフロー JSON ファイルをダウンロード + + ### 1. ワークフロー @@ -163,7 +179,7 @@ Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D 以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます ``` ComfyUI/ diff --git a/ja/tutorials/3d/triposplat.mdx b/ja/tutorials/3d/triposplat.mdx index 3c1dbffce..5cebf38ee 100644 --- a/ja/tutorials/3d/triposplat.mdx +++ b/ja/tutorials/3d/triposplat.mdx @@ -2,10 +2,10 @@ title: "TripoSplat 画像からガウシアンスプラット ComfyUI ワークフロー例" description: "TripoSplat を使用して、単一の 2D 画像から高品質な 3D ガウシアンスプラット表現を生成します。密度とレンダリング予算を制御可能。" sidebarTitle: "TripoSplat" -translationSourceHash: 5ca7b463 +translationSourceHash: 95c7decf translationFrom: tutorials/3d/triposplat.mdx translationBlockHashes: - "_intro": 07dba1db + "_intro": 77f8628a "How it works": 9da65e03 "Workflow node guide": 83cd15a0 "Steps to run": e4ad7fa4 @@ -14,6 +14,7 @@ translationBlockHashes: --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **TripoSplat** は、単一の 2D 画像から直接 **3D ガウシアンスプラット（Gaussian splat）** 表現を生成するオープンソースモデルです。VAST-AI によって開発され、オープンソースライセンスで公開されています。 @@ -24,9 +25,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - + + + Comfy Cloud でこのワークフローを即座に実行 + + JSON をダウンロード、またはテンプレートライブラリで "TripoSplat" を検索 + ## 仕組み @@ -106,23 +112,23 @@ TripoSplat は **フィードフォワードアーキテクチャ** を使用し TripoSplat モデルと必要なファイルをダウンロードします。対応する `models/` サブディレクトリに配置してください。 - + triposplat_fp16.safetensors — TripoSplat 拡散モデルチェックポイント - + triposplat_vae_decoder_fp16.safetensors — VAE デコーダー - + flux2-vae.safetensors — Flux.2 VAE、潜在表現エンコード用 - + dino_v3_vit_h.safetensors — CLIP ビジョンエンコーダー（DINOv2） - + birefnet.safetensors — 前処理用の背景除去モデル diff --git a/ja/tutorials/flux/flux-1-controlnet.mdx b/ja/tutorials/flux/flux-1-controlnet.mdx index 39c15df7e..8fd02ca11 100644 --- a/ja/tutorials/flux/flux-1-controlnet.mdx +++ b/ja/tutorials/flux/flux-1-controlnet.mdx @@ -2,17 +2,18 @@ title: "ComfyUI Flux.1 ControlNet の使用例" description: "本ガイドでは、Flux.1 ControlNet を用いたワークフローの使用例を紹介します。" sidebarTitle: "Flux.1 ControlNet" -translationSourceHash: 7511ec98 +translationSourceHash: a02fc86f translationFrom: tutorials/flux/flux-1-controlnet.mdx translationBlockHashes: "_intro": a69fbfae "FLUX.1 ControlNet Model Introduction": c242319b - "FLUX.1-Canny-dev Complete Version Workflow": c8623b99 - "FLUX.1-Depth-dev-lora Workflow": bf2160c6 + "FLUX.1-Canny-dev Complete Version Workflow": 0c4240ac + "FLUX.1-Depth-dev-lora Workflow": ebfe0fb5 "Community Versions of Flux Controlnets": 29728dcf --- + ![Flux.1 Canny Controlnet](/images/tutorial/flux/flux-1-canny-controlnet.png) ![Flux.1 Depth Controlnet](/images/tutorial/flux/flux-1-depth-controlnet.png) @@ -49,9 +50,14 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 ## FLUX.1-Canny-dev 完全版ワークフロー - -

Comfy Cloud で実行

- + + + Download JSON or search "Flux.1 Canny" in Template Library + + + Comfy Cloud で開く + + ### 1. ワークフローおよび関連アセット @@ -72,10 +78,10 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） ファイルの保存先ディレクトリ構成： ``` @@ -115,9 +121,14 @@ ComfyUI/ ## FLUX.1-Depth-dev-lora ワークフロー - -

Comfy Cloud で実行

- + + + Download JSON or search "Flux.1 Depth LoRA" in Template Library + + + Comfy Cloud で開く + + LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを追加したものであり、[Flux ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image) と比較して、対応する LoRA モデルを読み込むためのノードが追加されています。 @@ -138,11 +149,11 @@ LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) ファイルの保存先ディレクトリ構成： ``` diff --git a/ja/tutorials/flux/flux-1-fill-dev.mdx b/ja/tutorials/flux/flux-1-fill-dev.mdx index 9e8b9f12e..d9517d89f 100644 --- a/ja/tutorials/flux/flux-1-fill-dev.mdx +++ b/ja/tutorials/flux/flux-1-fill-dev.mdx @@ -2,17 +2,18 @@ title: "ComfyUI Flux.1 fill dev の使用例" description: "本ガイドでは、Flux.1 fill dev を用いた Inpainting（画像修復）および Outpainting（画像拡張）ワークフローの構築方法を解説します。" sidebarTitle: "Flux.1 fill dev" -translationSourceHash: 7774a137 +translationSourceHash: 5cba3176 translationFrom: tutorials/flux/flux-1-fill-dev.mdx translationBlockHashes: "_intro": 73fb6ba4 "Introduction to Flux.1 fill dev Model": 6dfbcfd4 - "Flux.1 Fill dev and related models installation": f1c9d9b6 - "Flux.1 Fill dev inpainting workflow": bc39310d - "Flux.1 Fill dev Outpainting Workflow": 13a777f6 + "Flux.1 Fill dev and related models installation": cf7747cd + "Flux.1 Fill dev inpainting workflow": 9589641c + "Flux.1 Fill dev Outpainting Workflow": dd827926 --- + ![Flux.1 fill dev](/images/tutorial/flux/flux-fill-dev-demo.jpeg) ## Flux.1 fill dev モデルの概要 @@ -39,10 +40,10 @@ Inpainting や Outpainting のワークフローについてまだご存じな ![Flux Agreement](/images/tutorial/flux/flux1_fill_dev_agreement.jpg) 必要なモデル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/resolve/main/flux1-fill-dev.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-fill-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev/blob/main/flux1-fill-dev.safetensors?download=true) ファイルの保存場所： ``` @@ -61,13 +62,23 @@ ComfyUI/ ### 1. Inpainting ワークフローおよび関連アセット - -

ワークフロー画像をダウンロード

- - - -

Comfy Cloud で実行

- + + + Download JSON or search "flux_fill_inpaint" in Template Library + + + Comfy Cloud で開く + + + + + + Download JSON or search "flux_fill_inpaint" in Template Library + + + Comfy Cloud で開く + + 以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 ![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) diff --git a/ja/tutorials/flux/flux-1-kontext-dev.mdx b/ja/tutorials/flux/flux-1-kontext-dev.mdx index b9537bda5..1831cd315 100644 --- a/ja/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ja/tutorials/flux/flux-1-kontext-dev.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Flux Kontext Dev ネイティブワークフローの例" description: "ComfyUI Flux Kontext Dev ネイティブワークフローの例。" sidebarTitle: "Flux.1 Kontext Dev" -translationSourceHash: f8c2bc9d +translationSourceHash: d2e6fb96 translationFrom: tutorials/flux/flux-1-kontext-dev.mdx translationBlockHashes: "_intro": 51c48128 - "About FLUX.1 Kontext Dev": 8f5451fd - "Model Download": 172ad3bc - "Flux.1 Kontext Dev Workflow": 7e22a0d2 + "About FLUX.1 Kontext Dev": bb2f00f7 + "Model Download": e6b40970 + "Flux.1 Kontext Dev Workflow": ffae9c25 --- + import PromptTechniques from "/snippets/ja/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -51,7 +52,7 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： **Diffusion Model（拡散モデル）** -- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/resolve/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) +- [flux1-dev-kontext_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/flux1-kontext-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-dev-kontext_fp8_scaled.safetensors) オリジナルの重み（weights）を使用したい場合は、Black Forest Labs の関連リポジトリからオリジナルモデルの重みを取得・利用できます。 @@ -62,7 +63,7 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： **Text Encoder（テキストエンコーダー）** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) または [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) または [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors) モデルの保存先 @@ -80,9 +81,14 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： ## Flux.1 Kontext Dev ワークフロー - -

Comfy Cloud で実行

- + + + Download JSON or search "Flux Kontext Dev" in Template Library + + + Comfy Cloud で開く + + このワークフローでは、編集対象の画像を読み込むために `Load Image(from output)` ノードを採用しており、編集後の画像を容易に取得・再利用できるため、複数回の反復編集がよりスムーズに行えます。 diff --git a/ja/tutorials/flux/flux-1-text-to-image.mdx b/ja/tutorials/flux/flux-1-text-to-image.mdx index 51b98dd27..4298fc725 100644 --- a/ja/tutorials/flux/flux-1-text-to-image.mdx +++ b/ja/tutorials/flux/flux-1-text-to-image.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Flux.1 テキストから画像へワークフローの例" description: "本ガイドでは、Flux.1 モデルについて簡潔に紹介し、フルバージョンおよび FP8 Checkpoint バージョンを含む、Flux.1 モデルを用いたテキストから画像への生成方法を解説します。" sidebarTitle: "Flux.1 テキストから画像へ" -translationSourceHash: 45e7a7c3 +translationSourceHash: 04546864 translationFrom: tutorials/flux/flux-1-text-to-image.mdx translationBlockHashes: "_intro": 36e44c17 - "Flux.1 Full Version Text-to-Image Example": 9eaf217e - "Flux.1 FP8 Checkpoint Version Text-to-Image Example": 14dfbf7e + "Flux.1 Full Version Text-to-Image Example": 78064d53 + "Flux.1 FP8 Checkpoint Version Text-to-Image Example": daf5c58e --- + ![Flux](/images/tutorial/flux/flux_example.png) Flux は、現時点で最も大規模なオープンソースのテキストから画像へ生成するモデルの一つであり、120億（12B）パラメータを有し、オリジナルファイルサイズは約23GBです。このモデルは、元 Stable Diffusion チームのメンバーによって設立された [Black Forest Labs](https://blackforestlabs.ai/) が開発しました。 Flux は、優れた画像品質と高い柔軟性で知られており、高品質かつ多様な画像を生成できます。 @@ -53,22 +54,27 @@ Flux は、優れた画像品質と高い柔軟性で知られており、高品 下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 ![Flux Dev オリジナルバージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - -

Comfy Cloud で実行

- + + + Comfy Cloud でこのワークフローを実行 + + + Download JSON or search "Flux.1 Dev" in Template Library + + #### 2. モデルの手動インストール - `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) のライセンス契約に同意する必要があります。 -- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 +- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 以下のモデルファイルをダウンロードしてください： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) 保存先のディレクトリ構成： ``` @@ -107,9 +113,14 @@ Flux の優れたプロンプト追従能力により、負のプロンプト（ ![Flux Schnell バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) - -

Comfy Cloud で実行

- + + + Comfy Cloud でこのワークフローを実行 + + + Download JSON or search "Flux.1 Schnell" in Template Library + + #### 2. モデルの手動インストール @@ -120,10 +131,10 @@ Flux の優れたプロンプト追従能力により、負のプロンプト（ 完全なモデルファイル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) ファイルの保存先ディレクトリ構成： ``` @@ -160,11 +171,22 @@ FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したも ![Flux Dev fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) - -

Comfy Cloud で実行

- - -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 + + + Comfy Cloud でこのワークフローを実行 + + + Download JSON or search "Flux.1 Dev FP8" in Template Library + + + Comfy Cloud でこのワークフローを実行 + + + Download JSON or search "Flux.1 Schnell FP8" in Template Library + + + +[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 対応する `Load Checkpoint` ノードが `flux1-dev-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 @@ -174,6 +196,6 @@ FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したも ![Flux Schnell fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) -[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/resolve/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 +[flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 対応する `Load Checkpoint` ノードが `flux1-schnell-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 diff --git a/ja/tutorials/flux/flux-1-uso.mdx b/ja/tutorials/flux/flux-1-uso.mdx index 9e8790c9a..5e2fa28b2 100644 --- a/ja/tutorials/flux/flux-1-uso.mdx +++ b/ja/tutorials/flux/flux-1-uso.mdx @@ -2,7 +2,7 @@ title: "ByteDance USO ComfyUI ネイティブワークフローの例" description: "ByteDance の USO モデルを用いた統一スタイル・主体駆動型生成" sidebarTitle: "ByteDance USO" -translationSourceHash: 85378cc8 +translationSourceHash: e41f0254 translationFrom: tutorials/flux/flux-1-uso.mdx --- @@ -40,9 +40,14 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff',

JSON ワークフローをダウンロード

- -

Comfy Cloud で実行

- + + + Download the workflow JSON and drag it into ComfyUI + + + Comfy Cloud でこのワークフローを実行 + + 以下の画像を入力画像として使用します。 @@ -52,18 +57,18 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', **checkpoints** -- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors) +- [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors) **loras** -- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) +- [uso-flux1-dit-lora-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/loras/uso-flux1-dit-lora-v1.safetensors) **model_patches** -- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/resolve/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) +- [uso-flux1-projector-v1.safetensors](https://huggingface.co/Comfy-Org/USO_1.0_Repackaged/blob/main/split_files/model_patches/uso-flux1-projector-v1.safetensors) **clip_visions** -- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors) +- [sigclip_vision_patch14_384.safetensors](https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors) 上記すべてのモデルをダウンロードし、以下のディレクトリ構造に配置してください： diff --git a/ja/tutorials/flux/flux-2-klein.mdx b/ja/tutorials/flux/flux-2-klein.mdx index c12f18487..bd9be116d 100644 --- a/ja/tutorials/flux/flux-2-klein.mdx +++ b/ja/tutorials/flux/flux-2-klein.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Flux.2 Klein 4B ガイド" description: "FLUX.2 [klein] 4B の概要と、ComfyUI におけるテキストから画像への生成および画像編集ワークフローの実行方法について解説します。" sidebarTitle: "Flux.2 Klein" -translationSourceHash: 25fd241d +translationSourceHash: 2b9ccba8 translationFrom: tutorials/flux/flux-2-klein.mdx translationBlockHashes: "_intro": 980a68e5 "About FLUX.2 [klein]": 77f716a7 - "Flux.2 Klein 4B Workflows": 153194d7 + "Flux.2 Klein 4B Workflows": 8c4a5de1 "Flux.2 Klein 4B Model Downloads": b438d16c - "Flux.2 Klein 9B Workflows": 6e7e8a01 + "Flux.2 Klein 9B Workflows": fbbce562 "Flux.2 Klein 9B Model Downloads": 0f38ad48 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -45,19 +46,19 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 ## Flux.2 Klein 4B モデルのダウンロード - + 4B モデル用のテキストエンコーダーです。 - + 拡散モデル（4B Base 版）。 - + 拡散モデル（4B 蒸留版）。 - + 4B モデル用のVAEです。 @@ -103,11 +104,11 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 拡散モデル（9B 蒸留版）。 - + 9B モデル用のテキストエンコーダーです。 - + 9B モデル用のVAEです。 diff --git a/ja/tutorials/flux/flux1-krea-dev.mdx b/ja/tutorials/flux/flux1-krea-dev.mdx index 827438849..1782f7754 100644 --- a/ja/tutorials/flux/flux1-krea-dev.mdx +++ b/ja/tutorials/flux/flux1-krea-dev.mdx @@ -2,7 +2,7 @@ title: "Flux.1 Krea Dev ComfyUI ワークフロー チュートリアル" description: "Black Forest Labs が Krea と共同で開発した、最高品質のオープンソース FLUX モデルです。独特な美意識と自然なディテールに重点を置き、「AIらしさ」を回避し、卓越したリアリズムと画像品質を実現します。" sidebarTitle: "Flux.1 Krea Dev" -translationSourceHash: e4ac73ab +translationSourceHash: 202e24cf translationFrom: tutorials/flux/flux1-krea-dev.mdx --- @@ -31,13 +31,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下の画像または JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 ![Flux Krea Dev ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Comfy Cloud でこのワークフローを実行 + + + Download JSON or search "Flux.1 Krea Dev" in Template Library + + + + + + Comfy Cloud でこのワークフローを実行 + + + Download JSON or search "Flux.1 Krea Dev" in Template Library + + #### 2. 手動によるモデルのインストール @@ -50,7 +60,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' より高品質な出力を求め、かつ十分な VRAM をお持ちの場合、オリジナルの重みファイルも試すことができます： -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) の利用規約に同意する必要があります。 @@ -59,12 +69,12 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以前に Flux 関連のワークフローをご利用済みの場合、以下のモデルは既に存在するため、再ダウンロードの必要はありません。 **テキストエンコーダー** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) — VRAM が 32GB を超える環境で推奨 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) — 低 VRAM 環境向け +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) — VRAM が 32GB を超える環境で推奨 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) — 低 VRAM 環境向け **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) ファイルの保存先： ``` diff --git a/ja/tutorials/image/hidream/hidream-e1.mdx b/ja/tutorials/image/hidream/hidream-e1.mdx index 47798c64a..bd9a0ee43 100644 --- a/ja/tutorials/image/hidream/hidream-e1.mdx +++ b/ja/tutorials/image/hidream/hidream-e1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI ネイティブ HiDream-E1、E1.1 ワークフローの例" sidebarTitle: "HiDream-e1" description: "本ガイドでは、ComfyUI ネイティブ版 HiDream-I1 のテキストから画像を生成するワークフローの例について、理解と実行方法を説明します。" -translationSourceHash: 76ea9fff +translationSourceHash: 7554c00c translationFrom: tutorials/image/hidream/hidream-e1.mdx translationBlockHashes: "_intro": 6829ae90 - "HiDream E1 and E1.1 Workflow Related Models": e648e936 - "HiDream E1.1 ComfyUI Native Workflow Example": 4134dfe8 - "HiDream E1 ComfyUI Native Workflow Example": 778b966a + "HiDream E1 and E1.1 Workflow Related Models": 25f9249b + "HiDream E1.1 ComfyUI Native Workflow Example": 4437f406 + "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ![HiDream-E1 デモ](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -40,8 +41,8 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン **Diffusion モデル** 両方のモデルを同時にダウンロードする必要はありません。E1.1 は E1 をベースとした改良版であり、実際のテスト結果から、品質およびパフォーマンスの両面で E1 を大幅に上回ることが確認されています。 -- [hidream_e1_1_bf16.safetensors（推奨）](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors（推奨）](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **テキストエンコーダー**： @@ -118,9 +119,14 @@ E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メ ## HiDream E1 の ComfyUI ネイティブ ワークフローの例 - -

Comfy Cloud で実行

- + + + Comfy Cloud で開く + + + Download JSON or search "HiDream E1 Full" in Template Library + + E1 は 2025年4月28日にリリースされたモデルで、**768×768 の固定解像度のみ** をサポートします。 diff --git a/ja/tutorials/image/hidream/hidream-i1.mdx b/ja/tutorials/image/hidream/hidream-i1.mdx index 2e7c76364..690d60e13 100644 --- a/ja/tutorials/image/hidream/hidream-i1.mdx +++ b/ja/tutorials/image/hidream/hidream-i1.mdx @@ -2,17 +2,18 @@ title: "ComfyUI ネイティブ版 HiDream-I1 テキストから画像へ変換するワークフローの例" sidebarTitle: "HiDream-I1" description: "本ガイドでは、ComfyUI ネイティブ版 HiDream-I1 のテキストから画像へ変換するワークフローの実行手順を詳しく説明します" -translationSourceHash: 590b8d6d +translationSourceHash: 73e39f2a translationFrom: tutorials/image/hidream/hidream-i1.mdx translationBlockHashes: "_intro": 59935bb2 "Model Features": 3ec455b0 "About This Workflow Example": 90c5d2b3 - "HiDream-I1 Workflow": 93e17294 + "HiDream-I1 Workflow": 5956f05c "Other Related Resources": e794d9ef --- + ![HiDream-I1 デモ](https://raw.githubusercontent.com/HiDream-ai/HiDream-I1/main/assets/demo.jpg) HiDream-I1 は、HiDream-ai 社が 2025 年 4 月 7 日に公式にオープンソース化したテキストから画像へ変換するモデルです。このモデルは 17B（170 億）パラメータを有し、[MIT ライセンス](https://github.com/HiDream-ai/HiDream-I1/blob/main/LICENSE) の下で公開されており、個人プロジェクト、学術研究、商用利用のすべてに対応しています。現在、複数のベンチマークテストにおいて優れた性能を発揮しています。 @@ -101,16 +102,21 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Full バージョンのワークフロー - -

Comfy Cloud で実行

- + + + セットアップ不要で Comfy Cloud で実行 + + + ワークフロー JSON ファイルをダウンロード + + #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード @@ -139,16 +145,21 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Dev バージョンのワークフロー - -

Comfy Cloud で実行

- + + + セットアップ不要で Comfy Cloud で実行 + + + ワークフロー JSON ファイルをダウンロード + + #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード @@ -177,16 +188,21 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ ### HiDream-I1 Fast バージョンのワークフロー - -

Comfy Cloud で実行

- + + + セットアップ不要で Comfy Cloud で実行 + + + ワークフロー JSON ファイルをダウンロード + + #### 1. モデルファイルのダウンロード ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 -- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） +- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） #### 2. ワークフローファイルのダウンロード diff --git a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 02b2594ef..024ee3d5b 100644 --- a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI NewBie-image-Exp0.1 ワークフロー例" description: "NewBie-image-Exp0.1 は、Next-DiT アーキテクチャを基盤とする 35 億パラメータのアニメスタイル文生成画像（text-to-image）モデルであり、XML 構造化プロンプトに対応し、高品質なアニメ画像生成に最適化されています。" sidebarTitle: "NewBie-image-Exp0.1" -translationSourceHash: f56facc0 +translationSourceHash: 22db4899 translationFrom: tutorials/image/newbie-image/newbie-image-exp-0-1.mdx translationBlockHashes: "_intro": 7f6f1284 - "NewBie-image text-to-image workflow": af50f161 + "NewBie-image text-to-image workflow": e7fbe7f5 "Model links": cd75bb7e "Prompt format": 5b819c6c --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **NewBie-image-Exp0.1** は、NewBieAI Lab が開発した 35 億パラメータの DiT（Diffusion Transformer）モデルで、アニメスタイルの文生成画像タスク専用に設計されています。Next-DiT アーキテクチャを採用しており、非常に詳細で視覚的に印象的なアニメ画像を生成できます。 @@ -30,13 +31,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## NewBie-image 文生成画像ワークフロー - -

JSON ワークフローファイルをダウンロード

- - - -

ComfyUI Cloud で実行

- + + + Download JSON or search "NewBie-image" in Template Library + + + Open in cloud + + + + + + Download JSON or search "NewBie-image" in Template Library + + + クラウドで開く + + @@ -44,16 +55,16 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/omnigen/omnigen2.mdx b/ja/tutorials/image/omnigen/omnigen2.mdx index f9c0a2d72..bc3f13257 100644 --- a/ja/tutorials/image/omnigen/omnigen2.mdx +++ b/ja/tutorials/image/omnigen/omnigen2.mdx @@ -2,17 +2,18 @@ title: "ComfyUI OmniGen2 ネイティブワークフローの例" description: "ComfyUI OmniGen2 ネイティブワークフローの例 — 文字から画像生成、画像編集、および複数画像の合成を統合したモデル。" sidebarTitle: "OmniGen2" -translationSourceHash: b061ff8c +translationSourceHash: 1429ffa8 translationFrom: tutorials/image/omnigen/omnigen2.mdx translationBlockHashes: "_intro": 3e9c5a21 "About OmniGen2": 2e6118e0 "OmniGen2 Model Download": 7997f8fe - "ComfyUI OmniGen2 Text-to-Image Workflow": f7650507 - "ComfyUI OmniGen2 Image Editing Workflow": 1859e94c + "ComfyUI OmniGen2 Text-to-Image Workflow": 5f9f63be + "ComfyUI OmniGen2 Image Editing Workflow": 1e06072e --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## OmniGen2 について @@ -42,13 +43,13 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 本記事では複数のワークフローを取り扱うため、対応するモデルファイルおよびインストール先は以下の通りです。各ワークフロー内にも、該当するモデルファイルのダウンロード情報が記載されています。 **拡散モデル（Diffusion Models）** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **テキストエンコーダー（Text Encoders）** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) ファイル保存先： ``` @@ -66,9 +67,11 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 ### 1. ワークフローファイルのダウンロード - -

Comfy Cloud で実行

- + + + Open and run this workflow directly in Comfy Cloud. + + ![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -94,9 +97,11 @@ OmniGen2 は豊富な画像編集機能を備えており、画像へのテキ ### 1. ワークフローファイルのダウンロード - -

Comfy Cloud で実行

- + + + Open and run this workflow directly in Comfy Cloud. + + ![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/ja/tutorials/image/ovis/ovis-image.mdx b/ja/tutorials/image/ovis/ovis-image.mdx index ecbad6885..2dd410dd0 100644 --- a/ja/tutorials/image/ovis/ovis-image.mdx +++ b/ja/tutorials/image/ovis/ovis-image.mdx @@ -2,7 +2,7 @@ title: "Ovis-Image ComfyUI ワークフロー例" description: "Ovis-Image は、高品質なテキストレンダリングに特化して最適化された7B規模のテキストから画像を生成するモデルであり、厳しい計算リソース制約下でも効率的に動作するよう設計されています。" sidebarTitle: "Ovis-Image" -translationSourceHash: 64c10a35 +translationSourceHash: 8d005c1d translationFrom: tutorials/image/ovis/ovis-image.mdx --- @@ -22,13 +22,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Ovis-Image のテキストから画像を生成するワークフロー - -

JSONワークフローファイルをダウンロード

- - - -

ComfyUI Cloud 上で実行

- + + + Comfy Cloud で開く + + + Download JSON or search "Ovis image" in Template Library + + + + + + Comfy Cloud で開く + + + Download JSON or search "Ovis image" in Template Library + + @@ -36,15 +46,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders（テキストエンコーダ）** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models（拡散モデル）** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/pixeldit/pixeldit.mdx b/ja/tutorials/image/pixeldit/pixeldit.mdx index 4cc67f103..287779c05 100644 --- a/ja/tutorials/image/pixeldit/pixeldit.mdx +++ b/ja/tutorials/image/pixeldit/pixeldit.mdx @@ -2,15 +2,16 @@ title: "PixelDiT ComfyUI ワークフロー例" description: "PixelDiT は NVIDIA のピクセル空間拡散トランスフォーマーで、1024px のテキストから画像への生成を行います。VAE エンコード/デコードを必要とせず、ピクセル空間で直接生成します。" sidebarTitle: "PixelDiT" -translationSourceHash: d5838b66 +translationSourceHash: b3280f94 translationFrom: tutorials/image/pixeldit/pixeldit.mdx translationBlockHashes: "_intro": ef3a0250 - "PixelDiT text-to-image workflow": 21b069a6 + "PixelDiT text-to-image workflow": 1879b3e8 "Model downloads": e4bafb0a --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **PixelDiT** は NVIDIA が開発したピクセル空間拡散トランスフォーマーで、**1024px** のテキストから画像への生成を行います。従来の潜在空間で動作する拡散モデルとは異なり、PixelDiT はデュアルレベル DiT アーキテクチャを使用してピクセル空間で直接画像を生成します——パッチレベル DiT とピクセルレベル DiT を組み合わせ、MM-DiT フュージョンによるテキストと画像トークン間の joint attention を実現します。 @@ -62,11 +63,11 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' PixelDiT はテキストエンコーダーと拡散モデルの 2 つのモデルファイルを使用します。 - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT テキストエンコーダー - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 拡散モデル diff --git a/ja/tutorials/image/qwen/qwen-image-2512.mdx b/ja/tutorials/image/qwen/qwen-image-2512.mdx index a144f0632..306db8595 100644 --- a/ja/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ja/tutorials/image/qwen/qwen-image-2512.mdx @@ -2,15 +2,16 @@ title: "Qwen-Image-2512 ComfyUI ネイティブワークフローの例" description: "Qwen-Image-2512 は、Qwen-Image のテキストから画像を生成する基盤モデルの12月アップデート版であり、人物のリアリズム向上、自然なディテールの精細化、およびテキストレンダリングの改善を特徴としています。" sidebarTitle: "Qwen-Image-2512" -translationSourceHash: dd214a11 +translationSourceHash: bdd1c703 translationFrom: tutorials/image/qwen/qwen-image-2512.mdx translationBlockHashes: "_intro": db22951d "Supported Aspect Ratios": 018121b0 - "Qwen-Image-2512 ComfyUI Native Workflow Example": 1221ba89 + "Qwen-Image-2512 ComfyUI Native Workflow Example": 50fbe75c --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image-2512** は、Qwen-Image のテキストから画像を生成する基盤モデルの12月アップデート版です。8月にリリースされたベース版 Qwen-Image モデルと比較して、Qwen-Image-2512 は画像品質およびリアリズムにおいて大幅な向上を実現しています。 @@ -55,28 +56,26 @@ ComfyUI を更新した後、テンプレートからワークフローファイ - **Text to Image (Qwen-Image 2512)**: 標準的な50ステップ生成 - **Text to Image (Qwen-Image 2512 4steps)**: Lightning LoRA を用いた高速4ステップ生成 - -

JSON ワークフローをダウンロード

- + ### 2. モデルのダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（任意：4ステップ Lightning 加速用）** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **拡散モデル** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（ほとんどのユーザーに推奨） -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（VRAM が十分に確保でき、より高品質な出力を求めている場合） +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（ほとんどのユーザーに推奨） +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（VRAM が十分に確保でき、より高品質な出力を求めている場合） **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx index 946f5c718..5c5038008 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -2,7 +2,7 @@ title: "Qwen-Image-Edit-2511 ComfyUI ネイティブワークフローの例" description: "Qwen-Image-Edit-2511 は、キャラクターの一貫性の向上、多人物編集、LoRA 機能の統合、および幾何学的推論能力の強化を特徴とする Qwen-Image-Edit の強化版です。" sidebarTitle: "Qwen-Image-Edit-2511" -translationSourceHash: be665671 +translationSourceHash: f1e0ff2e translationFrom: tutorials/image/qwen/qwen-image-edit-2511.mdx --- @@ -33,31 +33,41 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ComfyUI を更新した後、テンプレートからワークフローファイルを取得できます。あるいは、以下のワークフローを ComfyUI にドラッグ＆ドロップして読み込むこともできます。 - -

JSON 形式ワークフローをダウンロード

- - - -

ComfyUI Cloud 上で実行

- + + + Comfy Cloud で開く + + + Download JSON or search "Qwen-Image-Edit-2511" in Template Library + + + + + + Comfy Cloud で開く + + + Download JSON or search "Qwen-Image-Edit-2511" in Template Library + + ### 2. モデルのダウンロード **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（任意：4 ステップ Lightning 加速用）** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **拡散モデル** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image-edit.mdx b/ja/tutorials/image/qwen/qwen-image-edit.mdx index 01cbdb893..196316a34 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit.mdx @@ -2,15 +2,16 @@ title: "Qwen-Image-Edit ComfyUI ネイティブワークフローの例" description: "Qwen-Image-Edit は Qwen-Image の画像編集向けバージョンであり、20Bモデルを基にさらに学習が進められており、高精度なテキスト編集および意味／外観の両方を制御する編集機能をサポートします。" sidebarTitle: "Qwen-Image-Edit" -translationSourceHash: 9c641cf0 +translationSourceHash: 7c27546f translationFrom: tutorials/image/qwen/qwen-image-edit.mdx translationBlockHashes: "_intro": eee08e9a "ComfyOrg Qwen-Image-Edit Live Stream": 6e70a462 - "Qwen-Image-Edit ComfyUI Native Workflow Example": 7df4c813 + "Qwen-Image-Edit ComfyUI Native Workflow Example": 6703460f --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image-Edit** は、Qwen-Image の画像編集専用バージョンです。20B規模の Qwen-Image モデルを基に追加学習が行われており、Qwen-Image の特徴的なテキストレンダリング能力を編集タスクへと成功裏に拡張し、高精度なテキスト編集を実現しています。さらに、Qwen-Image-Edit では入力画像を Qwen2.5-VL（視覚的意味制御用）および VAE エンコーダ（視覚的外観制御用）の両方に同時に供給することで、意味と外観の両方を独立して制御可能な「二重編集機能」を実現しています。 @@ -47,13 +48,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ComfyUI を更新後、テンプレートからワークフローファイルを取得するか、下記のワークフローを ComfyUI へドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

JSON形式ワークフローをダウンロード

- - - -

ComfyUI Cloud 上で実行

- + + + Download JSON or search "image_qwen_image_edit" in Template Library + + + Run this workflow on Cloud GPUs with zero setup + + + + + + Download JSON or search "image_qwen_image_edit" in Template Library + + + Run this workflow on Cloud GPUs with zero setup + + 以下の画像を入力画像としてダウンロードしてください。 ![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -64,19 +75,19 @@ ComfyUI を更新後、テンプレートからワークフローファイルを **Diffusion モデル** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **テキストエンコーダ** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) モデルの保存場所 diff --git a/ja/tutorials/image/qwen/qwen-image-layered.mdx b/ja/tutorials/image/qwen/qwen-image-layered.mdx index b2c05499a..119889f0d 100644 --- a/ja/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ja/tutorials/image/qwen/qwen-image-layered.mdx @@ -2,17 +2,18 @@ title: "Qwen-Image-Layered ComfyUI ワークフロー例" description: "Qwen-Image-Layered は、画像を複数の RGBA レイヤーに分解できるモデルであり、レイヤー分解を通じて本質的な編集可能性を実現します。" sidebarTitle: "Qwen-Image-Layered" -translationSourceHash: bf2f2000 +translationSourceHash: 331a3de0 translationFrom: tutorials/image/qwen/qwen-image-layered.mdx translationBlockHashes: "_intro": 19900234 - "Qwen-Image-Layered workflow": ba937275 - "Model links": 98d12555 + "Qwen-Image-Layered workflow": 7a12a7ef + "Model links": 0635af99 "FP8 version": 6bffdd17 - "Workflow settings": b0f81aa2 + "Workflow settings": 098636f1 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image-Layered** は、アリババ社の通義千問（Qwen）チームが開発したモデルで、入力画像を複数の RGBA レイヤーに分解することができます。このレイヤー化された表現により、本質的な編集可能性が実現されます：各レイヤーを他のコンテンツに影響を与えることなく独立して操作できます。 @@ -30,13 +31,14 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Qwen-Image-Layered ワークフロー - -

JSON ワークフローファイルをダウンロード

- - - -

ComfyUI Cloud で実行

- + + + JSON ワークフローファイルをダウンロード + + + セットアップ不要で ComfyUI オンラインで実行 + + @@ -44,15 +46,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **モデルの保存場所** diff --git a/ja/tutorials/image/qwen/qwen-image.mdx b/ja/tutorials/image/qwen/qwen-image.mdx index 7fb6475cc..70c8068c3 100644 --- a/ja/tutorials/image/qwen/qwen-image.mdx +++ b/ja/tutorials/image/qwen/qwen-image.mdx @@ -2,18 +2,19 @@ title: "Qwen-Image ComfyUIネイティブワークフローの例" description: "Qwen-Imageは、Apache 2.0ライセンスのもとでオープンソース化された、20BパラメータのMMDiT（マルチモーダル拡散トランスフォーマー）モデルです。" sidebarTitle: "Qwen-Image" -translationSourceHash: f28191a9 +translationSourceHash: 4c682717 translationFrom: tutorials/image/qwen/qwen-image.mdx translationBlockHashes: "_intro": e096d7eb "ComfyOrg Qwen-Image live stream": 6436ac25 - "Qwen-Image Native Workflow Example": 73bab810 - "Qwen Image InstantX ControlNet Workflow": 741d3247 - "Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow": 45d40ad6 - "Qwen Image Union ControlNet LoRA Workflow": 333b4f67 + "Qwen-Image Native Workflow Example": bedb58a8 + "Qwen Image InstantX ControlNet Workflow": c5d39cf3 + "Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow": 6dc30273 + "Qwen Image Union ControlNet LoRA Workflow": a08d8e37 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image** は、アリババのQwenチームがリリースした初の画像生成基盤モデルです。これは、Apache 2.0ライセンスのもとでオープンソース化された20BパラメータのMMDiT（マルチモーダル拡散トランスフォーマー）モデルです。このモデルは、**複雑なテキストレンダリング**および**精密な画像編集**において顕著な進展を遂げており、英語や中国語など複数の言語において高忠実度の出力を実現しています。 @@ -82,14 +83,10 @@ GPU：RTX4090D（24GB） ComfyUIを更新後、テンプレートからワークフローファイルを検索するか、以下のワークフローをComfyUIにドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像へ変換するワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

Qwen-Image公式モデル用ワークフロー（JSON形式）をダウンロード

- + 蒸留版 - -

蒸留モデル用ワークフロー（JSON形式）をダウンロード

- + ### 2. モデルのダウンロード @@ -103,12 +100,12 @@ ComfyUIを更新後、テンプレートからワークフローファイルを **拡散モデル** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill（蒸留版） -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 蒸留版のオリジナル作者は、CFG値1.0で15ステップでの使用を推奨しています。 @@ -117,15 +114,15 @@ Qwen_image_distill（蒸留版） **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **テキストエンコーダー** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **モデルの保存場所** @@ -173,9 +170,14 @@ Qwen_image_distill（蒸留版） 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

JSON形式ワークフローをダウンロード

- + + + + + + + + 以下の画像を入力としてダウンロードしてください ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -183,7 +185,7 @@ Qwen_image_distill（蒸留版） 1. InstantX ControlNet -[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)をダウンロードし、`ComfyUI/models/controlnet/`フォルダーに保存してください +[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)をダウンロードし、`ComfyUI/models/controlnet/`フォルダーに保存してください 2. **Lotus Depthモデル** @@ -191,11 +193,11 @@ Qwen_image_distill（蒸留版） **拡散モデル** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAEモデル** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) または任意のSD1.5互換VAE +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) または任意のSD1.5互換VAE ``` ComfyUI/ @@ -234,9 +236,7 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして対応するワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

JSON形式ワークフローをダウンロード

- + 以下の画像を入力としてダウンロードしてください： @@ -246,9 +246,9 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http その他のモデルはQwen-Image基本ワークフローと同一です。以下のモデルのみをダウンロードし、`ComfyUI/models/model_patches`フォルダーに保存してください。 -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. ワークフローの使用方法 @@ -301,9 +301,7 @@ Comfy Org再ホストURL：[qwen_image_union_diffsynth_lora.safetensors](https:/ 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

JSON形式ワークフローをダウンロード

- + 以下の画像を入力としてダウンロードしてください diff --git a/ja/tutorials/image/z-image/z-image-turbo.mdx b/ja/tutorials/image/z-image/z-image-turbo.mdx index fe7d24df0..b75fe7cf1 100644 --- a/ja/tutorials/image/z-image/z-image-turbo.mdx +++ b/ja/tutorials/image/z-image/z-image-turbo.mdx @@ -2,15 +2,16 @@ title: "Z-Image-Turbo ComfyUI ワークフロー例" description: "Z-Image-Turbo は、蒸留済みの 6B パラメーターを備えた高効率画像生成モデルであり、サブセカンド（1秒未満）の推論遅延を実現します。" sidebarTitle: "Z-Image-Turbo" -translationSourceHash: 6ef05349 +translationSourceHash: b9ae7e8c translationFrom: tutorials/image/z-image/z-image-turbo.mdx translationBlockHashes: "_intro": 062e5fc1 "Z-Image-Turbo text-to-image workflow": 5057b237 - "Z-Image-Turbo Fun Union ControlNet workflow": 18f13b79 + "Z-Image-Turbo Fun Union ControlNet workflow": 34191a15 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -46,15 +47,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### Z-Image-Turbo モデルのダウンロード - + Z-Image-Turbo 専用のテキストエンコーダーです。 - + Z-Image-Turbo 専用の拡散モデルです。 - + Z-Image-Turbo 専用のVAE（変分オートエンコーダー）です。 @@ -81,7 +82,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### ControlNet 用の追加モデル - + Z-Image-Turbo 専用のControlNetモデルパッチです。 diff --git a/ja/tutorials/partner-nodes/anthropic/claude.mdx b/ja/tutorials/partner-nodes/anthropic/claude.mdx index d2d2a9daf..274b24809 100644 --- a/ja/tutorials/partner-nodes/anthropic/claude.mdx +++ b/ja/tutorials/partner-nodes/anthropic/claude.mdx @@ -2,7 +2,7 @@ title: "Anthropic Claude API ノード ComfyUI 公式サンプル" description: "本記事では、ComfyUI で Anthropic Claude パートナーノードを用いて対話機能を実現する方法について説明します" sidebarTitle: "Anthropic Claude" -translationSourceHash: 155df30f +translationSourceHash: b0368d1f translationFrom: tutorials/partner-nodes/anthropic/claude.mdx --- @@ -30,13 +30,16 @@ Anthropic Claude は、Anthropic が開発した強力な AI モデルファミ ## Anthropic Claude チャットワークフロー - +Anthropic Claude Chat workflow preview + + + Comfy Cloud を開く - - + JSON をダウンロードするか、テンプレートライブラリで "Anthropic Claude" を検索 + 対応するテンプレートでは、画像を解析して描画プロンプトへと変換するためのロールプロンプトを構築しています。 diff --git a/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 2b8615da5..be3f1ac6f 100644 --- a/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -2,15 +2,16 @@ title: "Beeble SwitchX Partner Nodes ComfyUI 公式例" description: "ComfyUI で Beeble SwitchX Partner Nodes を使用して AI による画像・動画の再照明と環境編集を行う方法を解説します" sidebarTitle: "Beeble SwitchX" -translationSourceHash: 3724b3ae +translationSourceHash: 8f168afe translationFrom: tutorials/partner-nodes/beeble/beeble-switchx.mdx translationBlockHashes: "_intro": 470a081a - "Beeble SwitchX: Image Edit": a252a19f - "Beeble SwitchX: Video Edit": 97df1f25 + "Beeble SwitchX: Image Edit": 245e4f8a + "Beeble SwitchX: Video Edit": 166c2e59 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -27,13 +28,14 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; Beeble SwitchX 画像編集ワークフロー - + + Comfy Cloud を開く - - + JSON をダウンロード、またはテンプレートライブラリで "Beeble SwitchX: Image Edit" を検索 + ### 仕組み @@ -67,13 +69,14 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; Beeble SwitchX 動画編集ワークフロー - + + Comfy Cloud を開く - - + JSON をダウンロード、またはテンプレートライブラリで "Beeble SwitchX: Video Edit" を検索 + ### 仕組み diff --git a/ja/tutorials/partner-nodes/bria/background-removal.mdx b/ja/tutorials/partner-nodes/bria/background-removal.mdx index 7dde64034..784ec4095 100644 --- a/ja/tutorials/partner-nodes/bria/background-removal.mdx +++ b/ja/tutorials/partner-nodes/bria/background-removal.mdx @@ -2,15 +2,16 @@ title: "ComfyUIでBriaの背景除去を使用する" description: "ComfyUIでBriaパートナーノードを使用して、画像・動画の背景除去、グリーンスクリーン、背景置換を行う方法を学びます" sidebarTitle: "Bria 背景除去" -translationSourceHash: c65b0474 +translationSourceHash: dbbb15e9 translationFrom: tutorials/partner-nodes/bria/background-removal.mdx translationBlockHashes: "_intro": f9000433 - "Image Background Removal": 14b7eb56 - "Video Background Processing": 3e621db5 + "Image Background Removal": 988ae876 + "Video Background Processing": aab02df5 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index 5d74706b8..3b1be6be6 100644 --- a/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -2,16 +2,17 @@ title: "ByteDance Seed Audio 1.0 - ユニバーサルオーディオ生成" description: "ComfyUIでSeed Audio 1.0を使用し、音声クローン、プリセット音声、キャラクタ駆動オーディオを用いて、単一のプロンプトから音声、音楽、効果音、複数話者による対話を生成します" sidebarTitle: "Seed Audio 1.0" -translationSourceHash: 3fe23cff +translationSourceHash: 86bb1688 translationFrom: tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx translationBlockHashes: - "_intro": 4ce27313 - "Key capabilities": d5c85b17 + "_intro": 7d1fd38c + "Key capabilities": f5c9eb87 "Available workflows": 37e20124 - "How to use Seed Audio 1.0 in ComfyUI": 37aa925d + "How to use Seed Audio 1.0 in ComfyUI": 5630659b "Get started": 1ae5cf45 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index c24203919..2e536fb9b 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -2,17 +2,18 @@ title: "Seedance 2.0 実在人物 - 実在人物の一貫性動画生成" description: "ComfyUIで1回のByteDanceライブネス（本人確認）を行い、Seedance 2.0で実在人物の一貫性と音声同期を保った動画を生成" sidebarTitle: "Seedance 2.0 実在人物" -translationSourceHash: 82069c10 +translationSourceHash: 9ea8f507 translationFrom: tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx translationBlockHashes: "_intro": 4890d35b "Video guide": 416e3ec3 "What's different": b67103d8 "How verification works in ComfyUI": ee1e1056 - "Available workflows": bc3d7526 + "Available workflows": 16803a18 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index 87278e101..65bbbd436 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -2,17 +2,18 @@ title: "Seedance 2.0 - AI動画生成" description: "ComfyUIでSeedance 2.0を使用して、テキスト、画像、動画、音声から同期音声、一貫したキャラクター、映画的カメラコントロールを備えた高品質動画を生成" sidebarTitle: "Seedance 2.0" -translationSourceHash: 4c9aec61 +translationSourceHash: 35902650 translationFrom: tutorials/partner-nodes/bytedance/seedance-2-0.mdx translationBlockHashes: "_intro": e1a3013d "Key capabilities": 6ca35ecf - "Available workflows": 4a9917b4 - "Seedance 2.0 Mini": c2cf0c21 + "Available workflows": d232c2a0 + "Seedance 2.0 Mini": db24d1c9 "Using real-person and AI-generated portraits in ComfyUI for Seedance 2.0": d9966fae --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index 75c615b0d..46bd485b3 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -2,17 +2,18 @@ title: "ByteDance Seedream 5.0 Lite — スマートAI画像生成" description: "ComfyUIでSeedream 5.0 Liteを用いて、ウェブ接続型検索および強化された指示追随機能による画像生成を行います" sidebarTitle: "Seedream 5.0 Lite" -translationSourceHash: ec3a4ae6 +translationSourceHash: 39b0d304 translationFrom: tutorials/partner-nodes/bytedance/seedream-5-lite.mdx translationBlockHashes: "_intro": 1991e54a "What's new in Seedream 5.0 lite": 02ac6a40 - "Seedream 5.0 lite image edit workflow": 44d7f3e4 - "Seedream 5.0 lite text to image workflow": 4fc4c666 + "Seedream 5.0 lite image edit workflow": e90cc958 + "Seedream 5.0 lite text to image workflow": f36785bf "Get started": d14874f4 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/ja/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index 3fdac3fcf..df522dc7b 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -2,15 +2,16 @@ title: "ByteDance Seedream 5.0 Pro - プロフェッショナルAI画像生成" description: "高品質な画像を生成し、強化された指示追従、精密な編集、プロフェッショナルグレードの出力をComfyUIで実現" sidebarTitle: "Seedream 5.0 Pro" -translationSourceHash: 125a57d0 +translationSourceHash: c10432af translationFrom: tutorials/partner-nodes/bytedance/seedream-5-pro.mdx translationBlockHashes: "_intro": 0b4a40d7 - "What sets Seedream 5.0 Pro apart": ae41f702 - "Seedream 5.0 Pro text to image workflow": 39d05a61 - "Seedream 5.0 Pro image edit workflow": 7c657db8 + "What sets Seedream 5.0 Pro apart": e1702d4b + "Seedream 5.0 Pro text-to-image workflow": 8dbc58df + "Seedream 5.0 Pro image edit workflow": e2cabdb9 "Get started": 709ac7fd --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx index c72de4c2b..1a3a4274c 100644 --- a/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -2,16 +2,17 @@ title: "Gemini Omni Flash: 会話型ビデオ生成" description: "Gemini Omni Flashは、Googleのマルチモーダルビデオモデルです。パートナーノードを通じてComfyUIで利用でき、自然言語でビデオを生成・編集できます。" sidebarTitle: "Gemini Omni Flash" -translationSourceHash: 2354813d +translationSourceHash: 6e260d1e translationFrom: tutorials/partner-nodes/google/gemini-omni-flash.mdx translationBlockHashes: "_intro": 3b6973dc "What Gemini Omni Flash offers": 215de783 - "Workflows": d363c40a + "Workflows": 149ee81a "Get started": 64517938 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/google/gemini.mdx b/ja/tutorials/partner-nodes/google/gemini.mdx index f1a6c2ec1..27fe14312 100644 --- a/ja/tutorials/partner-nodes/google/gemini.mdx +++ b/ja/tutorials/partner-nodes/google/gemini.mdx @@ -2,7 +2,7 @@ title: "Google Gemini API ノード ComfyUI 公式サンプル" description: "本記事では、ComfyUI で Google Gemini パートナーノードを用いて対話機能を実現する方法について説明します" sidebarTitle: "Google Gemini" -translationSourceHash: ddd327ee +translationSourceHash: 8cd6027c translationFrom: tutorials/partner-nodes/google/gemini.mdx --- @@ -22,13 +22,14 @@ Google Gemini は、Google が開発した強力な AI モデルであり、対 以下の JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式のワークフローファイルをダウンロード

- + + + Download Json Format Workflow File + + + Comfy Cloud で開く + + ### 2. ワークフロー実行の手順 diff --git a/ja/tutorials/partner-nodes/google/nano-banana-2.mdx b/ja/tutorials/partner-nodes/google/nano-banana-2.mdx index 772d077ed..82201f913 100644 --- a/ja/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/ja/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -2,17 +2,18 @@ title: "Nano Banana 2 — 高速AI画像生成" description: "ComfyUIでNano Banana 2を活用し、Flash速度でプロレベルの品質の画像を生成" sidebarTitle: "Nano Banana 2" -translationSourceHash: a146b8b0 +translationSourceHash: a86a55b2 translationFrom: tutorials/partner-nodes/google/nano-banana-2.mdx translationBlockHashes: "_intro": 379a03e0 "What's new in Nano Banana 2": 4048475a - "Nano Banana 2 image edit workflow": 085cc8cb + "Nano Banana 2 image edit workflow": ccade8b6 "Which model should you pick?": f16d5cc7 "Get started": f6189d9e --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/ja/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index 539f4b41d..3028e0226 100644 --- a/ja/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/ja/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -2,6 +2,8 @@ title: "Grok Imagine Video 1.5 画像から動画 ComfyUI 公式例" description: "ComfyUI で Grok Imagine Video 1.5 Partner Node を使用して、画像からネイティブオーディオ付きの高品質動画を生成する方法を解説します" sidebarTitle: "Grok Imagine Video 1.5" +translationSourceHash: 6a40c4e4 +translationFrom: tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx --- import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; diff --git a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index ac8427e0f..0355c647d 100644 --- a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -2,18 +2,19 @@ title: "ComfyUI での HappyHorse 1.0 動画生成" description: "ComfyUI でパートナーノードを通じて HappyHorse 1.0 を使用し、画像から動画、テキストから動画、参照から動画、動画編集を、シネマティックな美学とマルチショット一貫性で実現する方法を学びましょう" sidebarTitle: "HappyHorse 1.0" -translationSourceHash: a0fbfb64 +translationSourceHash: 0b14a684 translationFrom: tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx translationBlockHashes: "_intro": 4eba8bbb "Highlights": 376295d6 - "HappyHorse 1.0 image-to-video": 85d1766b - "HappyHorse 1.0 text-to-video": 29f5941d - "HappyHorse 1.0 reference-to-video": fba3d9ea - "HappyHorse 1.0 video edit": 20a4559a + "HappyHorse 1.0 image-to-video": d41d0b77 + "HappyHorse 1.0 text-to-video": 6c594c34 + "HappyHorse 1.0 reference-to-video": b8ae994c + "HappyHorse 1.0 video edit": 5b531078 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index bf2991c8a..6a9cfef2d 100644 --- a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -2,18 +2,19 @@ title: "ComfyUIでのHappyHorse 1.1ビデオ生成" description: "ComfyUIでパートナーノードを介してHappyHorse 1.1を使用し、画像から動画、テキストから動画、参照から動画でネイティブ同期オーディオと映画的なマルチショットストーリーテリングを実現する方法を学びます" sidebarTitle: "HappyHorse 1.1" -translationSourceHash: 1c2775c3 +translationSourceHash: 2fadefbd translationFrom: tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx translationBlockHashes: "_intro": 6ae1c8e8 "Highlights": 10f03d9f - "HappyHorse 1.1 text-to-video": 31c81d0f - "HappyHorse 1.1 image-to-video": 2fa5bff9 - "HappyHorse 1.1 reference-to-video": 983e4818 + "HappyHorse 1.1 text-to-video": 3551ba66 + "HappyHorse 1.1 image-to-video": 05ad4488 + "HappyHorse 1.1 reference-to-video": 39830de6 "Getting started": 27bbb438 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 1f3a8ba1f..50c0166c6 100644 --- a/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -2,19 +2,20 @@ title: "Hunyuan 3D API ノードによるモデル生成：ComfyUI 公式サンプル" description: "本記事では、ComfyUI 内で Hunyuan 3D ノードの API を使用して 3D モデルを生成する方法について説明します。" sidebarTitle: "Hunyuan 3D 3.0" -translationSourceHash: 8d72daac +translationSourceHash: c92f6932 translationFrom: tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx translationBlockHashes: "_intro": 312a61a4 "Use cases": 60e6d268 "Getting started": 62f4530d - "Text-to-3D workflow": 710fb1e6 - "Image-to-3D workflow": d7a3b87d - "Multi-view-to-3D workflow": 68952799 - "Advanced features": acc7cc22 + "Text-to-3D workflow": 5edf82d4 + "Image-to-3D workflow": ce7efef1 + "Multi-view-to-3D workflow": ff251560 + "Advanced features": 6b37a964 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index d83c4d398..a046dd488 100644 --- a/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -2,18 +2,19 @@ title: "Hunyuan 3D API ノードによるモデル生成：ComfyUI 公式サンプル" description: "本記事では、ComfyUI で Hunyuan 3D ノードの API を使用して 3D モデルを生成する方法について説明します。" sidebarTitle: "モデル生成" -translationSourceHash: cb7d1627 +translationSourceHash: aa9c5029 translationFrom: tutorials/partner-nodes/hunyuan3d/model-generation.mdx translationBlockHashes: "_intro": 312a61a4 "Use cases": 60e6d268 "Getting started": 62f4530d - "Text-to-3D workflow": 710fb1e6 - "Image-to-3D workflow": d7a3b87d - "Multi-view-to-3D workflow": 68952799 + "Text-to-3D workflow": 5edf82d4 + "Image-to-3D workflow": ce7efef1 + "Multi-view-to-3D workflow": ff251560 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index d0be8c19b..b7295c896 100644 --- a/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Ideogram 4.0 Partner Node チュートリアル" description: "ComfyUI で Ideogram 4.0 API Partner Node を使用する方法" sidebarTitle: "Ideogram 4.0" -translationSourceHash: 20fee685 +translationSourceHash: c4f854a7 translationFrom: tutorials/partner-nodes/ideogram/ideogram-v4.mdx translationBlockHashes: "_intro": 9a8e46c4 - "Ideogram 4.0 Partner Node Text-to-Image Workflow": ca6fe17b + "Ideogram 4.0 Partner Node Text-to-Image Workflow": ca41ad51 "Additional Notes": 5cc3cb89 "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx index ef05651b6..7e0f2fdc7 100644 --- a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -2,19 +2,20 @@ title: "Kling 2.6 Motion Control API ノード ComfyUI 公式サンプル" description: "ComfyUI で Kling 2.6 Motion Control パートナーノードを活用し、参照動画からキャラクター画像への高精度なモーション転送を行う方法を学びます" sidebarTitle: "Kling 2.6 Motion Control" -translationSourceHash: 89919ffc +translationSourceHash: e449a655 translationFrom: tutorials/partner-nodes/kling/kling-motion-control.mdx translationBlockHashes: "_intro": fac1cb1e "Product highlights": a82b89a7 "Character orientation modes": aa2f7dec "Model tiers": 565d97ae - "Kling 2.6 Motion Control workflow": c75b1c17 + "Kling 2.6 Motion Control workflow": 33f2f69c "Input requirements": 0ecbf467 "Tips for better results": b6fdceb5 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -51,9 +52,20 @@ Kling 2.6 Motion Control は、快手（Kuaishou）社が開発した専用の ## Kling 2.6 Motion Control のワークフロー - -

JSON形式のワークフローファイルをダウンロード

- + + + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. + + + Download the workflow JSON file for local use. + + + Download sample reference image + + + Download sample motion reference video + + ## 入力要件 diff --git a/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx b/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx index ad999bbc2..14c1048a5 100644 --- a/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -2,14 +2,14 @@ title: "Luma Uni-1 ガイド" description: "ComfyUI で Luma Uni-1 パートナーノードを使った画像生成・編集。" sidebarTitle: "Luma Uni-1" -translationSourceHash: 16838ef3 +translationSourceHash: 99d39585 translationFrom: tutorials/partner-nodes/luma/luma-uni-1.mdx translationBlockHashes: "_intro": 544c5d43 "What makes Uni-1 different": c0181542 "Strengths": 4dd065dd "The core distinction: Create vs Modify": c3ea208f - "Available workflows": a6cf5659 + "Available workflows": 9b93a043 "Core parameters": b43ac975 "Working with reference images": d20d34ae "Prompting guidelines": ba9ac50d @@ -23,6 +23,7 @@ translationBlockHashes: --- + **ComfyUI** では、Luma **Uni-1** は **パートナー API ノード**として利用します。**Create** はプロンプト（と任意の参照画像）から新規画像を生成し、**Modify** は入力画像を編集します。**Load Image** / **Save Image** などとノードを接続し、プロンプト・シード・アスペクト比・参照スロットを Luma ノード上で設定してグラフをキューするか、**Comfy Cloud** のテンプレートから開いて試せます。 Luma は Uni-1 を拡散モデルではないデコーダのみの自己回帰モデルとして説明しており、生成前にプロンプトを推論します。キャンバス上では **Create / Modify の選択**、参照画像の役割の明示、シードによる反復が実務上の要点です。 diff --git a/ja/tutorials/partner-nodes/meshy/meshy-6.mdx b/ja/tutorials/partner-nodes/meshy/meshy-6.mdx index ca9bfda00..e375710e1 100644 --- a/ja/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/ja/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -2,10 +2,17 @@ title: "Meshy 6 API ノードによる 3D モデル生成 — ComfyUI 公式サンプル" description: "本記事では、ComfyUI で Meshy 6 ノードの API を使用して 3D モデルを生成する方法について説明します。" sidebarTitle: "Meshy 6" -translationSourceHash: 302cd348 +translationSourceHash: f3036351 translationFrom: tutorials/partner-nodes/meshy/meshy-6.mdx +translationBlockHashes: + "_intro": 2dd46cd7 + "About Meshy 6": bc605135 + "Text-to-Model Workflow": f800f061 + "Image-to-Model Workflow": c148120d + "Multi-view to Model Workflow": 0cf6bb73 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -59,4 +66,4 @@ Meshy 6 の「画像→3D」機能を活用し、2D 画像を詳細な 3D モデ ローカル環境で使用するためのワークフロー JSON ファイルをダウンロードできます。 - \ No newline at end of file + diff --git a/ja/tutorials/partner-nodes/openai/chat.mdx b/ja/tutorials/partner-nodes/openai/chat.mdx index 2d195caf5..a8c9e7c4a 100644 --- a/ja/tutorials/partner-nodes/openai/chat.mdx +++ b/ja/tutorials/partner-nodes/openai/chat.mdx @@ -2,7 +2,7 @@ title: "OpenAI Chat API ノード ComfyUI 公式サンプル" description: "本記事では、ComfyUI で OpenAI Chat パートナーノードを活用して対話機能を実現する方法について説明します" sidebarTitle: "OpenAI Chat" -translationSourceHash: e4c0b73e +translationSourceHash: 77289d08 translationFrom: tutorials/partner-nodes/openai/chat.mdx --- @@ -22,13 +22,14 @@ OpenAI は生成AIに特化した企業であり、強力な対話機能を提 以下の JSON ファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式のワークフローファイルをダウンロード

- + + + Comfy Cloud で開く + + + Download the JSON format workflow file. + + ### 2. ワークフロー実行手順（ステップごと） diff --git a/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx b/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx index c4621795e..2c271dd1d 100644 --- a/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -3,18 +3,19 @@ title: "OpenAI GPT-Image-2 ノード" description: "ComfyUI で OpenAI GPT-Image-2 パートナーノードを用いて画像を生成する方法を学びます" sidebarTitle: "GPT-Image-2" icon: "image" -translationSourceHash: b99cdc88 +translationSourceHash: ede8405e translationFrom: tutorials/partner-nodes/openai/gpt-image-2.mdx translationBlockHashes: "_intro": 9f80a1fc "Node Overview": 84ac2d88 "Getting Started": ab5a2231 - "Available workflows": 08afb96c + "Available workflows": d687898b "Key Capabilities": df71dcca "Hybrid Pipelines": 9c6479f0 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/pricing.mdx b/ja/tutorials/partner-nodes/pricing.mdx index 40557945a..54c60fd37 100644 --- a/ja/tutorials/partner-nodes/pricing.mdx +++ b/ja/tutorials/partner-nodes/pricing.mdx @@ -3,7 +3,7 @@ title: "価格" description: "本記事では、現在提供中のパートナーノードの価格を一覧表示します。すべての価格はクレジット単位（211クレジット = 1米ドル）で表記されています。" sidebarTitle: "価格" mode: wide -translationSourceHash: c864b96d +translationSourceHash: 907be779 translationFrom: tutorials/partner-nodes/pricing.mdx translationBlockHashes: "_intro": 845592d4 @@ -16,7 +16,9 @@ translationBlockHashes: "Magnific": 4000e7e1 "Google": 21dc3165 "HappyHorse": bceb584a - "Hitpaw": e83827b3 + "Hitpaw": 320c0537 + "Ideogram": 9410bf72 + "Krea": fb20c257 "Kling": cdf2c899 "Lightricks": 4870964f "Luma": 655207a5 @@ -52,6 +54,7 @@ translationBlockHashes: + 以下の表には、現在提供中のパートナーノードの価格が記載されています。すべての価格はクレジット単位です。 211クレジット = 1米ドル diff --git a/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx b/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx index 1b11b5017..c80a0013e 100644 --- a/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -2,17 +2,18 @@ title: "ComfyUI での Recraft V4 による画像およびベクター生成" description: "ComfyUI で Recraft V4 を使用して、プロフェッショナルな画像および実用可能なベクター画像を生成します" sidebarTitle: "Recraft V4" -translationSourceHash: fe0eaf6f +translationSourceHash: da066915 translationFrom: tutorials/partner-nodes/recraft/recraft-v4.mdx translationBlockHashes: "_intro": 8e346040 "What's new in V4": 6bdb9842 - "Recraft V4 text to image workflow": 7273fda3 - "Recraft V4 text to vector workflow": d322ba5f + "Recraft V4 text to image workflow": 2ff998b0 + "Recraft V4 text to vector workflow": c3255bfc "Additional notes": d5f3f109 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/rodin/model-generation.mdx b/ja/tutorials/partner-nodes/rodin/model-generation.mdx index 99704e747..c9930daa5 100644 --- a/ja/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ja/tutorials/partner-nodes/rodin/model-generation.mdx @@ -2,16 +2,17 @@ title: "Rodin API ノードによるモデル生成：ComfyUI 公式サンプル" description: "本記事では、ComfyUI 内で Rodin ノードの API を使用して 3D モデルを生成する方法について説明します。" sidebarTitle: "モデル生成" -translationSourceHash: 8d58740b +translationSourceHash: 738362ca translationFrom: tutorials/partner-nodes/rodin/model-generation.mdx translationBlockHashes: "_intro": a35f46e8 - "Single-view Model Generation Workflow": d982caac - "Multi-view Model Generation Workflow": 8a727575 + "Single-view Model Generation Workflow": d7303db4 + "Multi-view Model Generation Workflow": 5c5d3550 "Other Related Nodes": c885ce31 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -32,13 +33,17 @@ ComfyUI は現在、Rodin のモデル生成 API をネイティブ統合して 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式ワークフローファイルをダウンロード

- + + + Single-view Model Generation (Json Format) + + + Comfy Cloud で開く + + + Download sample input image + + 以下の画像を入力画像としてダウンロードしてください。 @@ -66,13 +71,23 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込むことができます。 - -

JSON 形式ワークフローファイルをダウンロード

- + + + Multi-view Model Generation (Json Format) + + + Comfy Cloud で開く + + + Download front view + + + Download back view + + + Download left view + + 以下の画像を入力画像としてダウンロードしてください。 diff --git a/ja/tutorials/partner-nodes/topaz/astra-2.mdx b/ja/tutorials/partner-nodes/topaz/astra-2.mdx index 152e56f3c..e048c1eaf 100644 --- a/ja/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/ja/tutorials/partner-nodes/topaz/astra-2.mdx @@ -2,7 +2,7 @@ title: "Astra 2 - クリエイティブ拡散モデルによる動画アップスケーリング" description: "初めての方向けに Astra 2 の役割を説明し、ComfyUI でのクリエイティブ動画アップスケールと制御についてまとめます。" sidebarTitle: "Astra 2" -translationSourceHash: 2840b2ae +translationSourceHash: d28f921a translationFrom: tutorials/partner-nodes/topaz/astra-2.mdx --- diff --git a/ja/tutorials/partner-nodes/tripo/model-generation.mdx b/ja/tutorials/partner-nodes/tripo/model-generation.mdx index c9a3bb055..b237411cb 100644 --- a/ja/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ja/tutorials/partner-nodes/tripo/model-generation.mdx @@ -2,17 +2,18 @@ title: "Tripo API ノードによるモデル生成：ComfyUI 公式サンプル" description: "本記事では、ComfyUI 内で Tripo ノードの API を使用して 3D モデルを生成する方法について説明します" sidebarTitle: "モデル生成" -translationSourceHash: 91cfb63d +translationSourceHash: 9fa47d10 translationFrom: tutorials/partner-nodes/tripo/model-generation.mdx translationBlockHashes: "_intro": b329659d - "Text-to-Model Workflow": 1f82cb18 - "Image-to-Model Workflow": 20fac619 - "Multi-view Model Generation Workflow": b0e0a897 + "Text-to-Model Workflow": ae8e8384 + "Image-to-Model Workflow": 7463fca4 + "Multi-view Model Generation Workflow": c0a49c52 "Subsequent Task Processing for the Same Task": d24582e4 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -35,9 +36,14 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - -

JSON形式ワークフローファイルをダウンロード

- + + + Try the Text-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + ### 2. ワークフロー実行手順（ステップ・バイ・ステップ） @@ -58,9 +64,17 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - -

JSON形式ワークフローファイルをダウンロード

- + + + Try the Image-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + + Download sample input image + + 以下の画像を入力画像としてダウンロードしてください： @@ -85,9 +99,20 @@ ComfyUI は現在、Tripo の API をネイティブ統合済みであり、関 以下のファイルをダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込めます。 - -

JSON形式ワークフローファイルをダウンロード

- + + + Try the Multiview-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + + Download front view input + + + Download back view input + + 以下の画像を入力画像としてダウンロードしてください： diff --git a/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx index e103f05ac..8614183c8 100644 --- a/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -2,16 +2,17 @@ title: "Tripo 3.1 — 高精細 3D アセット生成 ComfyUI 公式ガイド" description: "ComfyUI のパートナーノードを使用して Tripo 3.1 で高精細な 3D アセットを生成する方法を紹介します。高密度ジオメトリと PBR 対応マテリアルによりプロダクション品質を実現します。" sidebarTitle: "Tripo 3.1" -translationSourceHash: c520daa5 +translationSourceHash: dbbf32c8 translationFrom: tutorials/partner-nodes/tripo/tripo-3-1.mdx translationBlockHashes: "_intro": ff0ef375 "Capabilities": 6187e278 "Use Cases": 10ad3fe3 - "Available Workflows": c68e812a + "Available Workflows": 5ec1d9c5 --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/wan/wan2-7.mdx b/ja/tutorials/partner-nodes/wan/wan2-7.mdx index ee7c7c7a6..d07aaffd9 100644 --- a/ja/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/ja/tutorials/partner-nodes/wan/wan2-7.mdx @@ -2,19 +2,20 @@ title: "ComfyUI での Wan2.7 動画生成" description: "ComfyUI でパートナーノードを使用して Wan2.7 による画像から動画、テキストから動画、参照から動画、動画の続き、動画編集を行う方法を学びましょう" sidebarTitle: "Wan2.7" -translationSourceHash: 3bb39265 +translationSourceHash: 42f5b589 translationFrom: tutorials/partner-nodes/wan/wan2-7.mdx translationBlockHashes: "_intro": 611e27ab "Key features": 6ca0965f "Highlights": cec936d1 - "Wan2.7 image-to-video": 29a49b02 - "Wan2.7 text-to-video": 25cfcacf - "Wan2.7 reference-to-video": cf70840e - "Wan2.7 video edit": ad198739 + "Wan2.7 image-to-video": af05fd5c + "Wan2.7 text-to-video": c9f23ab0 + "Wan2.7 reference-to-video": fe373e60 + "Wan2.7 video edit": 0d2511bb --- + import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/utility/moge.mdx b/ja/tutorials/utility/moge.mdx index e3e7397fa..755d1c026 100644 --- a/ja/tutorials/utility/moge.mdx +++ b/ja/tutorials/utility/moge.mdx @@ -2,19 +2,20 @@ title: "ComfyUI MoGe 使用例" description: "このガイドでは、ComfyUI で MoGe を使用して単眼幾何推定を行う方法を説明します：メートルスケールのポイントマップ、深度マップ、法線マップ、メッシュ生成。" sidebarTitle: "MoGe" -translationSourceHash: 7ae1c247 +translationSourceHash: ff036dcc translationFrom: tutorials/utility/moge.mdx translationBlockHashes: "_intro": f3974fd4 "Model Installation": fb228472 "Example Workflows": cf2f84c1 - "1. Depth Estimation": 0c7bafb8 - "2. Perspective to Mesh": 03608c5b - "3. Panorama to Mesh": bdbb850d + "1. Depth Estimation": 030e1494 + "2. Perspective to Mesh": 4fde2470 + "3. Panorama to Mesh": b273460c "Community Resources": d9fcd8cd --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # ComfyUI MoGe の紹介 @@ -50,8 +51,8 @@ ComfyUI は MoGe ノードをネイティブサポートしています。始め MoGe チェックポイントをダウンロードし、ComfyUI の該当フォルダに保存します： -- **MoGe-2（推奨）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1（ベースライン）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2（推奨）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1（ベースライン）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/bytedance/bernini-r.mdx b/ja/tutorials/video/bytedance/bernini-r.mdx index 13d6b0383..d3877c1ab 100644 --- a/ja/tutorials/video/bytedance/bernini-r.mdx +++ b/ja/tutorials/video/bytedance/bernini-r.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Bernini-R 公式サンプル" description: "ComfyUI で Bernini-R を使用した画像・動画編集（再照明、スタイル転送、被写体挿入など）を学びましょう。" sidebarTitle: "Bernini-R" -translationSourceHash: 84ecde34 +translationSourceHash: e0b08f10 translationFrom: tutorials/video/bytedance/bernini-r.mdx translationBlockHashes: - "_intro": efe73774 - "Model Installation": 84eedf7b + "_intro": b791048b + "Model Installation": 286fbd69 "Example Workflows": cf2f84c1 - "1. Image Editing": cd75c4ca - "2. Video Editing": 9e90382c - "Community Resources": 57f652a5 + "1. Image Editing": 0c53cda8 + "2. Video Editing": ef878ac4 + "Community Resources": 9b4a0aaf --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # ComfyUI Bernini-R 概要 @@ -49,16 +50,16 @@ ComfyUI は Bernini-R ノードをネイティブサポートしています。 必要なモデルウェイトをダウンロードし、対応する ComfyUI フォルダに保存します： **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 5871cea65..90b5e2d98 100644 --- a/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/ja/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -2,7 +2,7 @@ title: "Cosmos Predict2 Video2World ComfyUI 公式例" description: "このガイドでは、ComfyUI で Cosmos-Predict2 Video2World ワークフローを完了する方法を示します" sidebarTitle: "Cosmos-Predict2" -translationSourceHash: 7c20daa6 +translationSourceHash: 3386fb94 translationFrom: tutorials/video/cosmos/cosmos-predict2-video2world.mdx --- @@ -41,9 +41,14 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - -

Json 形式ワークフローファイルをダウンロード

- + + + Download the JSON format workflow file + + + Run this workflow on Comfy Cloud with pre-installed models + + 入力として以下の画像をダウンロードしてください。 @@ -55,17 +60,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **Diffusion model** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) +- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) 他の重みについては、[Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) にアクセスしてダウンロードしてください **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) ファイル保存場所 ``` diff --git a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 5ca22c6eb..531f25507 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -2,10 +2,18 @@ title: "HunyuanVideo 1.5" description: "消費者向け GPU で高品質な動画生成を実現する、軽量な 8.3B パラメータモデルである HunyuanVideo 1.5 の使用方法について学びます" sidebarTitle: "HunyuanVideo 1.5" -translationSourceHash: 0b15fd89 +translationSourceHash: 4379c5f7 translationFrom: tutorials/video/hunyuan/hunyuan-video-1-5.mdx +translationBlockHashes: + "_intro": 74d9d0ad + "Model highlights": a18b0a64 + "Common models for all workflows": 87181049 + "Hunyuan Video 1.5 Text-to-Video Workflow": 73a4afe7 + "Hunyuan Video 1.5 Image-to-Video Workflow": 42a0612d + "Super-resolution upscaler": 89338328 --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; [HunyuanVideo 1.5](https://github.com/Tencent/HunyuanVideo) は、テンセントの Hunyuan チームによって開発された、軽量な 8.3B パラメータモデルです。消費者向け GPU（24GB VRAM）でフラッグシップ級の動画生成を実現し、品質を損なうことなく参入障壁を劇的に下げます。 @@ -29,17 +37,17 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## モデルリンク **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **diffusion_models** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **vae** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) モデルの保存場所 @@ -54,4 +62,4 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; │ │ ├── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors │ │ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors │ └── :open_file_folder: vae/ -│ └── hunyuanvideo15_vae_fp16.safetensors \ No newline at end of file +│ └── hunyuanvideo15_vae_fp16.safetensors diff --git a/ja/tutorials/video/hunyuan/hunyuan-video.mdx b/ja/tutorials/video/hunyuan/hunyuan-video.mdx index 61bee272e..d6952e34e 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video.mdx @@ -2,17 +2,18 @@ title: "ComfyUI Hunyuan Video 使用例" description: "このガイドでは、ComfyUI において Hunyuan のテキストから動画 (T2V) および画像から動画 (I2V) のワークフローを使用する方法を説明します" sidebarTitle: "Hunyuan Video" -translationSourceHash: 045f4b68 +translationSourceHash: 2cbcac4c translationFrom: tutorials/video/hunyuan/hunyuan-video.mdx translationBlockHashes: "_intro": 1af8ce94 - "Common Models for All Workflows": 3e4807f2 - "Hunyuan Text-to-Video Workflow": 8c3a1c81 - "Hunyuan Image-to-Video Workflow": ef169a18 - "Try it yourself": d4ee74b3 + "Common Models for All Workflows": 004bda25 + "Hunyuan Text-to-Video Workflow": fdc74d6f + "Hunyuan Image-to-Video Workflow": fef3b225 + "Try It Yourself": 076fb43f --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; - -

JSON ワークフローファイルをダウンロード

- + + + Download the image below to use as the starting frame for the 1.3B workflow + + 14B バージョンをご利用になりたい場合は、単にモデルファイルを 14B バージョンに置き換えてください。ただし、VRAM の要件にご注意ください。 @@ -119,9 +122,11 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B.mp4" > - -

JSON ワークフローファイルをダウンロード

- + + + Download the image below to use as the starting frame for the 14B workflow + + **入力画像** ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) diff --git a/ja/tutorials/video/wan/fun-control.mdx b/ja/tutorials/video/wan/fun-control.mdx index c1001c086..f2c93852d 100644 --- a/ja/tutorials/video/wan/fun-control.mdx +++ b/ja/tutorials/video/wan/fun-control.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Wan2.1 Fun Control 動画示例" description: "本ガイドでは、ComfyUI で Wan2.1 Fun Control を使用して制御動画で動画を生成する方法を説明します" sidebarTitle: "Wan2.1 Fun Control" -translationSourceHash: 88043153 +translationSourceHash: f6e4d92d translationFrom: tutorials/video/wan/fun-control.mdx translationBlockHashes: "_intro": 9efc0241 - "About Wan2.1-Fun-Control": 19f05e50 - "Model Installation": aee3a183 - "ComfyUI Native Workflow": 13eb7888 - "Workflow Using Custom Nodes": cf408ff0 - "Usage Tips": ae173b3b + "About Wan2.1-Fun-Control": 2e562079 + "Model Installation": ef08dc73 + "ComfyUI Native Workflow": d64a3fa4 + "Workflow Using Custom Nodes": ca1f733a + "Usage Tips": 9e74c201 --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## Wan2.1-Fun-Control について @@ -55,18 +56,18 @@ ComfyUI は現在、Wan2.1 Fun Control モデルを**ネイティブサポート 対応するリンクをクリックしてダウンロードしてください。以前に Wan 関連のワークフローを使用したことがある場合は、**Diffusion models** のみをダウンロードする必要があります。 **Diffusion models** - 1.3B または 14B を選択。14B バージョンはファイルサイズが大きく（32GB）、VRAM 要件も高くなります： -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true)：ダウンロード後に `Wan2.1-Fun-14B-Control.safetensors` にリネームしてください **Text encoders** - 以下のモデルのいずれかを選択（fp16 精度はサイズが大きく、パフォーマンス要件が高くなります）： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存場所： ``` diff --git a/ja/tutorials/video/wan/fun-inp.mdx b/ja/tutorials/video/wan/fun-inp.mdx index ccae9db7b..8d65fc98c 100644 --- a/ja/tutorials/video/wan/fun-inp.mdx +++ b/ja/tutorials/video/wan/fun-inp.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Wan2.1 Fun InP 動画サンプル" description: "このガイドでは、ComfyUI で Wan2.1 Fun InP を使用して、最初と最後のフレームを制御した動画を生成する方法について説明します" sidebarTitle: "Wan2.1 Fun InP" -translationSourceHash: db96fa7d +translationSourceHash: 866f2586 translationFrom: tutorials/video/wan/fun-inp.mdx translationBlockHashes: "_intro": 9efc0241 - "About Wan2.1-Fun-InP": ab94df56 - "Wan2.1 Fun InP Workflow": 33ad43c2 + "About Wan2.1-Fun-InP": f67d80ee + "Wan2.1 Fun InP Workflow": 140545b2 "Other Wan2.1 Fun InP or video-related custom node packages": 4c96c810 --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## Wan2.1-Fun-InP について @@ -53,18 +54,18 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; 以下のモデルは [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) と [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) で見つかります。 **Diffusion models** - 1.3B または 14B を選択してください。14B バージョンはファイルサイズが大きく (32GB)、VRAM 要件も高くなります： -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): ダウンロード後、`Wan2.1-Fun-14B-InP.safetensors` にリネームしてください +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true): ダウンロード後、`Wan2.1-Fun-14B-InP.safetensors` にリネームしてください **Text encoders** - 以下のモデルのいずれかを選択してください（fp16 精度はサイズが大きく、パフォーマンス要件も高くなります）： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存場所： ``` diff --git a/ja/tutorials/video/wan/vace.mdx b/ja/tutorials/video/wan/vace.mdx index 67687320f..6320b3eea 100644 --- a/ja/tutorials/video/wan/vace.mdx +++ b/ja/tutorials/video/wan/vace.mdx @@ -2,20 +2,22 @@ title: "ComfyUI Wan2.1 VACE 動画サンプル" description: "本記事では、ComfyUI で Wan VACE 動画生成のサンプルを完了する方法を紹介します。" sidebarTitle: "Wan2.1 VACE" -translationSourceHash: 9b04cff5 +translationSourceHash: 882bb348 translationFrom: tutorials/video/wan/vace.mdx translationBlockHashes: - "_intro": f6c2c850 - "About VACE": 4fa63090 - "Model Download and Loading in Workflows": 61bc1fa6 - "VACE Text-to-Video Workflow": 0275c575 - "VACE Image-to-Video Workflow": 6dbc84f2 - "VACE Video-to-Video Workflow": 0b58d1d2 - "VACE Video Outpainting Workflow": 22da2682 - "VACE First-Last Frame Video Generation": 0f2de66b + "_intro": 4e003b58 + "About VACE": 3d2b2490 + "Model Download and Loading in Workflows": 0a6b52f4 + "1. VACE Text-to-Video": 1715a0ab + "2. VACE Image-to-Video": 0ade4d39 + "3. VACE Video-to-Video": 4642a1ae + "4. VACE Inpainting": 6535794d + "5. VACE Video Outpainting": 24f1f2ff + "6. VACE First-Last Frame Video Generation": e58f2ea1 --- + import CancelBypass from '/snippets/ja/interface/cancel-bypass.mdx' import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -47,7 +49,9 @@ VACE 14B は、アリババ Tongyi Wanxiang チームが公開したオープン 関連するモデル重みおよびコードリポジトリ： -- [VACE-1.3B](https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B) + + 1.3B model weights on Hugging Face + - [VACE-14B](https://huggingface.co/Wan-AI/Wan2.1-VACE-14B) - [Github](https://github.com/ali-vilab/VACE) - [VACE プロジェクトホームページ](https://ali-vilab.github.io/VACE-Page/) @@ -61,18 +65,18 @@ VACE 14B は、アリババ Tongyi Wanxiang チームが公開したオープン ### モデルのダウンロード **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 以前に Wan Video 関連のワークフローをご利用になったことがある場合、以下のモデルファイルはすでにダウンロード済みです。 **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **Text encoders** からいずれか 1 つのバージョンを選択してダウンロードしてください： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) ファイルの保存先： ``` diff --git a/ja/tutorials/video/wan/wan-alpha.mdx b/ja/tutorials/video/wan/wan-alpha.mdx index 13f90b3ff..a140a854d 100644 --- a/ja/tutorials/video/wan/wan-alpha.mdx +++ b/ja/tutorials/video/wan/wan-alpha.mdx @@ -2,10 +2,15 @@ title: "Wan-Alpha チュートリアル" description: "ComfyUI で Wan-Alpha を使用して、アルファチャンネル透過性を備えた動画を生成する方法を学びます" sidebarTitle: "Wan-Alpha" -translationSourceHash: 82295242 +translationSourceHash: 775fde1b translationFrom: tutorials/video/wan/wan-alpha.mdx +translationBlockHashes: + "_intro": a57abe06 + "Resources": afef8889 + "Wan-Alpha Text-to-Video Workflow (14B)": 32a5a973 --- + Wan-Alpha は、アルファチャンネル透過性を備えた高品質な動画を生成する専用のテキストから動画を生成するモデルです。ベースモデルとして Wan2.1-14B-T2V を採用しており、透過背景および半透明オブジェクトを含む動画を生成します。合成ワークフローに最適です。 このモデルは、透過背景、半透明オブジェクト（気泡、ガラス、水）、発光エフェクト、および適切なアルファチャンネルを備えた精細なディテール（髪の毛、煙、粒子）の生成に優れています。 @@ -29,4 +34,4 @@ Wan-Alpha は、アルファチャンネル透過性を備えた高品質な動 - [Wan-Alpha GitHub](https://github.com/WeChatCV/Wan-Alpha) - [Hugging Face モデル](https://huggingface.co/htdong/Wan-Alpha) - [ComfyUI 版](https://huggingface.co/htdong/Wan-Alpha_ComfyUI) -- [研究論文](https://arxiv.org/pdf/2509.24979) \ No newline at end of file +- [研究論文](https://arxiv.org/pdf/2509.24979) diff --git a/ja/tutorials/video/wan/wan-ati.mdx b/ja/tutorials/video/wan/wan-ati.mdx index d239223b3..e05c21192 100644 --- a/ja/tutorials/video/wan/wan-ati.mdx +++ b/ja/tutorials/video/wan/wan-ati.mdx @@ -2,15 +2,16 @@ title: "Wan ATI ComfyUI ネイティブワークフローチュートリアル" description: "動画生成における軌道制御の使用方法。" sidebarTitle: "Wan2.1 ATI" -translationSourceHash: a8716511 +translationSourceHash: 39439781 translationFrom: tutorials/video/wan/wan-ati.mdx translationBlockHashes: - "_intro": 247f391d + "_intro": 52f7de29 "Key Features": 67ce28af - "WAN ATI Trajectory Control Workflow Example": 01ea7d04 + "WAN ATI Trajectory Control Workflow Example": f704f2c5 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -46,17 +47,17 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ワークフローからモデルファイルを正常にダウンロードできていない場合、以下のリンクから手動でダウンロードしてみてください。 **Diffusionモデル** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **テキストエンコーダー**（以下のいずれか1つを選択） -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) ファイル保存先 ``` diff --git a/ja/tutorials/video/wan/wan-causal-forcing.mdx b/ja/tutorials/video/wan/wan-causal-forcing.mdx index 4b1420be4..3ec115972 100644 --- a/ja/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ja/tutorials/video/wan/wan-causal-forcing.mdx @@ -2,8 +2,17 @@ title: "Causal Forcing 画像から動画 ComfyUI ワークフロー例" description: "Wan2.1 の Causal Forcing または Causal Forcing++ を使用して、画像から動画を生成します。わずか 1 ステップの推論で滑らかで時間的に一貫性のある動画を実現します。" sidebarTitle: "Causal Forcing I2V" +translationSourceHash: 4f300f64 +translationFrom: tutorials/video/wan/wan-causal-forcing.mdx +translationBlockHashes: + "_intro": dbdb8d89 + "How it works": ea3eaaf0 + "Using the workflow": 5e6245a2 + "Steps to run": 3202924b + "Model downloads": c6ce079e --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Causal Forcing（因果強制）** は、推論時に **リカレント条件付け** を適用する動画生成技術です。各フレームが生成されるたびにモデルにフィードバックされ、次のフレームを予測するために使用されます。たった **1 〜 4 ステップの推論** で、1 枚の開始画像から滑らかで時間的に一貫性のある動画を生成できます。 @@ -83,10 +92,10 @@ Wan2.1 I2V モデルと必要なファイルをダウンロードします。対 ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B チェックポイント - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B チェックポイント（最小 8GB VRAM） @@ -94,10 +103,10 @@ Wan2.1 I2V モデルと必要なファイルをダウンロードします。対 ### CLIP と VAE - + google-bert/bert-base-uncased — CLIP テキストエンコーダー - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/ja/tutorials/video/wan/wan-dancer.mdx b/ja/tutorials/video/wan/wan-dancer.mdx index 6ea08b40f..be2c75f0e 100644 --- a/ja/tutorials/video/wan/wan-dancer.mdx +++ b/ja/tutorials/video/wan/wan-dancer.mdx @@ -2,16 +2,17 @@ title: "Wan Dancer: 音楽からダンスビデオ生成" description: "Wan 2.2 を基盤とした階層型オーディオ駆動フレームワーク Wan Dancer を使用して、音楽から分単位の一貫性のあるダンス動画を生成します。参照画像とオーディオを入力して、同期したダンスビデオを生成します。" sidebarTitle: "Wan Dancer" -translationSourceHash: ea7690e9 +translationSourceHash: 73c80f3f translationFrom: tutorials/video/wan/wan-dancer.mdx translationBlockHashes: "_intro": c890f065 - "Model Highlights": 06703fea + "Model Highlights": 0826f4bf "Workflow Overview": 0fa086f1 - "Wan Dancer Workflow": fbc54f05 - "Model Information": 97e02e88 + "Wan Dancer Workflow": 2251f138 + "Model Information": bc2b5961 "Report Issues": 271203b5 --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" Wan Dancerは、Wan 2.2アーキテクチャを基盤としたオーディオ駆動のダンス動画生成モデルです。グローバルモデルとローカルモデルからなる階層的フレームワークを使用し、入力された音楽に同期した一貫性のある表現豊かなダンス動画を生成します。 @@ -61,20 +62,20 @@ ComfyUI を最新バージョンに更新し、ワークフローファイルを ### 3. 手動でのモデルのダウンロード **拡散モデル** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **テキストエンコーダ** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP ビジョン** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan-flf.mdx b/ja/tutorials/video/wan/wan-flf.mdx index 7e4a45626..1d53b7575 100644 --- a/ja/tutorials/video/wan/wan-flf.mdx +++ b/ja/tutorials/video/wan/wan-flf.mdx @@ -2,7 +2,7 @@ title: "ComfyUI Wan2.1 FLF2V ネイティブ例" description: "本ガイドでは、ComfyUI で Wan2.1 FLF2V 動画生成のサンプルを実行する方法について説明します" sidebarTitle: "Wan2.1 FLF2V" -translationSourceHash: 9b049266 +translationSourceHash: 5b7a25cb translationFrom: tutorials/video/wan/wan-flf.mdx --- @@ -56,7 +56,7 @@ Wan FLF2V（First-Last Frame Video Generation：始終フレーム動画生成 本ガイドで使用するすべてのモデルは、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)から入手できます。 **diffusion_models**：ご使用のハードウェア環境に応じて、以下のいずれかのバージョンを選択してください。 -- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8: [wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -64,14 +64,14 @@ Wan FLF2V（First-Last Frame Video Generation：始終フレーム動画生成 **Text encoders**：以下のいずれか1つのバージョンをダウンロードしてください。 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイル保存先 diff --git a/ja/tutorials/video/wan/wan-move.mdx b/ja/tutorials/video/wan/wan-move.mdx index 02d87136c..cbd406952 100644 --- a/ja/tutorials/video/wan/wan-move.mdx +++ b/ja/tutorials/video/wan/wan-move.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan-Move ワークフロー例" description: "Wan-Move は、潜在トラジェクトリガイダンスを通じて運動を制御可能な動画生成モデルであり、画像から動画への生成において細粒度のポイントレベル運動制御を可能にします。" sidebarTitle: "Wan-Move" -translationSourceHash: 1b11ad45 +translationSourceHash: dda86ac3 translationFrom: tutorials/video/wan/wan-move.mdx translationBlockHashes: "_intro": 67a26359 - "Wan-Move image-to-video workflow": ec83b0e4 - "Model links": 25e1a28d + "Wan-Move image-to-video workflow": b77082ab + "Model links": 8e0ef166 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Wan-Move** は、アリババの通義実験室（Tongyi Lab）によって開発された、運動制御可能な動画生成フレームワークです。入力画像上でポイントのトラジェクトリ（軌跡）を指定することで、生成された動画内の物体の運動を制御でき、画像から動画への生成をより正確かつ制御可能にします。 @@ -28,13 +29,41 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Wan-Move 画像から動画へのワークフロー - -

JSON ワークフローファイルをダウンロード

- - - -

ComfyUI Cloud で実行

- + + + Comfy Cloud で開く + + + Download JSON or search "Wan-Move Motion-Control" in Template Library + + + Download the default input image, or use your own image as the starting frame. + + + Place in ComfyUI/models/diffusion_models/ + + + Place in ComfyUI/models/loras/ + + + Place in ComfyUI/models/text_encoders/ + + + Place in ComfyUI/models/clip_vision/ + + + Place in ComfyUI/models/vae/ + + + + + + Comfy Cloud で開く + + + Download JSON or search "Wan-Move Motion-Control" in Template Library + + @@ -42,23 +71,23 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **diffusion_models** -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) +- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **モデル保存場所** diff --git a/ja/tutorials/video/wan/wan-video.mdx b/ja/tutorials/video/wan/wan-video.mdx index 73df834b7..3cb045530 100644 --- a/ja/tutorials/video/wan/wan-video.mdx +++ b/ja/tutorials/video/wan/wan-video.mdx @@ -2,18 +2,19 @@ title: ComfyUI Wan2.1 動画生成のサンプル description: 「このガイドでは、ComfyUI で Wan2.1 Video を使用して動画の最初と最後のフレームを生成する方法を紹介します」 sidebarTitle: Wan2.1 -translationSourceHash: e0c0f714 +translationSourceHash: c11f515a translationFrom: tutorials/video/wan/wan-video.mdx translationBlockHashes: - "_intro": 228a7575 + "_intro": 376d341e "Wan2.1 ComfyUI Native Workflow Examples": b092012f - "Model Installation": 0bd3e4c2 - "Wan2.1 Text-to-Video Workflow": 611b9bf9 - "Wan2.1 Image-to-Video Workflow": bee5fc35 + "Model Installation": 21d2cbd6 + "Wan2.1 Text-to-Video Workflow (1.3B)": 268e4e8b + "Wan2.1 Image-to-Video Workflow (14B)": dfb19590 --- + Wan2.1 Video シリーズは、アリババ社が 2025 年 2 月に [Apache 2.0 ライセンス](https://github.com/Wan-Video/Wan2.1?tab=Apache-2.0-1-ov-file) の下でオープンソース化した動画生成モデルです。 このモデルには以下の 2 つのバージョンがあります： - 14B（140 億パラメータ） @@ -41,14 +42,16 @@ Wan2.1 Video シリーズは、アリババ社が 2025 年 2 月に [Apache 2.0 このガイドで言及されるすべてのモデルは、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files) から入手できます。以下は、このガイドのサンプルで使用する共通のモデルであり、事前にダウンロードしておくことを推奨します： **Text encoders** からいずれか 1 つのバージョンを選択してダウンロードしてください： -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) + + FP16 精度のテキストエンコーダー。ComfyUI/models/text_encoders/ に配置 + +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) ファイルの保存先ディレクトリ構成： ``` @@ -70,7 +73,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル ## Wan2.1 テキスト→動画（T2V）ワークフロー -ワークフローを開始する前に、[wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +ワークフローを開始する前に、[wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 > 他の T2V 精度バージョンが必要な場合は、[こちら](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) からダウンロードしてください。 @@ -107,7 +110,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル ![Wan2.1 画像→動画ワークフロー（14B、480P）の入力画像サンプル](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) #### 2. モデルのダウンロード -[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 #### 3. ワークフローをステップごとに実行 @@ -135,7 +138,7 @@ diffusion モデルについては、本ガイドでは fp16 精度のモデル #### 2. モデルのダウンロード -[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 +[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true) をダウンロードし、`ComfyUI/models/diffusion_models/` ディレクトリに保存してください。 #### 3. ワークフローをステップごとに実行 diff --git a/ja/tutorials/video/wan/wan2-2-animate.mdx b/ja/tutorials/video/wan/wan2-2-animate.mdx index 314d5e51b..4cef9f7ec 100644 --- a/ja/tutorials/video/wan/wan2-2-animate.mdx +++ b/ja/tutorials/video/wan/wan2-2-animate.mdx @@ -2,16 +2,16 @@ title: "Wan2.2 Animate ComfyUI ネイティブワークフロー" description: "正確な動きおよび表情の再現を実現する統一的人物アニメーション・置換フレームワーク。" sidebarTitle: "Wan2.2 Animate" -translationSourceHash: 8d8cb435 +translationSourceHash: 07849135 translationFrom: tutorials/video/wan/wan2-2-animate.mdx translationBlockHashes: "_intro": 96eb0424 "Model Highlights": fc4e7349 "ComfyOrg Wan2.2 Animate stream replay": a43ba31b - "About Wan2.2 Animate workflow": f3713999 - "Wan2.2 Anmate ComfyUI native workflow(without custom nodes)": fc4b2e18 + "About Wan2.2 Animate workflow": ea4c1d9a --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' Wan-Animate は、WAN チームが開発した人物アニメーションおよび置換のための統合フレームワークです。 @@ -55,13 +55,23 @@ Wan-Animate は、WAN チームが開発した人物アニメーションおよ 以下のワークフローファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして読み込んでください。 - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Comfy Cloud で開く + + + Download JSON or search "Wan2.2 Animate" in Template Library + + + + + + Comfy Cloud で開く + + + Download JSON or search "Wan2.2 Animate" in Template Library + + 以下の素材を入力としてダウンロードしてください： @@ -78,20 +88,20 @@ Wan-Animate は、WAN チームが開発した人物アニメーションおよ ### 2. モデルのダウンロードリンク **diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) — Kijai のリポジトリから提供されるモデル -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) — 元のモデル重み +- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) — Kijai のリポジトリから提供されるモデル +- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) — 元のモデル重み **clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) — 4ステップ高速化対応の LoRA +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) — 4ステップ高速化対応の LoRA **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-fun-camera.mdx b/ja/tutorials/video/wan/wan2-2-fun-camera.mdx index 2636c306a..81e812017 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -2,7 +2,7 @@ title: "ComfyUI Wan2.2 Fun Camera Control：カメラ制御による動画生成ワークフローの例" description: "本記事では、ComfyUI で Wan2.2 Fun Camera Control を用いてカメラ制御を活用した動画生成を行う方法を解説します。" sidebarTitle: "Wan2.2 Fun Camera" -translationSourceHash: c2df0664 +translationSourceHash: d7f3bd26 translationFrom: tutorials/video/wan/wan2-2-fun-camera.mdx --- @@ -46,9 +46,35 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

JSON ワークフローをダウンロード

- + + + Comfy Cloud で開く + + + Download JSON or search "Wan2.2 Fun Camera" in Template Library + + + Starting frame for the video generation. Download and use this image, or replace with your own. + + + High noise diffusion model for Wan2.2 Fun Camera + + + Low noise diffusion model for Wan2.2 Fun Camera + + + 4-step acceleration LoRA for high noise model + + + 4-step acceleration LoRA for low noise model + + + Wan2.1 VAE for encoding/decoding + + + FP8 scaled text encoder + + 以下の画像をダウンロードし、入力として使用します。 @@ -59,18 +85,18 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下のモデルは、[Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 **Diffusion モデル** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA（オプション：高速化用）** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **テキストエンコーダー** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ファイル保存先のディレクトリ構成： diff --git a/ja/tutorials/video/wan/wan2-2-fun-control.mdx b/ja/tutorials/video/wan/wan2-2-fun-control.mdx index 87fd9ec0c..30c9883a2 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-control.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan2.2 Fun Control 動画生成例" description: "この記事では、ComfyUI を使用して Wan2.2 Fun Control により制御動画を用いた動画生成を行う方法について紹介します。" sidebarTitle: "Wan2.2 Fun Control" -translationSourceHash: 59b97fdc +translationSourceHash: f1c35d6e translationFrom: tutorials/video/wan/wan2-2-fun-control.mdx translationBlockHashes: "_intro": 239e24b6 "ComfyOrg Wan2.2 Fun InP & Control Youtube Live Stream Replay": 22557a2e - "Wan2.2 Fun Control Video Generation Workflow Example": 68f53061 + "Wan2.2 Fun Control Video Generation Workflow Example": cbbb7456 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Wan2.2-Fun-Control** は、Alibaba PAI チームによってリリースされた次世代の動画生成・制御モデルです。革新的な Control Codes 機制を導入し、深層学習とマルチモーダル条件入力を組み合わせることで、预设された制御条件に準拠した高品質な動画を生成できます。本モデルは **Apache 2.0 ライセンス** でリリースされており、商用利用も可能です。 @@ -63,9 +64,38 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/wan2.2_14B_fun_inp.mp4" > - -

JSON ワークフローをダウンロード

- + + + Comfy Cloud で開く + + + Download JSON or search "Wan2.2 Fun Control" in Template Library + + + Start frame for video generation. Download and use this image, or replace with your own. + + + Preprocessed pose control video. Download and use this video, or replace with your own. + + + wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors: High noise diffusion model + + + wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors: Low noise diffusion model + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: High noise 4-step acceleration LoRA + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: Low noise 4-step acceleration LoRA + + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding/decoding + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + + 入力素材として以下の画像および動画をダウンロードしてください。 @@ -84,18 +114,18 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 以下のモデルは [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) で見つかります **Diffusion Model** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA (オプション、加速用)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx index b00d7fa5d..c8d400a7e 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan2.2 Fun Inp 首尾フレーム制御による動画生成の例" description: "本記事では、ComfyUI を用いて Wan2.2 Fun Inp の首尾フレーム制御による動画生成の例を実行する方法について説明します。" sidebarTitle: "Wan2.2 Fun Inp" -translationSourceHash: 9d81688f +translationSourceHash: 714331ba translationFrom: tutorials/video/wan/wan2-2-fun-inp.mdx translationBlockHashes: - "_intro": 75061370 + "_intro": 7a3595d7 "ComfyOrg Wan2.2 Fun InP & Control Youtube Live Stream Replay": 22557a2e - "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 39f39063 + "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 692b42a8 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Wan2.2-Fun-Inp** は、Alibaba PAI チームが開発・公開した首尾フレーム制御型動画生成モデルです。ユーザーは**開始フレーム画像と終了フレーム画像**を入力することで、それらの間を滑らかに遷移する中間動画を生成できます。これにより、クリエイターはより高度な創造的コントロールを実現できます。本モデルは **Apache 2.0 ライセンス**のもとで公開されており、商用利用も可能です。 @@ -61,13 +62,47 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem または、ComfyUI を最新版に更新した上で、以下のリンクからワークフローファイルをダウンロードし、ComfyUI の画面にドラッグ＆ドロップして読み込んでください。 - -

JSON形式ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Download JSON or search "Wan2.2 Fun Inp" in Template Library + + + Comfy Cloud で開く + + + Start frame for video generation. Download and use this image, or replace with your own. + + + End frame for video generation. Download and use this image, or replace with your own. + + + wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: High noise diffusion model for start-end frame inpainting + + + wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: Low noise diffusion model for start-end frame inpainting + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 4-step acceleration LoRA for high noise model + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 4-step acceleration LoRA for low noise model + + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + + + + + + Download JSON or search "Wan2.2 Fun Inp" in Template Library + + + Comfy Cloud で開く + + 以下の画像を開始フレームおよび終了フレームの素材としてご使用ください。 @@ -77,18 +112,18 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem ### 2. モデルの準備 **Diffusion モデル** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) **Lightning LoRA（任意：高速化用）** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **テキストエンコーダー** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ja/tutorials/video/wan/wan2-2-s2v.mdx b/ja/tutorials/video/wan/wan2-2-s2v.mdx index 872e49120..3fa4d0882 100644 --- a/ja/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ja/tutorials/video/wan/wan2-2-s2v.mdx @@ -2,7 +2,7 @@ title: Wan2.2-S2V 音声駆動型動画生成のための ComfyUI ネイティブワークフロー例 description: これは、ComfyUI における Wan2.2-S2V 音声駆動型動画生成のネイティブワークフローの例です。 sidebarTitle: "Wan2.2 S2V" -translationSourceHash: 8d1b0bf2 +translationSourceHash: 10604cda translationFrom: tutorials/video/wan/wan2-2-s2v.mdx --- @@ -35,38 +35,69 @@ Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

JSON ワークフローをダウンロード

- - - -

Comfy Cloud で実行

- + + + Comfy Cloud で開く + + + Download JSON or search "Wan2.2 S2V" in Template Library + + + Download the default input image, or use your own image. + + + Download the default input audio, or use your own audio. + + + FP8 scaled diffusion model. Place in ComfyUI/models/diffusion_models/ + + + BF16 diffusion model. Place in ComfyUI/models/diffusion_models/ + + + Audio encoder model. Place in ComfyUI/models/audio_encoders/ + + + Wan2.1 VAE model. Place in ComfyUI/models/vae/ + + + FP8 scaled text encoder. Place in ComfyUI/models/text_encoders/ + + + + + + Comfy Cloud で開く + + + Download JSON or search "Wan2.2 S2V" in Template Library + + 以下の画像および音声ファイルを入力としてダウンロードしてください: ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - -

入力音声をダウンロード

- + ### 2. モデルのダウンロードリンク すべてのモデルは、[当社のリポジトリ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 **diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) + + FP8 scaled diffusion model + +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) **audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) +- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -94,8 +125,8 @@ ComfyUI/ 両方のモデルは、[こちらのページ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) から入手可能です： -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) 本テンプレートでは `wan2.2_s2v_14B_fp8_scaled.safetensors` を使用しており、VRAM 使用量が少ないのが特徴です。ただし、品質劣化を抑えるために `wan2.2_s2v_14B_bf16.safetensors` を試すことも可能です。 diff --git a/ja/tutorials/video/wan/wan2_2.mdx b/ja/tutorials/video/wan/wan2_2.mdx index 7805a5892..ed58201ae 100644 --- a/ja/tutorials/video/wan/wan2_2.mdx +++ b/ja/tutorials/video/wan/wan2_2.mdx @@ -2,21 +2,22 @@ title: "Wan2.2 動画生成 ComfyUI 公式ネイティブワークフロー例" description: "ComfyUI における Alibaba Cloud Tongyi Wanxiang 2.2 動画生成モデルの公式使用ガイド" sidebarTitle: "Wan2.2" -translationSourceHash: 70e49209 +translationSourceHash: a75c822c translationFrom: tutorials/video/wan/wan2_2.mdx translationBlockHashes: "_intro": 3d02f6d3 "Model Highlights": fc7121c1 "Wan2.2 Open Source Model Versions": 7ed2d913 "ComfyOrg Wan2.2 Live Streams": d380a603 - "Wan2.2 TI2V 5B Hybrid Version Workflow Example": b0e108bd - "Wan2.2 14B T2V Text-to-Video Workflow Example": 9eaf9107 - "Wan2.2 14B I2V Image-to-Video Workflow Example": a006f8aa - "Wan2.2 14B FLF2V Workflow Example": 525e0946 + "Wan2.2 TI2V 5B Hybrid Version Workflow Example": a9d53e08 + "Wan2.2 14B T2V Text-to-Video Workflow Example": 6a6398c7 + "Wan2.2 14B I2V Image-to-Video Workflow Example": 356d6b7a + "Wan2.2 14B FLF2V Workflow Example": ca3b4239 "Community Resources": 7463b48b --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -45,19 +46,19 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 ## Flux.2 Klein 4B 모델 다운로드 - + 4B 모델용 텍스트 인코더입니다. - + 확산 모델(4B 베이스). - + 확산 모델(4B 정제). - + 4B 모델용 VAE입니다. @@ -103,11 +104,11 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 확산 모델(9B 정제). - + 9B 모델용 텍스트 인코더입니다. - + 9B 모델용 VAE입니다. diff --git a/ko/tutorials/flux/flux1-krea-dev.mdx b/ko/tutorials/flux/flux1-krea-dev.mdx index f1cba3b61..8940a7003 100644 --- a/ko/tutorials/flux/flux1-krea-dev.mdx +++ b/ko/tutorials/flux/flux1-krea-dev.mdx @@ -2,7 +2,7 @@ title: "Flux.1 Krea Dev ComfyUI 워크플로우 튜토리얼" description: "Black Forest Labs가 Krea와 협력해 개발한 최고의 오픈소스 FLUX 모델로, 독특한 미적 스타일과 자연스러운 디테일에 중점을 두며 AI 특유의 느낌을 피하고 뛰어난 사실성과 이미지 품질을 제공합니다." sidebarTitle: "Flux.1 Krea Dev" -translationSourceHash: e4ac73ab +translationSourceHash: 202e24cf translationFrom: tutorials/flux/flux1-krea-dev.mdx --- @@ -31,13 +31,11 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 아래 이미지 또는 JSON 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. ![Flux Krea Dev 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

JSON 워크플로우 다운로드

- + + 이 워크플로우를 Comfy Cloud에서 실행하세요 + JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Krea Dev" 검색 + - -

Comfy Cloud에서 실행하기

- #### 2. 수동 모델 설치 @@ -48,7 +46,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 더 높은 품질을 원하고 VRAM이 충분하다면 원본 모델 가중치를 사용해 보실 수 있습니다. -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) `flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 약정에 동의해야 합니다. @@ -57,12 +55,12 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 이전에 Flux 관련 워크플로우를 사용한 적이 있다면, 다음 모델들은 동일하므로 다시 다운로드할 필요가 없습니다. **텍스트 인코더** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/image/hidream/hidream-e1.mdx b/ko/tutorials/image/hidream/hidream-e1.mdx index 86a1e2e38..7aa3f3868 100644 --- a/ko/tutorials/image/hidream/hidream-e1.mdx +++ b/ko/tutorials/image/hidream/hidream-e1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI 네이티브 HiDream-E1, E1.1 워크플로우 예시" sidebarTitle: "HiDream-e1" description: "이 가이드는 ComfyUI 네이티브 HiDream-I1 텍스트 기반 이미지 생성 워크플로우 예시를 이해하고 완료하는 데 도움을 줄 것입니다." -translationSourceHash: 76ea9fff +translationSourceHash: 7554c00c translationFrom: tutorials/image/hidream/hidream-e1.mdx translationBlockHashes: "_intro": 6829ae90 - "HiDream E1 and E1.1 Workflow Related Models": e648e936 - "HiDream E1.1 ComfyUI Native Workflow Example": 4134dfe8 - "HiDream E1 ComfyUI Native Workflow Example": 778b966a + "HiDream E1 and E1.1 Workflow Related Models": 25f9249b + "HiDream E1.1 ComfyUI Native Workflow Example": 4437f406 + "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ![HiDream-E1 데모](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -42,8 +43,8 @@ HiDream-E1은 HiDream-ai가 공식적으로 오픈소스로 배포한 대규모 **Diffusion 모델** 두 모델을 모두 다운로드할 필요는 없습니다. E1.1은 E1을 기반으로 한 반복 버전이므로, 우리의 테스트 결과에 따르면 E1보다 품질과 성능이 크게 향상되었습니다. -- [hidream_e1_1_bf16.safetensors (권장)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors (권장)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **텍스트 인코더**: @@ -119,9 +120,10 @@ E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전 ## HiDream E1 ComfyUI 네이티브 워크플로우 예시 - -

Comfy Cloud에서 실행

- + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "HiDream E1 Full" 검색 + E1은 2025년 4월 28일에 출시된 모델입니다. 이 모델은 768*768 해상도만 지원합니다. diff --git a/ko/tutorials/image/hidream/hidream-i1.mdx b/ko/tutorials/image/hidream/hidream-i1.mdx index 2b27ec88d..64a823e2e 100644 --- a/ko/tutorials/image/hidream/hidream-i1.mdx +++ b/ko/tutorials/image/hidream/hidream-i1.mdx @@ -2,15 +2,16 @@ title: "ComfyUI 네이티브 HiDream-I1 텍스트 기반 이미지 생성 워크플로 예시" sidebarTitle: "HiDream-I1" description: "이 가이드에서는 ComfyUI 네이티브 HiDream-I1 텍스트 기반 이미지 생성 워크플로 예시를 완료하는 방법을 안내합니다." -translationSourceHash: 590b8d6d +translationSourceHash: 73e39f2a translationFrom: tutorials/image/hidream/hidream-i1.mdx translationBlockHashes: "_intro": 59935bb2 "Model Features": 3ec455b0 "About This Workflow Example": 90c5d2b3 - "HiDream-I1 Workflow": 93e17294 + "HiDream-I1 Workflow": 5956f05c "Other Related Resources": e794d9ef --- + ![HiDream-I1 데모](https://raw.githubusercontent.com/HiDream-ai/HiDream-I1/main/assets/demo.jpg) HiDream-I1은 2025년 4월 7일 HiDream-ai가 공식적으로 오픈소스로 공개한 텍스트 기반 이미지 생성 모델입니다. 이 모델은 170억 파라미터를 보유하며, [MIT 라이선스](https://github.com/HiDream-ai/HiDream-I1/blob/main/LICENSE)에 따라 배포되며 개인 프로젝트, 과학 연구 및 상업적 사용을 지원합니다. 현재 여러 벤치마크 테스트에서 뛰어난 성능을 발휘하고 있습니다. @@ -98,16 +99,19 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 풀버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + Run this workflow on Comfy Cloud with zero setup + + 워크플로우 JSON 파일 다운로드 + #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 @@ -136,15 +140,18 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Dev 버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + Run this workflow on Comfy Cloud with zero setup + + 워크플로우 JSON 파일 다운로드 + #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. @@ -172,15 +179,18 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Fast 버전 워크플로우 - -

Comfy Cloud에서 실행

- + + + Run this workflow on Comfy Cloud with zero setup + + 워크플로우 JSON 파일 다운로드 + #### 1. 모델 파일 다운로드 하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. -- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. +- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. #### 2. 워크플로우 파일 다운로드 아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. diff --git a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 798943559..d42173bc1 100644 --- a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI NewBie-image-Exp0.1 워크플로 예제" description: "NewBie-image-Exp0.1은 Next-DiT 아키텍처를 기반으로 한 3.5B 파라미터의 애니메이션 스타일 텍스트 기반 이미지 생성 모델로, XML 구조화된 프롬프트로 고품질 애니메이션 이미지 생성을 최적화했습니다." sidebarTitle: "NewBie-image-Exp0.1" -translationSourceHash: f56facc0 +translationSourceHash: 22db4899 translationFrom: tutorials/image/newbie-image/newbie-image-exp-0-1.mdx translationBlockHashes: "_intro": 7f6f1284 - "NewBie-image text-to-image workflow": af50f161 + "NewBie-image text-to-image workflow": e7fbe7f5 "Model links": cd75bb7e "Prompt format": 5b819c6c --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **NewBie-image-Exp0.1**은 NewBieAI Lab에서 개발한 3.5B 파라미터의 DiT 모델로, 애니메이션 스타일 텍스트 기반 이미지 생성을 위해 설계되었습니다. Next-DiT 아키텍처를 기반으로 하며, 뛰어난 디테일과 시각적으로 강렬한 애니메이션 이미지를 제공합니다. @@ -30,13 +31,11 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## NewBie-image 텍스트 기반 이미지 생성 워크플로 - -

JSON 워크플로 파일 다운로드

- + + JSON 다운로드 또는 템플릿 라이브러리에서 "NewBie-image" 검색 + 클라우드에서 열기 + - -

ComfyUI 클라우드에서 실행하기

- @@ -44,16 +43,16 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/omnigen/omnigen2.mdx b/ko/tutorials/image/omnigen/omnigen2.mdx index 78ed966cf..93d7e3ed6 100644 --- a/ko/tutorials/image/omnigen/omnigen2.mdx +++ b/ko/tutorials/image/omnigen/omnigen2.mdx @@ -2,17 +2,18 @@ title: "ComfyUI OmniGen2 네이티브 워크플로우 예시" description: "ComfyUI OmniGen2 네이티브 워크플로우 예시 - 통합 텍스트 기반 이미지 생성, 이미지 편집 및 다중 이미지 합성 모델" sidebarTitle: "OmniGen2" -translationSourceHash: b061ff8c +translationSourceHash: 1429ffa8 translationFrom: tutorials/image/omnigen/omnigen2.mdx translationBlockHashes: "_intro": 3e9c5a21 "About OmniGen2": 2e6118e0 "OmniGen2 Model Download": 7997f8fe - "ComfyUI OmniGen2 Text-to-Image Workflow": f7650507 - "ComfyUI OmniGen2 Image Editing Workflow": 1859e94c + "ComfyUI OmniGen2 Text-to-Image Workflow": 5f9f63be + "ComfyUI OmniGen2 Image Editing Workflow": 1e06072e --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## OmniGen2 소개 @@ -42,13 +43,13 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 본 문서에는 다양한 워크플로우가 포함되어 있으므로, 해당 모델 파일과 설치 위치는 다음과 같습니다. 모델 파일의 다운로드 정보는 각 워크플로우에도 포함되어 있습니다. **확산 모델** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) 파일 저장 위치: ``` @@ -66,9 +67,11 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 ### 1. 워크플로우 파일 다운로드 - -

Comfy Cloud에서 실행

- + + + Open and run this workflow directly in Comfy Cloud. + + ![텍스트 기반 이미지 생성 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -94,9 +97,11 @@ OmniGen2는 풍부한 이미지 편집 기능을 갖추고 있으며, 이미지 ### 1. 워크플로우 파일 다운로드 - -

Comfy Cloud에서 실행

- + + + Open and run this workflow directly in Comfy Cloud. + + ![이미지 편집 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) diff --git a/ko/tutorials/image/ovis/ovis-image.mdx b/ko/tutorials/image/ovis/ovis-image.mdx index 6586e2039..dc0ede6d9 100644 --- a/ko/tutorials/image/ovis/ovis-image.mdx +++ b/ko/tutorials/image/ovis/ovis-image.mdx @@ -2,7 +2,7 @@ title: "Ovis-Image ComfyUI 워크플로우 예시" description: "Ovis-Image는 고화질 텍스트 렌더링을 위해 특별히 최적화된 7B 텍스트 기반 이미지 생성 모델로, 엄격한 계산 제약 조건 하에서도 효율적으로 작동하도록 설계되었습니다." sidebarTitle: "Ovis-Image" -translationSourceHash: 64c10a35 +translationSourceHash: 8d005c1d translationFrom: tutorials/image/ovis/ovis-image.mdx --- @@ -22,13 +22,11 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Ovis-Image 텍스트 기반 이미지 생성 워크플로우 - -

JSON 워크플로우 파일 다운로드

- + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "Ovis image" 검색 + - -

ComfyUI 클라우드에서 실행하기

- @@ -36,15 +34,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/pixeldit/pixeldit.mdx b/ko/tutorials/image/pixeldit/pixeldit.mdx index bdd451a41..17a0cdd5b 100644 --- a/ko/tutorials/image/pixeldit/pixeldit.mdx +++ b/ko/tutorials/image/pixeldit/pixeldit.mdx @@ -2,15 +2,16 @@ title: "PixelDiT ComfyUI 워크플로우 예시" description: "PixelDiT는 1024px 텍스트 기반 이미지 생성을 위한 NVIDIA의 픽셀 공간 확산 변환기입니다. 이 모델은 직접 픽셀 공간에서 작동하며, VAE 인코딩/디코딩이 필요하지 않습니다." sidebarTitle: "PixelDiT" -translationSourceHash: d5838b66 +translationSourceHash: b3280f94 translationFrom: tutorials/image/pixeldit/pixeldit.mdx translationBlockHashes: "_intro": ef3a0250 - "PixelDiT text-to-image workflow": 21b069a6 + "PixelDiT text-to-image workflow": 1879b3e8 "Model downloads": e4bafb0a --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" **PixelDiT**는 1024px 텍스트 기반 이미지 생성을 위한 NVIDIA의 픽셀 공간 확산 변환기입니다. 잠재 공간에서 작동하는 전통적인 확산 모델과 달리 PixelDiT는 패치 수준 DiT와 픽셀 수준 DiT를 결합한 이중 레벨 DiT 아키텍처를 사용해 직접 픽셀 공간에서 이미지를 생성하며, MM-DiT 융합을 통해 텍스트와 이미지 토큰 간의 공동 주의를 구현합니다. @@ -62,11 +63,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" PixelDiT는 두 개의 모델 파일을 사용합니다: 텍스트 인코더와 확산 모델입니다. - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 텍스트 인코더 - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 확산 모델 diff --git a/ko/tutorials/image/qwen/qwen-image-2512.mdx b/ko/tutorials/image/qwen/qwen-image-2512.mdx index e8b9a4dc4..2509ec437 100644 --- a/ko/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ko/tutorials/image/qwen/qwen-image-2512.mdx @@ -2,15 +2,16 @@ title: "Qwen-Image-2512 ComfyUI 네이티브 워크플로우 예시" description: "Qwen-Image-2512는 Qwen-Image의 텍스트 기반 이미지 생성 모델의 12월 업데이트로, 향상된 인간의 사실성, 더 세밀한 자연 디테일, 개선된 텍스트 렌더링을 특징으로 합니다." sidebarTitle: "Qwen-Image-2512" -translationSourceHash: dd214a11 +translationSourceHash: bdd1c703 translationFrom: tutorials/image/qwen/qwen-image-2512.mdx translationBlockHashes: "_intro": db22951d "Supported Aspect Ratios": 018121b0 - "Qwen-Image-2512 ComfyUI Native Workflow Example": 1221ba89 + "Qwen-Image-2512 ComfyUI Native Workflow Example": 50fbe75c --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image-2512**는 Qwen-Image의 텍스트 기반 이미지 생성 모델의 12월 업데이트입니다. 지난 8월에 출시된 기본 Qwen-Image 모델과 비교해 Qwen-Image-2512는 이미지 품질과 사실성이 크게 향상되었습니다. @@ -43,9 +44,10 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - - Comfy Cloud에서 실행하기 - + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "Qwen-Image-2512" 검색 + ### 1. 워크플로우 파일 @@ -55,28 +57,24 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 - **텍스트 기반 이미지 생성 (Qwen-Image 2512)**: 표준 50단계 생성 - **텍스트 기반 이미지 생성 (Qwen-Image 2512 4단계)**: Lightning LoRA를 사용한 가속화된 4단계 생성 - -

JSON 워크플로우 다운로드

- - ### 2. 모델 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (선택사항 - 4단계 Lightning 가속화용)** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **디퓨전 모델** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (대부분의 사용자에게 권장됨) -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (충분한 VRAM을 보유하고 더 높은 품질을 원하는 경우) +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors) (대부분의 사용자에게 권장됨) +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors) (충분한 VRAM을 보유하고 더 높은 품질을 원하는 경우) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx index 8f3152cb1..b8562daa8 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -2,7 +2,7 @@ title: "Qwen-Image-Edit-2511 ComfyUI 네이티브 워크플로우 예시" description: "Qwen-Image-Edit-2511은 Qwen-Image-Edit의 향상된 버전으로, 개선된 문자 일관성, 다인 편집 기능, 통합 LoRA 기능 및 강화된 기하학적 추론을 특징으로 합니다." sidebarTitle: "Qwen-Image-Edit-2511" -translationSourceHash: be665671 +translationSourceHash: f1e0ff2e translationFrom: tutorials/image/qwen/qwen-image-edit-2511.mdx --- @@ -33,31 +33,31 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그하여 불러올 수 있습니다. - -

JSON 워크플로우 다운로드

- + + Comfy Cloud에서 열기 + + Download JSON or search "Qwen-Image-Edit-2511" in Template Library + + - -

ComfyUI 클라우드에서 실행

- ### 2. 모델 다운로드 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA (선택사항 - 4단계 라이트닝 가속화용)** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **디퓨전 모델** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image-edit.mdx b/ko/tutorials/image/qwen/qwen-image-edit.mdx index 23f22ef51..211ddfc83 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit.mdx @@ -2,15 +2,16 @@ title: "Qwen-Image-Edit ComfyUI 네이티브 워크플로우 예시" description: "Qwen-Image-Edit는 Qwen-Image의 이미지 편집 버전으로, 20B 모델을 기반으로 추가로 학습되어 정밀한 텍스트 편집과 이중 세미틱/외관 편집 기능을 지원합니다." sidebarTitle: "Qwen-Image-Edit" -translationSourceHash: 9c641cf0 +translationSourceHash: 7c27546f translationFrom: tutorials/image/qwen/qwen-image-edit.mdx translationBlockHashes: "_intro": eee08e9a "ComfyOrg Qwen-Image-Edit Live Stream": 6e70a462 - "Qwen-Image-Edit ComfyUI Native Workflow Example": 7df4c813 + "Qwen-Image-Edit ComfyUI Native Workflow Example": 6703460f --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image-Edit**는 Qwen-Image의 이미지 편집 버전입니다. 20B Qwen-Image 모델을 기반으로 추가로 학습되었으며, Qwen-Image만의 독특한 텍스트 렌더링 기능을 편집 작업에 성공적으로 확장해 정밀한 텍스트 편집이 가능합니다. 또한 Qwen-Image-Edit는 입력 이미지를 Qwen2.5-VL(시각적 세미틱 제어용)과 VAE 인코더(시각적 외관 제어용)에 동시에 입력하여 이중 세미틱 및 외관 편집 기능을 구현합니다. @@ -47,13 +48,13 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. ![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

JSON 워크플로우 다운로드

- + + JSON 다운로드 또는 템플릿 라이브러리에서 "image_qwen_image_edit" 검색 + + Run this workflow on Cloud GPUs with zero setup + + - -

ComfyUI 클라우드에서 실행

- 아래 이미지를 입력으로 다운로드하세요 ![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -64,19 +65,19 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 **디퓨전 모델** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) 모델 저장 위치 diff --git a/ko/tutorials/image/qwen/qwen-image-layered.mdx b/ko/tutorials/image/qwen/qwen-image-layered.mdx index 51aae2791..1029408d2 100644 --- a/ko/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ko/tutorials/image/qwen/qwen-image-layered.mdx @@ -2,17 +2,18 @@ title: "Qwen-Image-Layered ComfyUI 워크플로우 예시" description: "Qwen-Image-Layered는 이미지를 여러 개의 RGBA 레이어로 분해할 수 있는 모델로, 레이어 분해를 통해 본질적인 편집 가능성을 제공합니다." sidebarTitle: "Qwen-Image-Layered" -translationSourceHash: bf2f2000 +translationSourceHash: 331a3de0 translationFrom: tutorials/image/qwen/qwen-image-layered.mdx translationBlockHashes: "_intro": 19900234 - "Qwen-Image-Layered workflow": ba937275 - "Model links": 98d12555 + "Qwen-Image-Layered workflow": 7a12a7ef + "Model links": 0635af99 "FP8 version": 6bffdd17 - "Workflow settings": b0f81aa2 + "Workflow settings": 098636f1 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image-Layered**는 알리바바의 Qwen팀에서 개발한 모델로, 이미지를 여러 개의 RGBA 레이어로 분해할 수 있습니다. 이 계층적 표현은 각 레이어가 독립적으로 조작 가능하도록 하여 다른 콘텐츠에 영향을 주지 않고도 각 레이어를 개별적으로 수정할 수 있게 합니다. @@ -30,13 +31,16 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 워크플로우 - -

JSON 워크플로우 파일 다운로드

- +| +| +| Download the JSON workflow file +| +| +| +| Run ComfyUI online with zero setup +| +| - -

ComfyUI 클라우드에서 실행하기

- @@ -44,15 +48,15 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **모델 저장 위치** diff --git a/ko/tutorials/image/qwen/qwen-image.mdx b/ko/tutorials/image/qwen/qwen-image.mdx index 1a4c7c333..040c81323 100644 --- a/ko/tutorials/image/qwen/qwen-image.mdx +++ b/ko/tutorials/image/qwen/qwen-image.mdx @@ -2,18 +2,19 @@ title: "Qwen-Image ComfyUI 네이티브 워크플로우 예시" description: "Qwen-Image는 Apache 2.0 라이선스로 오픈소스화된 20B 파라미터 MMDiT(멀티모달 디퓨전 트랜스포머) 모델입니다." sidebarTitle: "Qwen-Image" -translationSourceHash: f28191a9 +translationSourceHash: 4c682717 translationFrom: tutorials/image/qwen/qwen-image.mdx translationBlockHashes: "_intro": e096d7eb "ComfyOrg Qwen-Image live stream": 6436ac25 - "Qwen-Image Native Workflow Example": 73bab810 - "Qwen Image InstantX ControlNet Workflow": 741d3247 - "Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow": 45d40ad6 - "Qwen Image Union ControlNet LoRA Workflow": 333b4f67 + "Qwen-Image Native Workflow Example": bedb58a8 + "Qwen Image InstantX ControlNet Workflow": c5d39cf3 + "Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow": 6dc30273 + "Qwen Image Union ControlNet LoRA Workflow": a08d8e37 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image**는 알리바바의 Qwen 팀이 발표한 최초의 이미지 생성 기반 모델입니다. 이 모델은 Apache 2.0 라이선스로 오픈소스화된 20B 파라미터 MMDiT(멀티모달 디퓨전 트랜스포머) 모델입니다. 이 모델은 **복잡한 텍스트 렌더링**과 **정밀한 이미지 편집**에서 큰 진전을 이루었으며, 영어와 중국어를 포함한 여러 언어에 대해 고화질 출력을 달성했습니다. @@ -58,9 +59,12 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - - Comfy Cloud에서 실행하기 - + + + + + + 이 문서에 첨부된 워크플로우에는 세 가지 다른 모델이 사용됩니다: 1. Qwen-Image 원본 모델 fp8_e4m3fn @@ -82,15 +86,9 @@ GPU: RTX4090D 24GB ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. ![Qwen-image Text-to-Image 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

Qwen-Image 공식 모델용 워크플로우 다운로드

- - 증류 버전 - -

증류 모델용 워크플로우 다운로드

- - + + ### 2. 모델 다운로드 **ComfyUI에서 사용 가능한 모델** @@ -103,12 +101,12 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 **디퓨전 모델** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 증류 버전의 원저자는 cfg 1.0에서 15단계 사용을 권장합니다. @@ -117,15 +115,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +[qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **모델 저장 위치** @@ -164,19 +162,18 @@ Qwen_image_distill 이것은 ControlNet 모델이므로 일반 ControlNet처럼 사용할 수 있습니다. - - Comfy Cloud에서 실행하기 - + + + + + + ### 1. 워크플로우 및 입력 이미지 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

JSON 형식 워크플로우 다운로드

- - 아래 이미지를 입력으로 다운로드하세요. ![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -184,7 +181,7 @@ Qwen_image_distill 1. InstantX Controlnet -[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)를 다운로드해 `ComfyUI/models/controlnet/` 폴더에 저장하세요. +[Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors)를 다운로드해 `ComfyUI/models/controlnet/` 폴더에 저장하세요. 2. **Lotus Depth 모델** @@ -192,11 +189,11 @@ Qwen_image_distill **디퓨전 모델** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAE 모델** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) 또는 어떤 SD1.5 VAE도 가능합니다. +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) 또는 어떤 SD1.5 VAE도 가능합니다. ``` ComfyUI/ @@ -220,9 +217,12 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets 모델 패치 워크플로우 - - Comfy Cloud에서 실행하기 - + + + + + + 이 모델은 실제로 ControlNet이 아니라, 캐니, 딥스, 인페인트 등 세 가지 다른 제어 모드를 지원하는 모델 패치입니다. @@ -235,10 +235,6 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( 아래 이미지를 다운로드해 ComfyUI로 드래그해 해당 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

JSON 형식 워크플로우 다운로드

- - 아래 이미지를 입력으로 다운로드하세요: ![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/input.png) @@ -247,9 +243,9 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( 다른 모델은 Qwen-Image 기본 워크플로우와 동일합니다. 아래 모델만 다운로드해 `ComfyUI/models/model_patches` 폴더에 저장하면 됩니다. -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. 워크플로우 사용 지침 @@ -292,9 +288,12 @@ ControlNet 관련 워크플로우를 처음 사용한다면, 제어 이미지는 ## Qwen Image Union ControlNet LoRA 워크플로우 - - Comfy Cloud에서 실행하기 - + + + + + + 원본 모델 주소: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 캐니, 딥스, 포즈, 라인아트, 소프트엣지, 노말, 오픈포즈 지원 이미지 구조 제어 LoRA @@ -303,10 +302,6 @@ Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](htt 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

JSON 형식 워크플로우 다운로드

- - 아래 이미지를 입력으로 다운로드하세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/input.png) diff --git a/ko/tutorials/image/z-image/z-image-turbo.mdx b/ko/tutorials/image/z-image/z-image-turbo.mdx index f67002698..039474d00 100644 --- a/ko/tutorials/image/z-image/z-image-turbo.mdx +++ b/ko/tutorials/image/z-image/z-image-turbo.mdx @@ -2,15 +2,16 @@ title: "Z-Image-Turbo ComfyUI 워크플로우 예시" description: "Z-Image-Turbo는 1초 미만의 추론 지연 시간을 제공하는, 6B 파라미터의 경량화된 효율적 이미지 생성 모델입니다." sidebarTitle: "Z-Image-Turbo" -translationSourceHash: 6ef05349 +translationSourceHash: b9ae7e8c translationFrom: tutorials/image/z-image/z-image-turbo.mdx translationBlockHashes: "_intro": 062e5fc1 "Z-Image-Turbo text-to-image workflow": 5057b237 - "Z-Image-Turbo Fun Union ControlNet workflow": 18f13b79 + "Z-Image-Turbo Fun Union ControlNet workflow": 34191a15 --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -46,15 +47,15 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### Z-Image-Turbo 모델 다운로드 - + Z-Image-Turbo용 텍스트 인코더입니다. - + Z-Image-Turbo용 디퓨전 모델입니다. - + Z-Image-Turbo용 VAE입니다. @@ -81,7 +82,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ### ControlNet용 추가 모델 - + Z-Image-Turbo용 ControlNet 모델 패치입니다. diff --git a/ko/tutorials/partner-nodes/anthropic/claude.mdx b/ko/tutorials/partner-nodes/anthropic/claude.mdx index 8d0b3d242..7f4670052 100644 --- a/ko/tutorials/partner-nodes/anthropic/claude.mdx +++ b/ko/tutorials/partner-nodes/anthropic/claude.mdx @@ -2,7 +2,7 @@ title: "Anthropic Claude 파트너 노드 ComfyUI 공식 예제" description: "이 글에서는 ComfyUI에서 Anthropic Claude 파트너 노드를 사용해 대화 기능을 완성하는 방법을 소개합니다." sidebarTitle: "Anthropic Claude" -translationSourceHash: 155df30f +translationSourceHash: b0368d1f translationFrom: tutorials/partner-nodes/anthropic/claude.mdx --- @@ -30,13 +30,13 @@ Anthropic Claude는 강력한 추론 능력, 안전성 및 긴 컨텍스트 처 ## Anthropic Claude 챗 워크플로우 - - Comfy Cloud에서 열기 - +Anthropic Claude Chat workflow preview + + + + + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Anthropic Claude" 검색 - 해당 템플릿에서는 역할 프롬프트를 분석하고 생성하는 프롬프트를 구축했습니다. 이를 통해 사용자의 이미지를 해당 드로잉 프롬프트로 해석합니다. diff --git a/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 82fe7954e..0961835a0 100644 --- a/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -2,15 +2,16 @@ title: "Beeble SwitchX 파트너 노드 ComfyUI 공식 예제" description: "이 가이드에서는 ComfyUI에서 Beeble SwitchX 파트너 노드를 사용해 AI 기반 이미지 및 비디오 재조명과 환경 편집을 수행하는 방법을 설명합니다." sidebarTitle: "Beeble SwitchX" -translationSourceHash: 3724b3ae +translationSourceHash: 8f168afe translationFrom: tutorials/partner-nodes/beeble/beeble-switchx.mdx translationBlockHashes: "_intro": 470a081a - "Beeble SwitchX: Image Edit": a252a19f - "Beeble SwitchX: Video Edit": 97df1f25 + "Beeble SwitchX: Image Edit": 245e4f8a + "Beeble SwitchX: Video Edit": 166c2e59 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -27,13 +28,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; Beeble SwitchX 이미지 편집 워크플로우 - - Comfy Cloud에서 열기 - + + + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Beeble SwitchX: 이미지 편집" 검색 - + ### 작동 방식 @@ -67,13 +66,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; Beeble SwitchX 비디오 편집 워크플로우 - - Comfy Cloud에서 열기 - + + + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Beeble SwitchX: 비디오 편집" 검색 - + ### 작동 방식 diff --git a/ko/tutorials/partner-nodes/bria/background-removal.mdx b/ko/tutorials/partner-nodes/bria/background-removal.mdx index 464934b7a..0a9179494 100644 --- a/ko/tutorials/partner-nodes/bria/background-removal.mdx +++ b/ko/tutorials/partner-nodes/bria/background-removal.mdx @@ -2,15 +2,16 @@ title: "ComfyUI에서 Bria 배경 제거 사용하기" description: "ComfyUI에서 Bria 파트너 노드를 사용하여 이미지 및 동영상 배경 제거, 그린스크린, 배경 교체를 수행하는 방법을 알아보세요" sidebarTitle: "Bria 배경 제거" -translationSourceHash: c65b0474 +translationSourceHash: dbbb15e9 translationFrom: tutorials/partner-nodes/bria/background-removal.mdx translationBlockHashes: "_intro": f9000433 - "Image Background Removal": 14b7eb56 - "Video Background Processing": 3e621db5 + "Image Background Removal": 988ae876 + "Video Background Processing": aab02df5 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -27,7 +28,7 @@ Bria의 AI 서비스를 사용하여 모든 이미지에서 배경을 제거합 워크플로우 다운로드 - + Comfy Cloud에서 사용해보기 @@ -44,7 +45,7 @@ Bria의 동영상 배경 처리 노드를 사용하면 동영상 배경을 제 워크플로우 다운로드 - + Comfy Cloud에서 사용해보기 @@ -57,7 +58,7 @@ Bria의 동영상 배경 처리 노드를 사용하면 동영상 배경을 제 워크플로우 다운로드 - + Comfy Cloud에서 사용해보기 @@ -70,7 +71,7 @@ Bria의 동영상 배경 처리 노드를 사용하면 동영상 배경을 제 워크플로우 다운로드 - + Comfy Cloud에서 사용해보기 @@ -83,7 +84,7 @@ Bria의 동영상 배경 처리 노드를 사용하면 동영상 배경을 제 워크플로우 다운로드 - + Comfy Cloud에서 사용해보기 diff --git a/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index cf8c6d8f3..b69566f3a 100644 --- a/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -2,16 +2,17 @@ title: "ByteDance Seed Audio 1.0: 보편적인 오디오 생성" description: "음성, 음악, 음향 효과 및 다중 화자 대화를 단일 프롬프트로 생성하며, 음성 복제, 사전 설정 음성 및 캐릭터 기반 오디오를 Seed Audio 1.0과 ComfyUI에서 활용합니다." sidebarTitle: "Seed Audio 1.0" -translationSourceHash: 3fe23cff +translationSourceHash: 86bb1688 translationFrom: tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx translationBlockHashes: - "_intro": 4ce27313 - "Key capabilities": d5c85b17 + "_intro": 7d1fd38c + "Key capabilities": f5c9eb87 "Available workflows": 37e20124 - "How to use Seed Audio 1.0 in ComfyUI": 37aa925d + "How to use Seed Audio 1.0 in ComfyUI": 5630659b "Get started": 1ae5cf45 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index b12eb0c0f..363ee621f 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -2,17 +2,18 @@ title: "Seedance 2.0 리얼 휴먼 - 검증된 실제 사람 영상 생성" description: "ComfyUI에서 Identity 일관성, 네이티브 오디오-비디오 싱크, 한 번의 ByteDance 생체 인증을 통해 Seedance 2.0에서 실제 사람을 등장시키는 영상을 생성하세요." sidebarTitle: "Seedance 2.0 리얼 휴먼" -translationSourceHash: 82069c10 +translationSourceHash: 9ea8f507 translationFrom: tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx translationBlockHashes: "_intro": 4890d35b "Video guide": 416e3ec3 "What's different": b67103d8 "How verification works in ComfyUI": ee1e1056 - "Available workflows": bc3d7526 + "Available workflows": 16803a18 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -99,14 +100,12 @@ ComfyUI에는 두 가지 사전 구축된 Seedance 2.0 리얼 휴먼 템플릿 인증된 초상화(선택적 추가 참조 이미지, 비디오, 오디오 포함)를 사용해 일관된 실제 사람 정체성을 갖춘 Seedance 2.0 비디오를 생성하세요. - - Seedance 2.0 리얼 휴먼 참조-비디오 워크플로우 파일을 가져오세요. - - -### Seedance 2.0 리얼 휴먼 첫 번째-마지막 프레임-비디오 - -인증된 시작 프레임과 끝 프레임을 제공해 그 사이의 비디오를 생성하면서도 실제 사람의 정체성을 유지하세요. +Seedance 2.0 Real Human R2V workflow preview - - Seedance 2.0 리얼 휴먼 첫 번째-마지막 프레임-비디오 워크플로우 파일을 가져오세요. + + + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. + + + diff --git a/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index d3d0fea26..fd51b563a 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -2,17 +2,18 @@ title: "Seedance 2.0 - AI 동영상 생성" description: "ComfyUI에서 Seedance 2.0을 사용해 텍스트, 이미지, 동영상, 오디오를 통합된 입력으로 받아 고화질 동영상을 생성하세요. 오디오와 동영상이 싱크되며 캐릭터가 일관되고 영화 같은 카메라 제어를 제공합니다." sidebarTitle: "Seedance 2.0" -translationSourceHash: 4c9aec61 +translationSourceHash: 35902650 translationFrom: tutorials/partner-nodes/bytedance/seedance-2-0.mdx translationBlockHashes: "_intro": e1a3013d "Key capabilities": 6ca35ecf - "Available workflows": 4a9917b4 - "Seedance 2.0 Mini": c2cf0c21 + "Available workflows": d232c2a0 + "Seedance 2.0 Mini": db24d1c9 "Using real-person and AI-generated portraits in ComfyUI for Seedance 2.0": d9966fae --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -37,37 +38,37 @@ Seedance 2.0은 ByteDance의 차세대 다중 모달 동영상 생성 모델로, 텍스트 프롬프트를 통해 동영상을 생성하며, Seedance 2.0이 장면, 움직임, 속도를 처리합니다. - - Comfy Cloud에서 즉시 텍스트 → 동영상 워크플로우를 시도해 보세요. - +Seedance 2.0 Text-to-Video workflow preview - - 워크플로우 JSON을 다운로드하세요. - + + + + + ### 참조 → 동영상 (R2V) 참조 이미지, 동영상 또는 오디오를 사용해 룩, 움직임, 리듬을 유도하면서 결과물이 일관되도록 유지합니다. - - Comfy Cloud에서 즉시 참조 → 동영상 워크플로우를 시도해 보세요. - +Seedance 2.0 Reference-to-Video workflow preview - - 워크플로우 JSON을 다운로드하세요. - + + + + + ### 첫 번째와 마지막 프레임 → 동영상 (FLF2V) 시작 프레임과 끝 프레임을 제공하면 Seedance 2.0이 그 사이의 움직임과 전환을 생성합니다. - - Comfy Cloud에서 즉시 첫 번째와 마지막 프레임 → 동영상 워크플로우를 시도해 보세요. - +Seedance 2.0 FLF2V workflow preview - - 워크플로우 JSON을 다운로드하세요. - + + + + + ## Seedance 2.0 Mini diff --git a/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index 89ea7f653..e8fc92f6e 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -2,17 +2,18 @@ title: "ByteDance Seedream 5.0 Lite - 스마트 AI 이미지 생성" description: "ComfyUI에서 Seedream 5.0 lite를 사용해 웹 연결 검색과 향상된 지시사항 처리 기능을 활용한 이미지 생성하기" sidebarTitle: "Seedream 5.0 Lite" -translationSourceHash: ec3a4ae6 +translationSourceHash: 39b0d304 translationFrom: tutorials/partner-nodes/bytedance/seedream-5-lite.mdx translationBlockHashes: "_intro": 1991e54a "What's new in Seedream 5.0 lite": 02ac6a40 - "Seedream 5.0 lite image edit workflow": 44d7f3e4 - "Seedream 5.0 lite text to image workflow": 4fc4c666 + "Seedream 5.0 lite image edit workflow": e90cc958 + "Seedream 5.0 lite text to image workflow": f36785bf "Get started": d14874f4 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -30,13 +31,13 @@ Seedream 5.0 lite는 BytePlus의 최신 이미지 생성 모델입니다. Seedre ## Seedream 5.0 lite 이미지 편집 워크플로우 - - Comfy Cloud에서 열기 - +Seedream 5.0 Lite Image Edit workflow preview + + + + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Seedream 5.0 lite" 검색 - + ### 이미지 편집 사례 @@ -65,13 +66,13 @@ Seedream 5.0 lite는 BytePlus의 최신 이미지 생성 모델입니다. Seedre ## Seedream 5.0 lite 텍스트 → 이미지 워크플로우 - - Comfy Cloud에서 열기 - +Seedream 5.0 Lite Text-to-Image workflow preview + + + + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Seedream 5.0 lite" 검색 - + ### 세계 지식 사례 diff --git a/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index 558cc3667..23af4bd7a 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -2,15 +2,16 @@ title: "ByteDance Seedream 5.0 Pro - 전문적인 AI 이미지 생성" description: "Seedream 5.0 Pro를 사용하여 ComfyUI에서 향상된 명령어 이해, 정밀한 편집, 전문가 수준의 출력으로 고품질 이미지를 생성하세요" sidebarTitle: "Seedream 5.0 Pro" -translationSourceHash: 125a57d0 +translationSourceHash: c10432af translationFrom: tutorials/partner-nodes/bytedance/seedream-5-pro.mdx translationBlockHashes: "_intro": 0b4a40d7 - "What sets Seedream 5.0 Pro apart": ae41f702 - "Seedream 5.0 Pro text to image workflow": 39d05a61 - "Seedream 5.0 Pro image edit workflow": 7c657db8 + "What sets Seedream 5.0 Pro apart": e1702d4b + "Seedream 5.0 Pro text-to-image workflow": 8dbc58df + "Seedream 5.0 Pro image edit workflow": e2cabdb9 "Get started": 709ac7fd --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -35,28 +36,17 @@ Seedream 5.0 Pro는 ByteDance의 전문가급 이미지 생성 모델로, Seedre Comfy Cloud에서 열기 - - JSON 다운로드 또는 템플릿 라이브러리에서 "Seedream 5.0 Pro" 검색 - - -![Seedream 5.0 Pro 텍스트 기반 이미지 생성 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/Seedream5.0_Pro_T2I.png) - -## Seedream 5.0 Pro 이미지 편집 워크플로 - -기존 이미지를 텍스트 명령어로 편집: 객체 변경, 스타일 교체, 색상 조정, 원본 구조를 유지하면서 장면 재구성 - - - Comfy Cloud에서 열기 - - - - JSON 다운로드 또는 템플릿 라이브러리에서 "Seedream 5.0 Pro" 검색 + + Comfy Cloud에서 열기 + ![Seedream 5.0 Pro 이미지 편집 입력](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/input/interior_college.png) ![Seedream 5.0 Pro 이미지 편집 출력](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/Seedream5.0_Pro_image_edit.png) + + ## 시작하기 1. ComfyUI를 최신 버전(0.15.0+)으로 업데이트하세요 diff --git a/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx index 7de36dd13..10d5dae94 100644 --- a/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -2,16 +2,17 @@ title: "Gemini Omni Flash: 대화형 비디오 생성" description: "Gemini Omni Flash를 사용하여 자연어로 비디오를 생성하고 편집하세요. Google의 멀티모달 비디오 모델로, ComfyUI에서 파트너 노드를 통해 사용 가능합니다." sidebarTitle: "Gemini Omni Flash" -translationSourceHash: 2354813d +translationSourceHash: 6e260d1e translationFrom: tutorials/partner-nodes/google/gemini-omni-flash.mdx translationBlockHashes: "_intro": 3b6973dc "What Gemini Omni Flash offers": 215de783 - "Workflows": d363c40a + "Workflows": 149ee81a "Get started": 64517938 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -41,6 +42,8 @@ Gemini Omni Flash는 Google DeepMind의 고품질, 비용 효율적인 비디오 +![Gemini Omni Flash Text to Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_t2v-1.webp) + 자연어 프롬프트로 시네마틱 비디오를 생성합니다. 텍스트 설명을 세계 인식 모션, 조명 및 사운드가 포함된 비디오 출력으로 변환합니다. 소셜 미디어 콘텐츠 생성, 빠른 비디오 프로토타이핑 및 반복적인 시각적 스토리텔링에 이상적입니다. ### 이미지 기반 비디오 생성 @@ -60,6 +63,8 @@ Gemini Omni Flash는 Google DeepMind의 고품질, 비용 효율적인 비디오 +![Gemini Omni Flash Image to Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_i2v-1.webp) + Gemini Omni Flash를 사용하여 두 이미지로 비디오를 생성합니다. 자연어 프롬프트를 해석하여 지속 시간과 화면 비율을 제어합니다. 짧은 브랜드 클립, 다이나믹한 소셜 미디어 콘텐츠 제작 및 대화형 프롬프트를 통한 반복적인 비디오 편집에 적합합니다. ### 비디오 편집 @@ -76,6 +81,8 @@ Gemini Omni Flash를 사용하여 두 이미지로 비디오를 생성합니다. +![Gemini Omni Flash Video Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_video_edit-1.webp) + Gemini Omni Flash를 사용하여 자연어로 비디오를 편집합니다. 하나의 입력 비디오를 설명 지침에 따라 하나의 편집된 출력으로 변환합니다. 프롬프트에서 지속 시간과 화면 비율을 지정합니다. 빠른 소셜 미디어 리믹스, 시네마틱 장면 조정 및 반복적인 비디오 다듬기에 이상적입니다. ## 시작하기 diff --git a/ko/tutorials/partner-nodes/google/gemini.mdx b/ko/tutorials/partner-nodes/google/gemini.mdx index a3e14d274..bd905a5e4 100644 --- a/ko/tutorials/partner-nodes/google/gemini.mdx +++ b/ko/tutorials/partner-nodes/google/gemini.mdx @@ -2,7 +2,7 @@ title: "Google Gemini 파트너 노드 ComfyUI 공식 예제" description: "이 기사에서는 ComfyUI에서 Google Gemini 파트너 노드를 사용해 대화 기능을 완성하는 방법을 소개합니다." sidebarTitle: "Google Gemini" -translationSourceHash: ddd327ee +translationSourceHash: 8cd6027c translationFrom: tutorials/partner-nodes/google/gemini.mdx --- @@ -22,13 +22,10 @@ Google Gemini는 구글이 개발한 강력한 AI 모델로, 대화 및 텍스 아래 Json 파일을 다운로드한 후, ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. - -

Json 형식 워크플로우 파일 다운로드

- + + JSON 형식 워크플로우 파일 다운로드 + Comfy Cloud에서 열기 + ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/google/nano-banana-2.mdx b/ko/tutorials/partner-nodes/google/nano-banana-2.mdx index 68cec8b81..57b6d3e05 100644 --- a/ko/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/ko/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -2,17 +2,18 @@ title: "나노 바나나 2 - 빠른 AI 이미지 생성" description: "ComfyUI에서 나노 바나나 2를 사용해 프로 수준의 품질을 플래시 속도로 고화질 이미지로 생성하세요." sidebarTitle: "나노 바나나 2" -translationSourceHash: a146b8b0 +translationSourceHash: a86a55b2 translationFrom: tutorials/partner-nodes/google/nano-banana-2.mdx translationBlockHashes: "_intro": 379a03e0 "What's new in Nano Banana 2": 4048475a - "Nano Banana 2 image edit workflow": 085cc8cb + "Nano Banana 2 image edit workflow": ccade8b6 "Which model should you pick?": f16d5cc7 "Get started": f6189d9e --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -39,6 +40,12 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; JSON 다운로드 또는 템플릿 라이브러리에서 "나노 바나나 2" 검색 +![Nano Banana 2 workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_nano_banana2_image_edit-1.webp) + + + 이 워크플로의 예제 입력 이미지 가져오기 + + ### 프로 수준 품질 샘플 ![품질 비교](https://substack-post-media.s3.amazonaws.com/public/images/1f92ae2e-14d8-4a3b-9ed7-e57dacd2584f_1825x1696.png) diff --git a/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index 6080613d8..f17dbacce 100644 --- a/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -2,6 +2,8 @@ title: "Grok Imagine Video 1.5 이미지 투 비디오 ComfyUI 공식 예제" description: "이 가이드는 ComfyUI에서 Grok Imagine Video 1.5 Partner Node를 사용하여 이미지에서 기본 오디오가 포함된 고품질 비디오를 생성하는 방법을 설명합니다" sidebarTitle: "Grok Imagine Video 1.5" +translationSourceHash: 6a40c4e4 +translationFrom: tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx --- import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; @@ -32,7 +34,11 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; -### 워크플로우 개요 + + 이 워크플로의 예제 입력 이미지를 가져오세요. + + +### 워크플로 개요 이 워크플로우는 세 개의 노드를 사용합니다: - **LoadImage** — 시작 이미지 프레임 제공 diff --git a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index f8d9ef1df..c0f1a7a32 100644 --- a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -2,18 +2,19 @@ title: "ComfyUI에서 HappyHorse 1.0 동영상 생성" description: "ComfyUI의 파트너 노드를 통해 HappyHorse 1.0을 사용하여 이미지-동영상, 텍스트-동영상, 참조-동영상 및 시네마틱한 미학과 다중 샷 일관성을 갖춘 동영상 편집 방법을 배워보세요." sidebarTitle: "HappyHorse 1.0" -translationSourceHash: a0fbfb64 +translationSourceHash: 0b14a684 translationFrom: tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx translationBlockHashes: "_intro": 4eba8bbb "Highlights": 376295d6 - "HappyHorse 1.0 image-to-video": 85d1766b - "HappyHorse 1.0 text-to-video": 29f5941d - "HappyHorse 1.0 reference-to-video": fba3d9ea - "HappyHorse 1.0 video edit": 20a4559a + "HappyHorse 1.0 image-to-video": d41d0b77 + "HappyHorse 1.0 text-to-video": 6c594c34 + "HappyHorse 1.0 reference-to-video": b8ae994c + "HappyHorse 1.0 video edit": 5b531078 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -48,13 +49,13 @@ HappyHorse 1.0은 강력한 미학, 다중 샷 시퀀싱, 그리고 강력한 > - - HappyHorse 1.0 이미지-동영상 워크플로우 파일을 받으세요. - - - - Comfy Cloud에서 즉시 이미지-동영상 워크플로우를 체험해 보세요. + + + Try the Image-to-Video workflow instantly on Comfy Cloud. + + + ## HappyHorse 1.0 텍스트-동영상 @@ -66,34 +67,34 @@ HappyHorse 1.0은 강력한 미학, 다중 샷 시퀀싱, 그리고 강력한 src="https://github.com/Comfy-Org/example_workflows/raw/refs/heads/main/api_nodes/happy_horse/1.0/t2v_1.mp4" > - - HappyHorse 1.0 텍스트-동영상 워크플로우 파일을 받으세요. - - - - Comfy Cloud에서 즉시 텍스트-동영상 워크플로우를 체험해 보세요. + + + Try the Text-to-Video workflow instantly on Comfy Cloud. + + + ## HappyHorse 1.0 참조-동영상 참조 피사체를 활용해 동영상 생성을 유도하며, 영화 같은 다중 샷 시퀀스에서도 정체성을 유지합니다. - - HappyHorse 1.0 참조-동영상 워크플로우 파일을 받으세요. - - - - Comfy Cloud에서 즉시 참조-동영상 워크플로우를 체험해 보세요. + + + Try the Reference-to-Video workflow instantly on Comfy Cloud. + + + ## HappyHorse 1.0 동영상 편집 기존 영상을 변형하거나 피사체를 교체/삽입하면서 모션과 구성을 유지하는 V2V 및 SV2V 편집 워크플로우를 활용하세요. - - HappyHorse 1.0 동영상 편집 워크플로우 파일을 받으세요. - - - - Comfy Cloud에서 즉시 동영상 편집 워크플로우를 체험해 보세요. + + + Try the Video Edit workflow instantly on Comfy Cloud. + + + diff --git a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index 9f79306a8..ea0ec41a9 100644 --- a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -2,18 +2,19 @@ title: "ComfyUI에서 HappyHorse 1.1 비디오 생성" description: "ComfyUI에서 파트너 노드로 HappyHorse 1.1을 사용하여 이미지 기반 비디오 생성, 텍스트 기반 비디오 생성, 레퍼런스 기반 비디오 생성을 수행하고 네이티브 동기화 오디오와 시네마틱 멀티샷 스토리텔링을 활용하는 방법을 알아보세요." sidebarTitle: "HappyHorse 1.1" -translationSourceHash: 1c2775c3 +translationSourceHash: 2fadefbd translationFrom: tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx translationBlockHashes: "_intro": 6ae1c8e8 "Highlights": 10f03d9f - "HappyHorse 1.1 text-to-video": 31c81d0f - "HappyHorse 1.1 image-to-video": 2fa5bff9 - "HappyHorse 1.1 reference-to-video": 983e4818 + "HappyHorse 1.1 text-to-video": 3551ba66 + "HappyHorse 1.1 image-to-video": 05ad4488 + "HappyHorse 1.1 reference-to-video": 39830de6 "Getting started": 27bbb438 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -51,6 +52,8 @@ HappyHorse 1.1은 Alibaba의 프로덕션급 비디오 생성 모델의 최신 +![HappyHorse 1.1 Text-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_1_t2v-1.webp) + ## HappyHorse 1.1 이미지 기반 비디오 생성 정적 첫 프레임을 애니메이션화하세요. 이미지가 이미 스타일을 담고 있으므로 모션과 카메라 움직임을 설명합니다. HappyHorse 1.1이 오디오가 포함된 비디오를 반환합니다. @@ -67,6 +70,8 @@ HappyHorse 1.1은 Alibaba의 프로덕션급 비디오 생성 모델의 최신 +![HappyHorse 1.1 Image-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_1_i2v-1.webp) + ## HappyHorse 1.1 레퍼런스 기반 비디오 생성 멀티 캐릭터 무대극을 구성하세요. 캐릭터와 장면을 레퍼런스 이미지에 매핑한 다음, 타임스탬프가 찍힌 스토리보드와 캐릭터별 대사를 통해 연출합니다. 생성당 최대 9개의 레퍼런스 이미지를 지원하며, 캐릭터와 장면 레퍼런스가 분리되어 배경이 변경되어도 캐릭터가 일관되게 유지됩니다. @@ -86,6 +91,8 @@ HappyHorse 1.1은 Alibaba의 프로덕션급 비디오 생성 모델의 최신 +![HappyHorse 1.1 Reference-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_1_r2v-1.webp) + ## 시작하기 1. ComfyUI를 최신 버전으로 업데이트하세요. diff --git a/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 892a93f8c..ef6f12e16 100644 --- a/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -2,19 +2,20 @@ title: "Hunyuan 3D 파트너 노드 모델 생성 ComfyUI 공식 예제" description: "이 글에서는 ComfyUI에서 Hunyuan 3D 노드의 API를 사용해 3D 모델을 생성하는 방법을 소개합니다." sidebarTitle: "Hunyuan 3D 3.0" -translationSourceHash: 8d72daac +translationSourceHash: c92f6932 translationFrom: tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx translationBlockHashes: "_intro": 312a61a4 "Use cases": 60e6d268 "Getting started": 62f4530d - "Text-to-3D workflow": 710fb1e6 - "Image-to-3D workflow": d7a3b87d - "Multi-view-to-3D workflow": 68952799 - "Advanced features": acc7cc22 + "Text-to-3D workflow": 5edf82d4 + "Image-to-3D workflow": ce7efef1 + "Multi-view-to-3D workflow": ff251560 + "Advanced features": 6b37a964 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index c3704fa67..80807e285 100644 --- a/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -2,18 +2,19 @@ title: "Hunyuan 3D 파트너 노드 모델 생성 ComfyUI 공식 예제" description: "이 글에서는 ComfyUI에서 Hunyuan 3D 노드의 API를 사용해 3D 모델을 생성하는 방법을 소개합니다." sidebarTitle: "모델 생성" -translationSourceHash: cb7d1627 +translationSourceHash: aa9c5029 translationFrom: tutorials/partner-nodes/hunyuan3d/model-generation.mdx translationBlockHashes: "_intro": 312a61a4 "Use cases": 60e6d268 "Getting started": 62f4530d - "Text-to-3D workflow": 710fb1e6 - "Image-to-3D workflow": d7a3b87d - "Multi-view-to-3D workflow": 68952799 + "Text-to-3D workflow": 5edf82d4 + "Image-to-3D workflow": ce7efef1 + "Multi-view-to-3D workflow": ff251560 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 25cc9d5a7..827c26d0b 100644 --- a/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Ideogram 4.0 파트너 노드 튜토리얼" description: "ComfyUI에서 Ideogram 4.0 API 파트너 노드를 사용하는 방법을 알아보세요" sidebarTitle: "Ideogram 4.0" -translationSourceHash: 20fee685 +translationSourceHash: c4f854a7 translationFrom: tutorials/partner-nodes/ideogram/ideogram-v4.mdx translationBlockHashes: "_intro": 9a8e46c4 - "Ideogram 4.0 Partner Node Text-to-Image Workflow": ca6fe17b + "Ideogram 4.0 Partner Node Text-to-Image Workflow": ca41ad51 "Additional Notes": 5cc3cb89 "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -20,7 +21,9 @@ Ideogram 4.0은 Ideogram의 최신 텍스트 기반 이미지 생성 모델로, -## Ideogram 4.0 파트너 노드 텍스트 기반 이미지 생성 워크플로우 +## Ideogram 4.0 파트너 노드 텍스트-이미지 워크플로우 + +Ideogram 4.0 Text-to-Image workflow preview Comfy Cloud에서 열기 diff --git a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx index af7f71b93..44a16ee92 100644 --- a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -2,19 +2,20 @@ title: "Kling 2.6 모션 컨트롤 파트너 노드 ComfyUI 공식 예제" description: "ComfyUI에서 Kling 2.6 모션 컨트롤 파트너 노드를 사용해 참조 동영상에서 캐릭터 이미지로 정밀한 모션 전송하는 방법을 알아보세요" sidebarTitle: "Kling 2.6 모션 컨트롤" -translationSourceHash: 89919ffc +translationSourceHash: e449a655 translationFrom: tutorials/partner-nodes/kling/kling-motion-control.mdx translationBlockHashes: "_intro": fac1cb1e "Product highlights": a82b89a7 "Character orientation modes": aa2f7dec "Model tiers": 565d97ae - "Kling 2.6 Motion Control workflow": c75b1c17 + "Kling 2.6 Motion Control workflow": 33f2f69c "Input requirements": 0ecbf467 "Tips for better results": b6fdceb5 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -51,9 +52,29 @@ Kling 2.6 모션 컨트롤은 Kuaishou가 개발한 특수 다중모달 모델 ## Kling 2.6 모션 컨트롤 작업 흐름 - -

워크플로우 파일을 JSON 형식으로 다운로드하세요

- +Kling 2.6 Motion Control workflow preview + + + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. + + + + Download the workflow JSON file for local use. + + +
+Input materials + +Download these sample input files to try the workflow: + + + + Download sample reference image + + + Download sample motion reference video + + ## 입력 요구사항 diff --git a/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx b/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx index 8c3ed0c1f..8f5dd261e 100644 --- a/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -2,14 +2,14 @@ title: "Luma Uni-1 가이드" description: "ComfyUI에서 Luma Uni-1 파트너 노드를 사용해 Create 및 Modify 이미지 워크플로를 진행하는 방법을 안내합니다." sidebarTitle: "Luma Uni-1" -translationSourceHash: 16838ef3 +translationSourceHash: 99d39585 translationFrom: tutorials/partner-nodes/luma/luma-uni-1.mdx translationBlockHashes: "_intro": 544c5d43 "What makes Uni-1 different": c0181542 "Strengths": 4dd065dd "The core distinction: Create vs Modify": c3ea208f - "Available workflows": a6cf5659 + "Available workflows": 9b93a043 "Core parameters": b43ac975 "Working with reference images": d20d34ae "Prompting guidelines": ba9ac50d @@ -23,6 +23,7 @@ translationBlockHashes: --- + **ComfyUI**에서 **Luma Uni-1** 은 통합 이미지 작업을 위한 **파트너 노드**로 제공됩니다: **Create** 그래프는 프롬프트(선택적 레퍼런스 포함)로 새 이미지를 생성하고, **Modify** 그래프는 기존 이미지를 입력으로 받아 편집합니다. 두 모드 모두 평소 워크플로처럼 **Load Image** / **Save Image** 노드를 연결하고, Luma 노드에서 프롬프트, 시드, 종횡비, 레퍼런스 슬롯을 설정한 후 로컬에서 그래프를 실행하거나 **Comfy Cloud**에서 템플릿을 열면 됩니다. Luma는 Uni-1을 비확산, 디코더 전용 자기회귀 모델로 설명하며, 그리기 전에 프롬프트를 추론합니다. 캔버스에서 가장 중요한 것은 **Create vs Modify** 모드를 선택하고, 레퍼런스를 명확히 라벨링하고, 시드로 반복 작업하는 것입니다. @@ -74,26 +75,26 @@ Uni-1의 모든 것은 한 가지 질문으로 시작합니다: **새로운 것 ### Image Create 워크플로 - - Comfy Cloud에서 바로 Image Create 워크플로를 사용해 보세요. - +Luma Uni-1 Image Create workflow preview + + + + - - JSON을 다운로드하거나 템플릿 라이브러리에서 "Luma UNI-1 Image Create"를 검색하세요 - + ### Image Edit 워크플로 - - Comfy Cloud에서 바로 Image Edit 워크플로를 사용해 보세요. - +Luma Uni-1 Image Edit workflow preview - - JSON을 다운로드하거나 템플릿 라이브러리에서 "Luma UNI-1 Image Edit"를 검색하세요 - + + + 워크플로는 간단합니다: **프롬프트 → 평가 → 다듬기**. 탐색 단계에서는 시드를 비워 두세요. 마음에 드는 결과를 찾으면 시드를 고정하고 거기서부터 반복하세요. + + ## 핵심 파라미터 | 파라미터 | 설명 | diff --git a/ko/tutorials/partner-nodes/meshy/meshy-6.mdx b/ko/tutorials/partner-nodes/meshy/meshy-6.mdx index ee77cd605..f4c91fc0c 100644 --- a/ko/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/ko/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -2,10 +2,17 @@ title: "Meshy 6 파트너 노드 모델 생성 ComfyUI 공식 예제" description: "이 기사에서는 ComfyUI에서 Meshy 6 노드의 API를 사용해 3D 모델을 생성하는 방법을 소개합니다." sidebarTitle: "Meshy 6" -translationSourceHash: 302cd348 +translationSourceHash: f3036351 translationFrom: tutorials/partner-nodes/meshy/meshy-6.mdx +translationBlockHashes: + "_intro": 2dd46cd7 + "About Meshy 6": bc605135 + "Text-to-Model Workflow": f800f061 + "Image-to-Model Workflow": c148120d + "Multi-view to Model Workflow": 0cf6bb73 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -53,10 +60,20 @@ Meshy 6의 이미지 기반 3D 생성 기능을 활용해 2D 이미지를 세부 여러 뷰 이미지로부터 3D 모델을 생성해 더욱 정확한 지오메트리와 텍스처 재구성을 실현하세요. - - Comfy Cloud에서 즉시 다중 뷰 워크플로우를 실행하세요. + + Run the multi-view workflow instantly on Comfy Cloud. - - 로컬에서 사용할 수 있도록 워크플로우 JSON 파일을 다운로드하세요. - \ No newline at end of file +![Meshy 6 Multi-view to Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_multi_image_to_model-1.webp) + + + + Get the example front view image. + + + Get the example back view image. + + + Get the example side view image. + + diff --git a/ko/tutorials/partner-nodes/openai/chat.mdx b/ko/tutorials/partner-nodes/openai/chat.mdx index e431f97a4..e167ba5ba 100644 --- a/ko/tutorials/partner-nodes/openai/chat.mdx +++ b/ko/tutorials/partner-nodes/openai/chat.mdx @@ -2,7 +2,7 @@ title: "OpenAI 채팅 파트너 노드 ComfyUI 공식 예제" description: "이 글에서는 ComfyUI에서 OpenAI 채팅 파트너 노드를 사용하여 대화 기능을 완성하는 방법을 소개합니다." sidebarTitle: "OpenAI 채팅" -translationSourceHash: e4c0b73e +translationSourceHash: 77289d08 translationFrom: tutorials/partner-nodes/openai/chat.mdx --- @@ -22,13 +22,10 @@ OpenAI는 생성형 AI에 중점을 둔 회사로, 강력한 대화 기능을 아래 Json 파일을 다운로드한 후, ComfyUI로 드래그하여 해당 워크플로우를 불러오세요. - -

Json 형식 워크플로우 파일 다운로드

- + + Comfy Cloud에서 열기 + JSON 형식 워크플로우 파일을 다운로드하세요 + ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx b/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx index 01ead003b..e0f109c94 100644 --- a/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -3,18 +3,19 @@ title: "OpenAI GPT-Image-2 노드" description: "ComfyUI에서 이미지를 생성하는 데 OpenAI GPT-Image-2 파트너 노드를 사용하는 방법을 알아보세요" sidebarTitle: "GPT-Image-2" icon: "image" -translationSourceHash: b99cdc88 +translationSourceHash: ede8405e translationFrom: tutorials/partner-nodes/openai/gpt-image-2.mdx translationBlockHashes: "_intro": 9f80a1fc "Node Overview": 84ac2d88 "Getting Started": ab5a2231 - "Available workflows": 08afb96c + "Available workflows": d687898b "Key Capabilities": df71dcca "Hybrid Pipelines": 9c6479f0 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/pricing.mdx b/ko/tutorials/partner-nodes/pricing.mdx index 8ee4f7601..54a4667e1 100644 --- a/ko/tutorials/partner-nodes/pricing.mdx +++ b/ko/tutorials/partner-nodes/pricing.mdx @@ -3,7 +3,7 @@ title: "가격 정책" description: "이 문서에는 현재 파트너 노드의 가격 정책이 나와 있습니다. 모든 가격은 크레딧 단위로 표시됩니다(211 크레딧 = 1 USD)." sidebarTitle: "가격 정책" mode: wide -translationSourceHash: c864b96d +translationSourceHash: 907be779 translationFrom: tutorials/partner-nodes/pricing.mdx translationBlockHashes: "_intro": 845592d4 @@ -16,7 +16,9 @@ translationBlockHashes: "Magnific": 4000e7e1 "Google": 21dc3165 "HappyHorse": bceb584a - "Hitpaw": e83827b3 + "Hitpaw": 320c0537 + "Ideogram": 9410bf72 + "Krea": fb20c257 "Kling": cdf2c899 "Lightricks": 4870964f "Luma": 655207a5 @@ -52,6 +54,7 @@ translationBlockHashes: + 다음 표에는 현재 파트너 노드의 가격 정책이 나와 있습니다. 모든 가격은 크레딧 단위로 표시됩니다. 211 크레딧 = 1 USD diff --git a/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx b/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx index 4eb844625..32540aee1 100644 --- a/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -2,17 +2,18 @@ title: "ComfyUI에서 Recraft V4 이미지 및 벡터 생성" description: "ComfyUI에서 Recraft V4로 전문적인 이미지와 제작 준비가 완료된 벡터 생성하기" sidebarTitle: "Recraft V4" -translationSourceHash: fe0eaf6f +translationSourceHash: da066915 translationFrom: tutorials/partner-nodes/recraft/recraft-v4.mdx translationBlockHashes: "_intro": 8e346040 "What's new in V4": 6bdb9842 - "Recraft V4 text to image workflow": 7273fda3 - "Recraft V4 text to vector workflow": d322ba5f + "Recraft V4 text to image workflow": 2ff998b0 + "Recraft V4 text to vector workflow": c3255bfc "Additional notes": d5f3f109 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -41,6 +42,8 @@ Recraft V4는 전문 디자인 작업을 위해 개발된 새로운 이미지 JSON 다운로드 또는 템플릿 라이브러리에서 "Recraft V4 텍스트 기반 이미지 생성" 검색 +![Recraft V4 Text to Image workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_t2i-1.webp) + ### 워크플로우 실행 단계 1. (선택사항) `Recraft Style` 노드를 수정해 시각적 스타일을 조정하세요 @@ -72,6 +75,8 @@ Recraft V4는 직접 제작 준비가 완료된 SVG 벡터 출력을 생성할 JSON 다운로드 또는 템플릿 라이브러리에서 "Recraft V4 텍스트 기반 벡터 생성" 검색 +![Recraft V4 Text to Vector workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_text_to_vector-1.webp) + ### 벡터 쇼케이스 ![옥상 일러스트레이션](https://substack-post-media.s3.amazonaws.com/public/images/571c5214-5c7f-43f9-81c4-80cad1e5578b_1792x1024.png) diff --git a/ko/tutorials/partner-nodes/rodin/model-generation.mdx b/ko/tutorials/partner-nodes/rodin/model-generation.mdx index ad386d281..82121b3b4 100644 --- a/ko/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ko/tutorials/partner-nodes/rodin/model-generation.mdx @@ -2,16 +2,17 @@ title: "Rodin 파트너 노드 모델 생성 ComfyUI 공식 예제" description: "이 글에서는 ComfyUI에서 Rodin 노드의 API를 사용해 모델을 생성하는 방법을 소개합니다." sidebarTitle: "모델 생성" -translationSourceHash: 8d58740b +translationSourceHash: 738362ca translationFrom: tutorials/partner-nodes/rodin/model-generation.mdx translationBlockHashes: "_intro": a35f46e8 - "Single-view Model Generation Workflow": d982caac - "Multi-view Model Generation Workflow": 8a727575 + "Single-view Model Generation Workflow": d7303db4 + "Multi-view Model Generation Workflow": 5c5d3550 "Other Related Nodes": c885ce31 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -28,21 +29,32 @@ ComfyUI는 이제 해당 Rodin 모델 생성 API를 기본적으로 통합하여 ## 단일뷰 모델 생성 워크플로우 -### 1. 워크플로우 파일 다운로드 +Generate a 3D model from a single image input with Rodin. + +Rodin Single-view Model Generation workflow preview -아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. + + + Single-view Model Generation (Json Format) + + + Comfy Cloud에서 열기 + + - -

Json 형식 워크플로우 파일 다운로드

- +
+입력 자료 -입력 이미지로 아래 이미지를 다운로드하세요. +Download this sample input image to try the workflow: -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/doll.jpg) + + + 샘플 입력 이미지 다운로드 + + +
+ +![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/doll.jpg) ### 2. 워크플로우를 단계별로 완료하세요 @@ -60,25 +72,38 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', 해당 `Rodin 3D Generate - Regular Generate`는 최대 5개의 이미지 입력을 지원합니다. -### 1. 워크플로우 파일 다운로드 - 단일뷰 워크플로우를 다중뷰 워크플로우로 수정하거나, 아래 워크플로우 파일을 직접 다운로드할 수 있습니다. -아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - - -

Json 형식 워크플로우 파일 다운로드

- - -입력 이미지로 아래 이미지를 다운로드하세요. - -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/front.jpg) -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/back.jpg) -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/left.jpg) +Rodin Multi-view Model Generation workflow preview + + + + Multi-view Model Generation (Json Format) + + Comfy Cloud에서 열기 + + +
+입력 자료 + +Download these sample input images to try the workflow: + + + + 앞면 다운로드 + + + 뒷면 다운로드 + + + 왼쪽면 다운로드 + + +
+ +![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/front.jpg) +![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/back.jpg) +![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/rodin/multiview_to_model/left.jpg) ### 2. 워크플로우를 단계별로 완료하세요 diff --git a/ko/tutorials/partner-nodes/topaz/astra-2.mdx b/ko/tutorials/partner-nodes/topaz/astra-2.mdx index 79129ab92..51dbdc03b 100644 --- a/ko/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/ko/tutorials/partner-nodes/topaz/astra-2.mdx @@ -2,7 +2,7 @@ title: "Astra 2 - 창의적인 디퓨전 동영상 업스케일링" description: "ComfyUI에서 Astra 2가 동영상에 제공하는 기능과, 세부사항과 스타일을 제어하며 수행하는 창의적인 업스케일링에 대해 알아보세요." sidebarTitle: "Astra 2" -translationSourceHash: 2840b2ae +translationSourceHash: d28f921a translationFrom: tutorials/partner-nodes/topaz/astra-2.mdx --- @@ -38,6 +38,8 @@ Astra 1을 기반으로 하며, 세부사항과 스타일화가 어떻게 생성 **동영상 로드** → **Topaz Video Enhance** → **동영상 저장**. 템플릿을 열고 다른 사전 설정이 선택되어 있다면 업스케일러에서 **Astra 2**를 선택하세요. +Astra 2 workflow preview + Comfy Cloud에서 바로 워크플로우를 열어보세요. @@ -46,6 +48,18 @@ Astra 1을 기반으로 하며, 세부사항과 스타일화가 어떻게 생성 로컬 ComfyUI용 워크플로우 JSON을 다운로드하세요. +
+입력 자료 + +이 샘플 입력 동영상을 다운로드하여 워크플로를 시도해보세요: + + + + 샘플 입력 동영상 다운로드 + + +
+ ## 관련 항목 - [ComfyUI에서 동영상 업스케일링](/ko/tutorials/utility/video-upscale) — 업스케일링 방식 선택 가이드. \ No newline at end of file diff --git a/ko/tutorials/partner-nodes/tripo/model-generation.mdx b/ko/tutorials/partner-nodes/tripo/model-generation.mdx index f51ee9ef9..8c129071a 100644 --- a/ko/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ko/tutorials/partner-nodes/tripo/model-generation.mdx @@ -2,17 +2,18 @@ title: "Tripo 파트너 노드 모델 생성 ComfyUI 공식 예제" description: "이 글에서는 ComfyUI에서 Tripo 노드의 API를 사용해 모델을 생성하는 방법을 소개합니다." sidebarTitle: "모델 생성" -translationSourceHash: 91cfb63d +translationSourceHash: 9fa47d10 translationFrom: tutorials/partner-nodes/tripo/model-generation.mdx translationBlockHashes: "_intro": b329659d - "Text-to-Model Workflow": 1f82cb18 - "Image-to-Model Workflow": 20fac619 - "Multi-view Model Generation Workflow": b0e0a897 + "Text-to-Model Workflow": ae8e8384 + "Image-to-Model Workflow": 7463fca4 + "Multi-view Model Generation Workflow": c0a49c52 "Subsequent Task Processing for the Same Task": d24582e4 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -35,9 +36,14 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + Try the Text-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + ### 2. 워크플로우를 단계별로 완료하세요 @@ -58,9 +64,14 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- + + + Try the Image-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + 아래 이미지를 입력 이미지로 다운로드하세요. @@ -86,9 +97,18 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - -

Json 형식 워크플로우 파일 다운로드

- +Generate a 3D model from multiple view images for enhanced accuracy. + +Tripo Multi-view Model Generation workflow preview + + + + Try the Multiview-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + 아래 이미지를 입력 이미지로 다운로드하세요. diff --git a/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx index a49262239..38758feb5 100644 --- a/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -2,16 +2,17 @@ title: "Tripo 3.1: 고해상도 3D 에셋 생성 ComfyUI 공식 가이드" description: "파트너 노드를 통해 ComfyUI에서 Tripo 3.1을 사용해 고해상도 3D 에셋을 생성하는 방법을 알아보세요. 높은 밀도의 지오메트리와 제작용 PBR 준비 완료 재료를 제공합니다." sidebarTitle: "Tripo 3.1" -translationSourceHash: c520daa5 +translationSourceHash: dbbf32c8 translationFrom: tutorials/partner-nodes/tripo/tripo-3-1.mdx translationBlockHashes: "_intro": ff0ef375 "Capabilities": 6187e278 "Use Cases": 10ad3fe3 - "Available Workflows": c68e812a + "Available Workflows": 5ec1d9c5 --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -40,33 +41,39 @@ Tripo 3.1은 ComfyUI의 Tripo 파트너 노드를 통해 이용 가능한 최신 ### 텍스트 기반 모델 - - Comfy Cloud에서 텍스트 기반 모델 워크플로를 즉시 시험해 보세요. - +Generate a high-detail 3D model from a text prompt using Tripo 3.1. + +Tripo 3.1 Text-to-Model workflow preview - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Tripo 3.1 Text-to-Model"을 검색하세요. - + + + + + ### 이미지 기반 모델 - - Comfy Cloud에서 이미지 기반 모델 워크플로를 즉시 시험해 보세요. - +Generate a high-detail 3D model from an image input using Tripo 3.1. + +Tripo 3.1 Image-to-Model workflow preview - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Tripo 3.1 Image-to-Model"을 검색하세요. - + + + + + ### 멀티뷰 기반 모델 - - Comfy Cloud에서 멀티뷰 기반 모델 워크플로를 즉시 시험해 보세요. - +Generate a high-detail 3D model from multiple view images using Tripo 3.1. + +Tripo 3.1 Multiview-to-Model workflow preview + + + + - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Tripo 3.1 Multiview-to-Model"을 검색하세요. - + ### 버전 비교 diff --git a/ko/tutorials/partner-nodes/wan/wan2-7.mdx b/ko/tutorials/partner-nodes/wan/wan2-7.mdx index 65ad80d80..39ded22ed 100644 --- a/ko/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/ko/tutorials/partner-nodes/wan/wan2-7.mdx @@ -2,19 +2,20 @@ title: "ComfyUI에서 Wan2.7 동영상 생성" description: "ComfyUI의 파트너 노드를 통해 Wan2.7을 사용하여 이미지-동영상, 텍스트-동영상, 참조-동영상, 동영상 연속성 및 동영상 편집을 하는 방법을 알아보세요." sidebarTitle: "Wan2.7" -translationSourceHash: 3bb39265 +translationSourceHash: 42f5b589 translationFrom: tutorials/partner-nodes/wan/wan2-7.mdx translationBlockHashes: "_intro": 611e27ab "Key features": 6ca0965f "Highlights": cec936d1 - "Wan2.7 image-to-video": 29a49b02 - "Wan2.7 text-to-video": 25cfcacf - "Wan2.7 reference-to-video": cf70840e - "Wan2.7 video edit": ad198739 + "Wan2.7 image-to-video": af05fd5c + "Wan2.7 text-to-video": c9f23ab0 + "Wan2.7 reference-to-video": fe373e60 + "Wan2.7 video edit": 0d2511bb --- + import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -44,46 +45,46 @@ Wan2.7은 알리바바의 최신 동영상 생성 모델로, 현재 ComfyUI의 이미지 입력을 기반으로 동영상을 생성합니다. 첫 프레임, 첫 번째와 마지막 프레임, 오디오 기반 생성 모드를 지원합니다. - - Wan2.7 이미지-동영상 워크플로우 파일을 받으세요. - +Wan2.7 I2V workflow preview + + + + - - Comfy Cloud에서 즉시 이미지-동영상 워크플로우를 사용해 보세요. - + ## Wan2.7 텍스트-동영상 순수한 텍스트 프롬프트를 기반으로 동영상을 생성합니다. 선택적으로 오디오 입력과 다중 샷 내레이션을 추가해 더 풍부한 스토리텔링을 즐길 수 있습니다. - - Wan2.7 텍스트-동영상 워크플로우 파일을 받으세요. - +Wan2.7 T2V workflow preview + + + + - - Comfy Cloud에서 즉시 텍스트-동영상 워크플로우를 사용해 보세요. - + ## Wan2.7 참조-동영상 피사체의 시각적 모습을 나타낸 참조 이미지와 선택적 목소리 음색 참조를 사용하세요. 최대 5명의 실제 인물 입력을 지원해 다중 캐릭터 상호작용 장면을 구현할 수 있습니다. - - Wan2.7 참조-동영상 워크플로우 파일을 받으세요. - +Wan2.7 R2V workflow preview - - Comfy Cloud에서 즉시 참조-동영상 워크플로우를 사용해 보세요. - + + + + + ## Wan2.7 동영상 편집 텍스트 프롬프트, 참조 이미지 또는 스타일 전송을 사용해 기존 동영상을 편집하거나 복제하세요. - - Wan2.7 동영상 편집 워크플로우 파일을 받으세요. - +Wan2.7 Video Edit workflow preview + + + + - - Comfy Cloud에서 즉시 동영상 편집 워크플로우를 사용해 보세요. - + diff --git a/ko/tutorials/utility/moge.mdx b/ko/tutorials/utility/moge.mdx index af99d09df..2611f89f9 100644 --- a/ko/tutorials/utility/moge.mdx +++ b/ko/tutorials/utility/moge.mdx @@ -2,19 +2,20 @@ title: "ComfyUI MoGe 예제" description: "이 가이드에서는 ComfyUI에서 MoGe를 사용해 단안 기하학 추정, 즉 계량 점 지도, 깊이 지도, 법선 지도 및 메시 생성 방법을 보여줍니다." sidebarTitle: "MoGe" -translationSourceHash: 7ae1c247 +translationSourceHash: ff036dcc translationFrom: tutorials/utility/moge.mdx translationBlockHashes: "_intro": f3974fd4 "Model Installation": fb228472 "Example Workflows": cf2f84c1 - "1. Depth Estimation": 0c7bafb8 - "2. Perspective to Mesh": 03608c5b - "3. Panorama to Mesh": bdbb850d + "1. Depth Estimation": 030e1494 + "2. Perspective to Mesh": 4fde2470 + "3. Panorama to Mesh": b273460c "Community Resources": d9fcd8cd --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" # ComfyUI MoGe 소개 @@ -50,8 +51,8 @@ ComfyUI는 이제 MoGe 노드를 기본적으로 지원합니다. 시작하기 MoGe 체크포인트를 다운로드하고 해당 ComfyUI 폴더에 저장하세요: -- **MoGe-2 (권장)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1 (기준)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2 (권장)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1 (기준)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/bytedance/bernini-r.mdx b/ko/tutorials/video/bytedance/bernini-r.mdx index ba520d810..6c9c0b56b 100644 --- a/ko/tutorials/video/bytedance/bernini-r.mdx +++ b/ko/tutorials/video/bytedance/bernini-r.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Bernini-R 공식 예제" description: "ComfyUI에서 Bernini-R을 사용한 이미지 및 비디오 편집(재조명, 스타일 변환, 피사체 삽입 등)을 알아보세요." sidebarTitle: "Bernini-R" -translationSourceHash: 84ecde34 +translationSourceHash: e0b08f10 translationFrom: tutorials/video/bytedance/bernini-r.mdx translationBlockHashes: - "_intro": efe73774 - "Model Installation": 84eedf7b + "_intro": b791048b + "Model Installation": 286fbd69 "Example Workflows": cf2f84c1 - "1. Image Editing": cd75c4ca - "2. Video Editing": 9e90382c - "Community Resources": 57f652a5 + "1. Image Editing": 0c53cda8 + "2. Video Editing": ef878ac4 + "Community Resources": 9b4a0aaf --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' # ComfyUI Bernini-R 소개 @@ -49,16 +50,16 @@ ComfyUI는 이제 Bernini-R 노드를 기본 지원합니다. 시작하기 전 필요한 모델 가중치를 다운로드하여 해당 ComfyUI 폴더에 저장합니다: **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index ce6479416..b2a154998 100644 --- a/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/ko/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -2,7 +2,7 @@ title: "Cosmos Predict2 Video2World ComfyUI 공식 예제" description: "이 가이드는 ComfyUI에서 Cosmos-Predict2 Video2World 워크플로우를 완료하는 방법을 보여줍니다." sidebarTitle: "Cosmos-Predict2" -translationSourceHash: 7c20daa6 +translationSourceHash: 3386fb94 translationFrom: tutorials/video/cosmos/cosmos-predict2-video2world.mdx --- @@ -42,9 +42,10 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - -

Json 형식 워크플로우 파일 다운로드

- + + JSON 형식 워크플로우 파일을 다운로드하세요 + 사전 설치된 모델로 Comfy Cloud에서 이 워크플로우 실행 + 다음 이미지를 입력으로 다운로드하세요: @@ -56,17 +57,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **확산 모델** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) +- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) 기타 가중치는 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged)에서 다운로드하세요. **텍스트 인코더** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 414f7b0d2..7a75aa8d7 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -2,10 +2,18 @@ title: "HunyuanVideo 1.5" description: "소비자용 GPU에서 고품질 비디오 생성을 위한 경량 83억 파라미터 모델인 HunyuanVideo 1.5를 사용하는 방법을 알아보세요." sidebarTitle: "HunyuanVideo 1.5" -translationSourceHash: 0b15fd89 +translationSourceHash: 4379c5f7 translationFrom: tutorials/video/hunyuan/hunyuan-video-1-5.mdx +translationBlockHashes: + "_intro": 74d9d0ad + "Model highlights": a18b0a64 + "Common models for all workflows": 87181049 + "Hunyuan Video 1.5 Text-to-Video Workflow": 73a4afe7 + "Hunyuan Video 1.5 Image-to-Video Workflow": 42a0612d + "Super-resolution upscaler": 89338328 --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; [HunyuanVideo 1.5](https://github.com/Tencent/HunyuanVideo)는 텐센트의 Hunyuan팀이 개발한 경량 83억 파라미터 모델입니다. 이 모델은 소비자용 GPU(24GB VRAM)에서 플래그십 수준의 비디오 생성 성능을 제공하며, 품질을 저하시키지 않으면서 진입 장벽을 대폭 낮췄습니다. @@ -29,17 +37,17 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 모델 링크 **텍스트 인코더** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **디퓨전 모델** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **VAE** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) 모델 저장 위치 @@ -55,4 +63,4 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; │ │ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors │ └── :open_file_folder: vae/ │ └── hunyuanvideo15_vae_fp16.safetensors -``` \ No newline at end of file +``` diff --git a/ko/tutorials/video/hunyuan/hunyuan-video.mdx b/ko/tutorials/video/hunyuan/hunyuan-video.mdx index 16902e496..ad5f0cf7a 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video.mdx @@ -2,17 +2,18 @@ title: "ComfyUI 훈위안 비디오 예시" description: "이 가이드는 ComfyUI에서 훈위안 텍스트-투-비디오 및 이미지-투-비디오 워크플로우를 사용하는 방법을 보여줍니다." sidebarTitle: "훈위안 비디오" -translationSourceHash: 045f4b68 +translationSourceHash: 2cbcac4c translationFrom: tutorials/video/hunyuan/hunyuan-video.mdx translationBlockHashes: "_intro": 1af8ce94 - "Common Models for All Workflows": 3e4807f2 - "Hunyuan Text-to-Video Workflow": 8c3a1c81 - "Hunyuan Image-to-Video Workflow": ef169a18 - "Try it yourself": d4ee74b3 + "Common Models for All Workflows": 004bda25 + "Hunyuan Text-to-Video Workflow": fdc74d6f + "Hunyuan Image-to-Video Workflow": fef3b225 + "Try It Yourself": 076fb43f --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; - -

Json 워크플로우 파일 다운로드

- - 14B 버전을 사용하고 싶다면 모델 파일을 14B 버전으로 교체하면 되지만, VRAM 요구 사항에 유의해 주세요. -#### 1.2 입력 이미지 다운로드 - -아래 이미지를 다운로드해 시작 프레임으로 사용하세요: +### 2. 입력 자료 다운로드 -![입력 참조 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_1.3B_input.jpg) + + + 1.3B 워크플로우의 시작 프레임으로 사용할 이미지를 다운로드하세요. + + -### 2. 워크플로우 단계별 완료하기 +### 3. 워크플로우 단계별 완료하기 ![Wan2.1 Fun Camera 워크플로우 단계](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -111,7 +145,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 7. `WanCameraEmbedding` 노드에서 카메라 모션을 설정하세요. 8. `Run` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 생성을 실행하세요. -## ComfyUI Wan2.1 Fun Camera 14B 워크플로우 및 입력 이미지 +## ComfyUI Wan2.1 Fun Camera 14B 워크플로우 + +### 1. 워크플로우 다운로드 + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan 2.1 Fun Camera 14B" 검색 + + - -

Json 워크플로우 파일 다운로드

- +### 2. 입력 자료 다운로드 -**입력 이미지** -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) + + + 14B 워크플로우의 시작 프레임으로 사용할 이미지를 다운로드하세요. + + ## 성능 참고사항 diff --git a/ko/tutorials/video/wan/fun-control.mdx b/ko/tutorials/video/wan/fun-control.mdx index 9cb0cc3d9..b72c3d8bc 100644 --- a/ko/tutorials/video/wan/fun-control.mdx +++ b/ko/tutorials/video/wan/fun-control.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Wan2.1 Fun Control 동영상 예시" description: "이 가이드에서는 ComfyUI에서 Wan2.1 Fun Control을 사용해 제어 동영상을 생성하는 방법을 보여줍니다." sidebarTitle: "Wan2.1 Fun Control" -translationSourceHash: 88043153 +translationSourceHash: f6e4d92d translationFrom: tutorials/video/wan/fun-control.mdx translationBlockHashes: "_intro": 9efc0241 - "About Wan2.1-Fun-Control": 19f05e50 - "Model Installation": aee3a183 - "ComfyUI Native Workflow": 13eb7888 - "Workflow Using Custom Nodes": cf408ff0 - "Usage Tips": ae173b3b + "About Wan2.1-Fun-Control": 2e562079 + "Model Installation": ef08dc73 + "ComfyUI Native Workflow": d64a3fa4 + "Workflow Using Custom Nodes": ca1f733a + "Usage Tips": 9e74c201 --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## Wan2.1-Fun-Control 소개 @@ -52,18 +53,18 @@ ComfyUI는 현재 Wan2.1 Fun Control 모델을 **네이티브로 지원**합니 해당 링크를 클릭해 다운로드하세요. 이전에 Wan 관련 워크플로우를 사용한 적이 있다면 **Diffusion 모델**만 다운로드하면 됩니다. **Diffusion 모델** - 1.3B 또는 14B 중 선택하세요. 14B 버전은 파일 크기가 더 크고 (32GB), VRAM 요구 사항도 높습니다: -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-Control.safetensors`로 이름 변경 **텍스트 인코더** - 다음 모델 중 하나를 선택하세요 (fp16 정밀도는 파일 크기가 더 크고 성능 요구 사항도 높습니다): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/video/wan/fun-inp.mdx b/ko/tutorials/video/wan/fun-inp.mdx index dc9d11849..319f8544a 100644 --- a/ko/tutorials/video/wan/fun-inp.mdx +++ b/ko/tutorials/video/wan/fun-inp.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Wan2.1 Fun InP 동영상 예시" description: "이 가이드에서는 ComfyUI에서 Wan2.1 Fun InP를 사용해 첫 프레임과 마지막 프레임 제어 기능을 갖춘 동영상을 생성하는 방법을 보여줍니다." sidebarTitle: "Wan2.1 Fun InP" -translationSourceHash: db96fa7d +translationSourceHash: 866f2586 translationFrom: tutorials/video/wan/fun-inp.mdx translationBlockHashes: "_intro": 9efc0241 - "About Wan2.1-Fun-InP": ab94df56 - "Wan2.1 Fun InP Workflow": 33ad43c2 + "About Wan2.1-Fun-InP": f67d80ee + "Wan2.1 Fun InP Workflow": 140545b2 "Other Wan2.1 Fun InP or video-related custom node packages": 4c96c810 --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## Wan2.1-Fun-InP 소개 @@ -53,18 +54,18 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 다음 모델들은 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 및 [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334)에서 찾을 수 있습니다. **디퓨전 모델** - 1.3B 또는 14B를 선택하세요. 14B 버전은 파일 크기가 더 크고(32GB), VRAM 요구 사항도 높습니다: -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-InP.safetensors`로 이름 변경 +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true): 다운로드 후 `Wan2.1-Fun-14B-InP.safetensors`로 이름 변경 **텍스트 인코더** - 다음 모델 중 하나를 선택하세요 (fp16 정밀도는 크기가 더 크고 성능 요구 사항도 높습니다): -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` diff --git a/ko/tutorials/video/wan/vace.mdx b/ko/tutorials/video/wan/vace.mdx index 7069223b3..cd62a0ef9 100644 --- a/ko/tutorials/video/wan/vace.mdx +++ b/ko/tutorials/video/wan/vace.mdx @@ -2,18 +2,20 @@ title: "ComfyUI Wan2.1 VACE 비디오 예제" description: "이 기사에서는 ComfyUI에서 Wan VACE 비디오 생성 예제를 완료하는 방법을 소개합니다." sidebarTitle: "Wan2.1 VACE" -translationSourceHash: 9b04cff5 +translationSourceHash: 882bb348 translationFrom: tutorials/video/wan/vace.mdx translationBlockHashes: - "_intro": f6c2c850 - "About VACE": 4fa63090 - "Model Download and Loading in Workflows": 61bc1fa6 - "VACE Text-to-Video Workflow": 0275c575 - "VACE Image-to-Video Workflow": 6dbc84f2 - "VACE Video-to-Video Workflow": 0b58d1d2 - "VACE Video Outpainting Workflow": 22da2682 - "VACE First-Last Frame Video Generation": 0f2de66b + "_intro": 4e003b58 + "About VACE": 3d2b2490 + "Model Download and Loading in Workflows": 0a6b52f4 + "1. VACE Text-to-Video": 1715a0ab + "2. VACE Image-to-Video": 0ade4d39 + "3. VACE Video-to-Video": 4642a1ae + "4. VACE Inpainting": 6535794d + "5. VACE Video Outpainting": 24f1f2ff + "6. VACE First-Last Frame Video Generation": e58f2ea1 --- + import CancelBypass from '/snippets/ko/interface/cancel-bypass.mdx' import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -59,18 +61,18 @@ VACE 14B는 알리바바 통이 완샹팀이 출시한 오픈소스 통합 비 ### 모델 다운로드 **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 이전에 Wan Video 관련 워크플로를 사용하셨다면 이미 다음 모델 파일을 다운로드하셨을 것입니다. **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **텍스트 인코더** 중 하나를 선택해 다운로드하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/wan/wan-alpha.mdx b/ko/tutorials/video/wan/wan-alpha.mdx index c6a212e56..c2ffa8ec1 100644 --- a/ko/tutorials/video/wan/wan-alpha.mdx +++ b/ko/tutorials/video/wan/wan-alpha.mdx @@ -2,10 +2,15 @@ title: "Wan-Alpha 튜토리얼" description: "ComfyUI에서 Wan-Alpha를 사용해 알파 채널 투명도가 적용된 동영상을 생성하는 방법을 배워보세요." sidebarTitle: "Wan-Alpha" -translationSourceHash: 82295242 +translationSourceHash: 775fde1b translationFrom: tutorials/video/wan/wan-alpha.mdx +translationBlockHashes: + "_intro": a57abe06 + "Resources": afef8889 + "Wan-Alpha Text-to-Video Workflow (14B)": 32a5a973 --- + Wan-Alpha는 알파 채널 투명도가 적용된 고품질 동영상을 생성하는 특수한 텍스트 기반 비디오 생성 모델입니다. Wan2.1-14B-T2V 베이스 모델을 기반으로 하며, 투명한 배경과 반투명 객체를 포함한 동영상을 만들어 합성 작업에 완벽합니다. 이 모델은 투명한 배경, 반투명 객체(기포, 유리, 물), 빛나는 효과 및 정확한 알파 채널을 갖춘 세밀한 디테일(머리카락, 연기, 입자)을 잘 생성합니다. @@ -29,4 +34,4 @@ Wan-Alpha는 알파 채널 투명도가 적용된 고품질 동영상을 생성 - [Wan-Alpha GitHub](https://github.com/WeChatCV/Wan-Alpha) - [Hugging Face 모델](https://huggingface.co/htdong/Wan-Alpha) - [ComfyUI 버전](https://huggingface.co/htdong/Wan-Alpha_ComfyUI) -- [연구 논문](https://arxiv.org/pdf/2509.24979) \ No newline at end of file +- [연구 논문](https://arxiv.org/pdf/2509.24979) diff --git a/ko/tutorials/video/wan/wan-ati.mdx b/ko/tutorials/video/wan/wan-ati.mdx index 345de9b22..f5b9de21f 100644 --- a/ko/tutorials/video/wan/wan-ati.mdx +++ b/ko/tutorials/video/wan/wan-ati.mdx @@ -2,15 +2,16 @@ title: "Wan ATI ComfyUI 네이티브 워크플로우 튜토리얼" description: "트래젝터리 제어를 사용한 비디오 생성" sidebarTitle: "WAN2l1 ATI" -translationSourceHash: a8716511 +translationSourceHash: 39439781 translationFrom: tutorials/video/wan/wan-ati.mdx translationBlockHashes: - "_intro": 247f391d + "_intro": 52f7de29 "Key Features": 67ce28af - "WAN ATI Trajectory Control Workflow Example": 01ea7d04 + "WAN ATI Trajectory Control Workflow Example": f704f2c5 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' @@ -47,18 +48,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 워크플로우에서 모델 파일을 성공적으로 다운로드하지 못했다면, 아래 링크를 이용해 수동으로 다운로드해 보세요. **디퓨전 모델** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **텍스트 인코더** 다음 모델 중 하나를 선택하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) 파일 저장 위치 ``` diff --git a/ko/tutorials/video/wan/wan-causal-forcing.mdx b/ko/tutorials/video/wan/wan-causal-forcing.mdx index 7f90de227..8c0970a51 100644 --- a/ko/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ko/tutorials/video/wan/wan-causal-forcing.mdx @@ -2,8 +2,17 @@ title: "Causal Forcing I2V ComfyUI 워크플로우 예제" description: "Causal Forcing 또는 Causal Forcing++와 Wan2.1을 사용하여 이미지에서 비디오를 생성합니다 — 단 1회의 추론 단계로 부드럽고 시간적으로 일관된 비디오 구현" sidebarTitle: "Causal Forcing I2V" +translationSourceHash: 4f300f64 +translationFrom: tutorials/video/wan/wan-causal-forcing.mdx +translationBlockHashes: + "_intro": dbdb8d89 + "How it works": ea3eaaf0 + "Using the workflow": 5e6245a2 + "Steps to run": 3202924b + "Model downloads": c6ce079e --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Causal Forcing**은 추론 중 **반복적 조건화**를 적용하는 비디오 생성 기술로, 생성된 각 프레임이 모델에 다시 입력되어 다음 프레임을 예측합니다. 이를 통해 단일 시작 이미지에서 **1~4회의 추론 단계**만으로 부드럽고 시간적으로 일관된 비디오를 생성할 수 있습니다. @@ -83,10 +92,10 @@ Wan2.1 I2V 모델과 필요한 파일을 다운로드합니다. 해당 `models/` ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B 체크포인트 - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B 체크포인트 (최소 8GB VRAM) @@ -94,10 +103,10 @@ Wan2.1 I2V 모델과 필요한 파일을 다운로드합니다. 해당 `models/` ### CLIP 및 VAE - + google-bert/bert-base-uncased — CLIP 텍스트 인코더 - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/ko/tutorials/video/wan/wan-dancer.mdx b/ko/tutorials/video/wan/wan-dancer.mdx index 87022fb2d..1f134c5d7 100644 --- a/ko/tutorials/video/wan/wan-dancer.mdx +++ b/ko/tutorials/video/wan/wan-dancer.mdx @@ -2,16 +2,17 @@ title: "Wan Dancer: 음악에서 춤 비디오 생성" description: "Wan Dancer를 사용하여 음악에서 분 단위의 일관된 춤 비디오를 생성하세요. Wan Dancer는 Wan 2.2를 기반으로 구축된 계층적 오디오 기반 프레임워크입니다. 참조 이미지와 오디오를 입력하여 동기화된 춤 비디오를 생성합니다." sidebarTitle: "Wan Dancer" -translationSourceHash: ea7690e9 +translationSourceHash: 73c80f3f translationFrom: tutorials/video/wan/wan-dancer.mdx translationBlockHashes: "_intro": c890f065 - "Model Highlights": 06703fea + "Model Highlights": 0826f4bf "Workflow Overview": 0fa086f1 - "Wan Dancer Workflow": fbc54f05 - "Model Information": 97e02e88 + "Wan Dancer Workflow": 2251f138 + "Model Information": bc2b5961 "Report Issues": 271203b5 --- + import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" Wan Dancer는 Wan 2.2 아키텍처를 기반으로 구축된 오디오 기반 댄스 비디오 생성 모델입니다. 글로벌 및 로컬 전문가 모델을 포함하는 계층적 프레임워크를 사용하여 입력 음악에 동기화된 일관되고 표현력 있는 댄스 비디오를 생성합니다. @@ -61,20 +62,20 @@ ComfyUI를 최신 버전으로 업데이트한 후 워크플로 파일을 다운 ### 3. 모델 수동 다운로드 **확산 모델** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **텍스트 인코더** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan-flf.mdx b/ko/tutorials/video/wan/wan-flf.mdx index 825284aa2..76ae0d4d9 100644 --- a/ko/tutorials/video/wan/wan-flf.mdx +++ b/ko/tutorials/video/wan/wan-flf.mdx @@ -2,7 +2,7 @@ title: "ComfyUI Wan2.1 FLF2V 네이티브 예제" description: "이 가이드는 ComfyUI에서 Wan2.1 FLF2V 동영상 생성 예제를 완료하는 방법을 설명합니다." sidebarTitle: "Wan2.1 FLF2V" -translationSourceHash: 9b049266 +translationSourceHash: 5b7a25cb translationFrom: tutorials/video/wan/wan-flf.mdx --- @@ -58,7 +58,7 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 이 가이드에 포함된 모든 모델은 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)에서 확인할 수 있습니다. **diffusion_models** 하드웨어 환경에 따라 버전을 선택하세요. -- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16: [wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8: [wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -66,14 +66,14 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 **Text encoders**에서 버전을 하나 선택해 다운로드하세요. -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치 diff --git a/ko/tutorials/video/wan/wan-move.mdx b/ko/tutorials/video/wan/wan-move.mdx index b767247e3..d3630d0cb 100644 --- a/ko/tutorials/video/wan/wan-move.mdx +++ b/ko/tutorials/video/wan/wan-move.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan-Move 워크플로우 예시" description: "Wan-Move는 잠재적 궤적 안내를 통해 모션을 제어 가능한 비디오 생성 모델로, 이미지에서 비디오 생성 시 세밀한 포인트 단위의 모션 제어를 가능하게 합니다." sidebarTitle: "Wan-Move" -translationSourceHash: 1b11ad45 +translationSourceHash: dda86ac3 translationFrom: tutorials/video/wan/wan-move.mdx translationBlockHashes: "_intro": 67a26359 - "Wan-Move image-to-video workflow": ec83b0e4 - "Model links": 25e1a28d + "Wan-Move image-to-video workflow": b77082ab + "Model links": 8e0ef166 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Wan-Move**는 알리바바의 퉁이 랩에서 개발한 모션 제어가 가능한 비디오 생성 프레임워크입니다. 사용자는 입력 이미지 위에 점의 궤적을 지정함으로써 생성된 비디오에서 객체의 움직임을 제어할 수 있어 이미지에서 비디오 생성을 더욱 정밀하고 제어 가능하게 만듭니다. @@ -28,37 +29,48 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Wan-Move 이미지에서 비디오 워크플로우 - -

JSON 워크플로우 파일 다운로드

- - - -

ComfyUI 클라우드에서 실행하기

- - - - -## 모델 링크 +Preview the workflow output: -**텍스트 인코더** +![Wan-Move preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_wanmove_480p-1.webp) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan-Move Motion-Control" 검색 + + -**클립 비전** +### 입력 이미지 -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) + + + 기본 입력 이미지를 다운로드하거나, 자신의 이미지를 시작 프레임으로 사용하세요. + + -**로라** - -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) - -**디퓨전 모델** - -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) + -**VAE** +## 모델 링크 -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) + + + ComfyUI/models/diffusion_models/에 배치 + + + ComfyUI/models/loras/에 배치 + + + ComfyUI/models/text_encoders/에 배치 + + + ComfyUI/models/clip_vision/에 배치 + + + ComfyUI/models/vae/에 배치 + + **모델 저장 위치** diff --git a/ko/tutorials/video/wan/wan-video.mdx b/ko/tutorials/video/wan/wan-video.mdx index 657814f1b..127b4bcb1 100644 --- a/ko/tutorials/video/wan/wan-video.mdx +++ b/ko/tutorials/video/wan/wan-video.mdx @@ -1,19 +1,20 @@ --- -title: ComfyUI Wan2.1 동영상 예시 +title: "ComfyUI Wan2.1 동영상 예시" description: "이 가이드는 ComfyUI에서 Wan2.1 Video를 사용해 첫 프레임과 마지막 프레임을 포함한 동영상을 생성하는 방법을 보여줍니다." -sidebarTitle: Wan2.1 -translationSourceHash: e0c0f714 +sidebarTitle: "Wan2.1" +translationSourceHash: c11f515a translationFrom: tutorials/video/wan/wan-video.mdx translationBlockHashes: - "_intro": 228a7575 + "_intro": 376d341e "Wan2.1 ComfyUI Native Workflow Examples": b092012f - "Model Installation": 0bd3e4c2 - "Wan2.1 Text-to-Video Workflow": 611b9bf9 - "Wan2.1 Image-to-Video Workflow": bee5fc35 + "Model Installation": 21d2cbd6 + "Wan2.1 Text-to-Video Workflow (1.3B)": 268e4e8b + "Wan2.1 Image-to-Video Workflow (14B)": dfb19590 --- + Wan2.1 Video 시리즈는 2025년 2월 알리바바가 [Apache 2.0 라이선스](https://github.com/Wan-Video/Wan2.1?tab=Apache-2.0-1-ov-file)로 오픈소스로 공개한 동영상 생성 모델입니다. 두 가지 버전을 제공합니다: - 14B (140억 파라미터) @@ -41,14 +42,14 @@ Wan2.1 Video 시리즈는 2025년 2월 알리바바가 [Apache 2.0 라이선스] 이 가이드에서 언급된 모든 모델은 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)에서 확인할 수 있습니다. 아래는 이 가이드의 예시에 필요한 일반적인 모델들로, 미리 다운로드해 두시면 됩니다: **텍스트 인코더**에서 하나의 버전을 선택해 다운로드하세요: -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP 비전** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 파일 저장 위치: ``` @@ -70,7 +71,7 @@ ComfyUI/ ## Wan2.1 텍스트 투 비디오 워크플로우 -워크플로우를 시작하기 전에 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +워크플로우를 시작하기 전에 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. > 다른 t2v 정밀도 버전이 필요하시다면 [여기](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models)를 방문해 다운로드해 주세요. @@ -108,7 +109,7 @@ ComfyUI/ #### 2. 모델 다운로드 -[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. #### 3. 워크플로우 단계별 완료 @@ -136,7 +137,7 @@ ComfyUI/ #### 2. 모델 다운로드 -[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. +[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)를 다운로드해 `ComfyUI/models/diffusion_models/` 디렉토리에 저장해 주세요. #### 3. 워크플로우 단계별 완료 diff --git a/ko/tutorials/video/wan/wan2-2-animate.mdx b/ko/tutorials/video/wan/wan2-2-animate.mdx index 9155a436e..a4581fa4c 100644 --- a/ko/tutorials/video/wan/wan2-2-animate.mdx +++ b/ko/tutorials/video/wan/wan2-2-animate.mdx @@ -2,16 +2,16 @@ title: "Wan2.2 Animate ComfyUI 기본 워크플로우" description: "정밀한 모션과 표현을 정확히 재현하는 통합 캐릭터 애니메이션 및 교체 프레임워크." sidebarTitle: "Wan2.2 Animate" -translationSourceHash: 8d8cb435 +translationSourceHash: 07849135 translationFrom: tutorials/video/wan/wan2-2-animate.mdx translationBlockHashes: "_intro": 96eb0424 "Model Highlights": fc4e7349 "ComfyOrg Wan2.2 Animate stream replay": a43ba31b - "About Wan2.2 Animate workflow": f3713999 - "Wan2.2 Anmate ComfyUI native workflow(without custom nodes)": fc4b2e18 + "About Wan2.2 Animate workflow": ea4c1d9a --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 위한 통합 프레임워크입니다. @@ -55,13 +55,11 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 다음 워크플로우 파일을 다운로드해 ComfyUI로 끌어다 놓으면 워크플로우가 로드됩니다. - -

JSON 워크플로우 다운로드

- + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Animate" 검색 + - -

Comfy Cloud에서 실행

- 아래 자료를 입력으로 다운로드하세요: @@ -78,20 +76,20 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 ### 2. 모델 링크 **diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) Kijai의 리포지토리에서 가져온 모델입니다. -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 원본 모델 가중치 +- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) Kijai의 리포지토리에서 가져온 모델입니다. +- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 원본 모델 가중치 **clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 4단계 가속 Lora입니다. +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 4단계 가속 Lora입니다. **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-fun-camera.mdx b/ko/tutorials/video/wan/wan2-2-fun-camera.mdx index ae9e27c58..df4c404dd 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -2,7 +2,7 @@ title: "ComfyUI Wan2.2 Fun 카메라 제어: 동영상 생성 워크플로우 예시" description: "이 기사에서는 ComfyUI에서 Wan2.2 Fun 카메라 제어를 사용해 동영상 생성을 위한 카메라 제어 방법을 보여줍니다." sidebarTitle: "Wan2.2 Fun 카메라" -translationSourceHash: c2df0664 +translationSourceHash: d7f3bd26 translationFrom: tutorials/video/wan/wan2-2-fun-camera.mdx --- @@ -46,9 +46,12 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

JSON 워크플로우 다운로드

- +Wan2.2 Fun Camera Control workflow + + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Camera" 검색 + 아래 이미지를 다운로드해 입력으로 사용하세요. @@ -59,18 +62,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 다음 모델들은 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인할 수 있습니다: **디퓨전 모델** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA (선택사항, 가속화용)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) 파일 저장 위치 diff --git a/ko/tutorials/video/wan/wan2-2-fun-control.mdx b/ko/tutorials/video/wan/wan2-2-fun-control.mdx index b60dbd1cc..ac9c7507f 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-control.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan2.2 펀 컨트롤 비디오 생성 예시" description: "이 기사에서는 ComfyUI를 사용해 컨트롤 비디오를 활용한 Wan2.2 펀 컨트롤 비디오 생성 방법을 소개합니다." sidebarTitle: "Wan2.2 펀 컨트롤" -translationSourceHash: 59b97fdc +translationSourceHash: f1c35d6e translationFrom: tutorials/video/wan/wan2-2-fun-control.mdx translationBlockHashes: "_intro": 239e24b6 "ComfyOrg Wan2.2 Fun InP & Control Youtube Live Stream Replay": 22557a2e - "Wan2.2 Fun Control Video Generation Workflow Example": 68f53061 + "Wan2.2 Fun Control Video Generation Workflow Example": cbbb7456 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Wan2.2-Fun-Control**은 알리바바 PAI팀이 출시한 차세대 비디오 생성 및 제어 모델입니다. 혁신적인 Control Codes 메커니즘과 딥러닝, 다중 모달 조건 입력을 결합해 미리 설정된 제어 조건에 부합하는 고품질 비디오를 생성할 수 있습니다. 이 모델은 **Apache 2.0 라이선스**로 공개되며 상업적 사용도 지원합니다. @@ -63,9 +64,10 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/wan2.2_14B_fun_inp.mp4" > - -

JSON 워크플로우 다운로드

- + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Control" 검색 + 다음 이미지와 비디오를 입력 자료로 다운로드해 주세요. @@ -84,18 +86,18 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 아래 모델들은 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인하실 수 있습니다. **디퓨전 모델** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA (선택사항, 가속화용)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx index 6a076123b..4f035f807 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan2.2 Fun Inp 시작-끝 프레임 동영상 생성 예제" description: "이 기사는 ComfyUI를 사용해 Wan2.2 Fun Inp 시작-끝 프레임 동영상 생성 예제를 완성하는 방법을 소개합니다." sidebarTitle: "Wan2.2 Fun Inp" -translationSourceHash: 9d81688f +translationSourceHash: 714331ba translationFrom: tutorials/video/wan/wan2-2-fun-inp.mdx translationBlockHashes: - "_intro": 75061370 + "_intro": 7a3595d7 "ComfyOrg Wan2.2 Fun InP & Control Youtube Live Stream Replay": 22557a2e - "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 39f39063 + "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 692b42a8 --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Wan2.2-Fun-Inp**는 알리바바 PAI 팀이 출시한 시작-끝 프레임 제어 동영상 생성 모델입니다. 이 모델은 **시작 및 끝 프레임 이미지**를 입력해 중간 전환 동영상을 생성할 수 있어 크리에이터들이 더욱 창의적인 제어를 할 수 있도록 지원합니다. 이 모델은 **Apache 2.0 라이선스**로 배포되며 상업적 사용도 가능합니다. @@ -61,13 +62,11 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> 또는 ComfyUI를 최신 버전으로 업데이트한 후 아래 워크플로우를 다운로드해 ComfyUI에 드래그하여 로드하세요. - -

JSON 워크플로우 다운로드

- - -

Comfy Cloud에서 실행

- + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Inp" 검색 + Comfy Cloud에서 열기 + 다음 자료를 시작 및 끝 프레임으로 사용하세요. @@ -77,18 +76,18 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> ### 2. 모델 **디퓨전 모델** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) **Lightning LoRA (선택사항, 가속화용)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **텍스트 인코더** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan2-2-s2v.mdx b/ko/tutorials/video/wan/wan2-2-s2v.mdx index bd4747e65..2d7079bec 100644 --- a/ko/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ko/tutorials/video/wan/wan2-2-s2v.mdx @@ -1,8 +1,8 @@ --- -title: Wan2.2-S2V 오디오 기반 비디오 생성 ComfyUI 네이티브 워크플로우 예시 +title: "Wan2.2-S2V 오디오 기반 비디오 생성 ComfyUI 네이티브 워크플로우 예시" description: "이는 ComfyUI에서 Wan2.2-S2V 오디오 기반 비디오 생성을 위한 네이티브 워크플로우 예시입니다." sidebarTitle: "Wan2.2 S2V" -translationSourceHash: 8d1b0bf2 +translationSourceHash: 10604cda translationFrom: tutorials/video/wan/wan2-2-s2v.mdx --- @@ -35,38 +35,39 @@ Wan2.2 S2V 모델: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

JSON 워크플로우 다운로드

- + + Comfy Cloud에서 열기 + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 S2V" 검색 + - -

Comfy Cloud에서 실행

- 다음 이미지와 오디오를 입력으로 다운로드하세요: ![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - -

입력 오디오 다운로드

- + + 기본 입력 이미지를 다운로드하거나 자신의 이미지를 사용하세요 + + Download the default input audio, or use your own audio. + + ### 2. 모델 링크 모델들은 [우리 리포지토리](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged)에서 확인하실 수 있습니다. **diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) **audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) +- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -94,8 +95,8 @@ ComfyUI/ 두 모델 모두 [여기](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models)에서 확인하실 수 있습니다: -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) 이 템플릿에서는 `wan2.2_s2v_14B_fp8_scaled.safetensors`를 사용하며, 이 모델은 더 적은 VRAM을 필요로 합니다. 하지만 품질 저하를 줄이기 위해 `wan2.2_s2v_14B_bf16.safetensors`를 시도해볼 수도 있습니다. diff --git a/ko/tutorials/video/wan/wan2_2.mdx b/ko/tutorials/video/wan/wan2_2.mdx index 1303255f4..9ad610bac 100644 --- a/ko/tutorials/video/wan/wan2_2.mdx +++ b/ko/tutorials/video/wan/wan2_2.mdx @@ -1,20 +1,21 @@ --- title: "Wan2.2 비디오 생성 ComfyUI 공식 네이티브 워크플로 예시" description: "ComfyUI에서 Alibaba Cloud Tongyi Wanxiang 2.2 비디오 생성 모델의 공식 사용 가이드" -sidebarTitle: Wan2.2 -translationSourceHash: 70e49209 +sidebarTitle: "Wan2.2" +translationSourceHash: a75c822c translationFrom: tutorials/video/wan/wan2_2.mdx translationBlockHashes: "_intro": 3d02f6d3 "Model Highlights": fc7121c1 "Wan2.2 Open Source Model Versions": 7ed2d913 "ComfyOrg Wan2.2 Live Streams": d380a603 - "Wan2.2 TI2V 5B Hybrid Version Workflow Example": b0e108bd - "Wan2.2 14B T2V Text-to-Video Workflow Example": 9eaf9107 - "Wan2.2 14B I2V Image-to-Video Workflow Example": a006f8aa - "Wan2.2 14B FLF2V Workflow Example": 525e0946 + "Wan2.2 TI2V 5B Hybrid Version Workflow Example": a9d53e08 + "Wan2.2 14B T2V Text-to-Video Workflow Example": 6a6398c7 + "Wan2.2 14B I2V Image-to-Video Workflow Example": 356d6b7a + "Wan2.2 14B FLF2V Workflow Example": ca3b4239 "Community Resources": 7463b48b --- + import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' @@ -45,19 +46,19 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 ## Flux.2 Klein 4B 模型下载 - + 4B 模型文本编码器。 - + 扩散模型（4B Base）。 - + 扩散模型（4B 蒸馏版）。 - + 4B 模型 VAE。 @@ -103,11 +104,11 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 扩散模型（9B 蒸馏版）。 - + 9B 模型文本编码器。 - + 9B 模型 VAE。 diff --git a/zh/tutorials/flux/flux1-krea-dev.mdx b/zh/tutorials/flux/flux1-krea-dev.mdx index 23f2514c3..d85fe90ee 100644 --- a/zh/tutorials/flux/flux1-krea-dev.mdx +++ b/zh/tutorials/flux/flux1-krea-dev.mdx @@ -2,7 +2,7 @@ title: "Flux.1 Krea Dev ComfyUI 工作流教程" description: "Black Forest Labs 与 Krea 合作开发的最佳开源 FLUX 模型，专注于独特美学风格和自然细节，避免 AI 感，提供卓越的真实感和图像质量。" sidebarTitle: "Flux.1 Krea Dev" -translationSourceHash: e4ac73ab +translationSourceHash: 202e24cf translationFrom: tutorials/flux/flux1-krea-dev.mdx --- @@ -31,13 +31,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下载下面的图片或JSON，并拖入 ComfyUI 以加载对应工作流 ![Flux Krea Dev 工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - -

下载 JSON 格式工作流

- + + + 在 Comfy Cloud 上运行此工作流 + + + Download JSON or search "Flux.1 Krea Dev" in Template Library + + - -

在 Comfy Cloud 上运行

- #### 2. 模型链接 @@ -47,7 +49,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面这个版本是原始权重，如果你追求更高质量有足够的显存，可以尝试这个版本 -- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/resolve/main/flux1-krea-dev.safetensors) +- [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) - `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 的协议后才能使用浏览器进行下载。 @@ -55,12 +57,12 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 如果你使用过 Flux 相关的工作流，下面的模型是相同的，不需要重复下载 **Text encoders** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM **VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) 文件保存位置： ``` diff --git a/zh/tutorials/image/hidream/hidream-e1.mdx b/zh/tutorials/image/hidream/hidream-e1.mdx index 8a9fd0afa..924d4fe4b 100644 --- a/zh/tutorials/image/hidream/hidream-e1.mdx +++ b/zh/tutorials/image/hidream/hidream-e1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI 原生版本 HiDream-E1, E1.1 工作流示例" sidebarTitle: "HiDream-e1" description: "本篇将引导了解并完成 ComfyUI 原生版本 HiDream-I1 文生图工作流实例" -translationSourceHash: 76ea9fff +translationSourceHash: 7554c00c translationFrom: tutorials/image/hidream/hidream-e1.mdx translationBlockHashes: "_intro": 6829ae90 - "HiDream E1 and E1.1 Workflow Related Models": e648e936 - "HiDream E1.1 ComfyUI Native Workflow Example": 4134dfe8 - "HiDream E1 ComfyUI Native Workflow Example": 778b966a + "HiDream E1 and E1.1 Workflow Related Models": 25f9249b + "HiDream E1.1 ComfyUI Native Workflow Example": 4437f406 + "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ![HiDream-E1 演示](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -40,8 +41,8 @@ HiDream-E1 是智象未来(HiDream-ai) 正式开源的交互式图像编辑大 **Diffusion Model** 你不用同时下载这两个模型，由于 E1.1 是基于 E1 的迭代版本，在实际测试中它的质量和效果较 E1 都有较大提升 -- [hidream_e1_1_bf16.safetensors(推荐)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB -- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB +- [hidream_e1_1_bf16.safetensors(推荐)](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_1_bf16.safetensors) 34.2GB +- [hidream_e1_full_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_e1_full_bf16.safetensors) 34.2GB **Text Encoder**： @@ -115,9 +116,14 @@ E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一 ## HiDream E1 ComfyUI 原生 工作流示例 - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "HiDream E1.1" in Template Library + + E1 是于 2025 年 4 月 28 日发布的，这个模型只支持 768*768 的分辨率 diff --git a/zh/tutorials/image/hidream/hidream-i1.mdx b/zh/tutorials/image/hidream/hidream-i1.mdx index 040270b12..7e1487336 100644 --- a/zh/tutorials/image/hidream/hidream-i1.mdx +++ b/zh/tutorials/image/hidream/hidream-i1.mdx @@ -2,17 +2,18 @@ title: "ComfyUI 原生版本 HiDream-I1 文生图工作流示例" sidebarTitle: "HiDream-I1" description: "本篇将引导了解并完成 ComfyUI 原生版本 HiDream-I1 文生图工作流实例" -translationSourceHash: 590b8d6d +translationSourceHash: 73e39f2a translationFrom: tutorials/image/hidream/hidream-i1.mdx translationBlockHashes: "_intro": 59935bb2 "Model Features": 3ec455b0 "About This Workflow Example": 90c5d2b3 - "HiDream-I1 Workflow": 93e17294 + "HiDream-I1 Workflow": 5956f05c "Other Related Resources": e794d9ef --- + ![HiDream-I1 演示](https://raw.githubusercontent.com/HiDream-ai/HiDream-I1/main/assets/demo.jpg) HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图模型。该模型拥有17B参数规模，采用 [MIT 许可证](https://github.com/HiDream-ai/HiDream-I1/blob/main/LICENSE) 发布，支持用于个人项目、科学研究以及商用，目前在多项基准测试中该模型表现优异。 @@ -97,15 +98,20 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 full 版本工作流 - -

Run on Comfy Cloud

- + + + Run this workflow on Comfy Cloud with zero setup + + + 下载工作流 JSON 文件 + + #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 @@ -134,15 +140,20 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 dev 版本工作流 - -

Run on Comfy Cloud

- + + + Run this workflow on Comfy Cloud with zero setup + + + 下载工作流 JSON 文件 + + #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 @@ -170,15 +181,24 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 fast 版本工作流 - -

Run on Comfy Cloud

- + + + + Run this workflow on Comfy Cloud with zero setup + + + 下载工作流 JSON 文件 + + +- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM + #### 1. 模型文件下载 请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 -- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 +- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 #### 2. 工作流文件下载 请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 diff --git a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 969ac372c..d1a8a5156 100644 --- a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI NewBie-image-Exp0.1 工作流示例" description: "NewBie-image-Exp0.1 是一个基于 Next-DiT 架构的 35 亿参数动漫风格文生图模型，针对高质量动漫图像生成进行了优化，支持 XML 结构化提示词。" sidebarTitle: "NewBie-image-Exp0.1" -translationSourceHash: f56facc0 +translationSourceHash: 22db4899 translationFrom: tutorials/image/newbie-image/newbie-image-exp-0-1.mdx translationBlockHashes: "_intro": 7f6f1284 - "NewBie-image text-to-image workflow": af50f161 + "NewBie-image text-to-image workflow": e7fbe7f5 "Model links": cd75bb7e "Prompt format": 5b819c6c --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **NewBie-image-Exp0.1** 是由 NewBieAI Lab 开发的 35 亿参数 DiT 模型，专为动漫风格文生图设计。基于 Next-DiT 架构构建，能够生成细节丰富、视觉效果出色的动漫图像。 @@ -30,13 +31,17 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## NewBie-image 文生图工作流 - -

下载 JSON 工作流文件

- + + + + Download JSON or search "NewBie-image" in Template Library + + + Open in cloud + + + - -

在 ComfyUI Cloud 上运行

- @@ -44,16 +49,16 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **模型存放位置** diff --git a/zh/tutorials/image/omnigen/omnigen2.mdx b/zh/tutorials/image/omnigen/omnigen2.mdx index eb09cac91..f80b90415 100644 --- a/zh/tutorials/image/omnigen/omnigen2.mdx +++ b/zh/tutorials/image/omnigen/omnigen2.mdx @@ -2,17 +2,18 @@ title: "ComfyUI OmniGen2 原生工作流示例" description: "ComfyUI OmniGen2 原生工作流示例 - 统一的文生图、图像编辑和多图像合成模型。" sidebarTitle: "OmniGen2" -translationSourceHash: b061ff8c +translationSourceHash: 1429ffa8 translationFrom: tutorials/image/omnigen/omnigen2.mdx translationBlockHashes: "_intro": 3e9c5a21 "About OmniGen2": 2e6118e0 "OmniGen2 Model Download": 7997f8fe - "ComfyUI OmniGen2 Text-to-Image Workflow": f7650507 - "ComfyUI OmniGen2 Image Editing Workflow": 1859e94c + "ComfyUI OmniGen2 Text-to-Image Workflow": 5f9f63be + "ComfyUI OmniGen2 Image Editing Workflow": 1e06072e --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 OmniGen2 @@ -42,13 +43,13 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 由于本文涉及不同工作流，对应的模型文件及安装位置如下，对应工作流中也已包含了模型文件下载信息： **Diffusion Models）** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **Text Encoders）** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) 文件保存位置： @@ -67,9 +68,11 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 ### 1. 工作流文件下载 - -

在 Comfy Cloud 上运行

- + + + Open and run this workflow directly in Comfy Cloud. + + ![文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) @@ -95,9 +98,11 @@ OmniGen2 有丰富的图像编辑能力，并且支持为图像添加文本 ### 1. 工作流文件下载 - -

在 Comfy Cloud 上运行

- + + + Open and run this workflow directly in Comfy Cloud. + + ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) 下载下面的图片，我们将使用它作为输入图片。 diff --git a/zh/tutorials/image/ovis/ovis-image.mdx b/zh/tutorials/image/ovis/ovis-image.mdx index b28c481ee..9d4d06823 100644 --- a/zh/tutorials/image/ovis/ovis-image.mdx +++ b/zh/tutorials/image/ovis/ovis-image.mdx @@ -2,7 +2,7 @@ title: "Ovis-Image ComfyUI 工作流示例" description: "Ovis-Image 是一个 7B 文生图模型，专门针对高质量文本渲染进行优化，旨在严格的计算约束下高效运行。" sidebarTitle: "Ovis-Image" -translationSourceHash: 64c10a35 +translationSourceHash: 8d005c1d translationFrom: tutorials/image/ovis/ovis-image.mdx --- @@ -22,13 +22,17 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Ovis-Image 文生图工作流 - -

下载 JSON 工作流文件

- + + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Ovis image" in Template Library + + + - -

在 ComfyUI Cloud 上运行

- @@ -36,15 +40,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders（文本编码器）** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models（扩散模型）** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/pixeldit/pixeldit.mdx b/zh/tutorials/image/pixeldit/pixeldit.mdx index 9daacab20..18f597d59 100644 --- a/zh/tutorials/image/pixeldit/pixeldit.mdx +++ b/zh/tutorials/image/pixeldit/pixeldit.mdx @@ -2,15 +2,16 @@ title: "PixelDiT ComfyUI 工作流示例" description: "PixelDiT 是 NVIDIA 的像素空间扩散变换器，用于生成 1024px 文本到图像。它直接在像素空间生成图像，无需 VAE 编解码。" sidebarTitle: "PixelDiT" -translationSourceHash: d5838b66 +translationSourceHash: b3280f94 translationFrom: tutorials/image/pixeldit/pixeldit.mdx translationBlockHashes: "_intro": ef3a0250 - "PixelDiT text-to-image workflow": 21b069a6 + "PixelDiT text-to-image workflow": 1879b3e8 "Model downloads": e4bafb0a --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **PixelDiT** 是 NVIDIA 开发的像素空间扩散变换器，用于 **1024px** 文本到图像生成。与传统在潜空间操作的扩散模型不同，PixelDiT 使用双层 DiT 架构直接在像素空间生成图像——结合了 patch 级 DiT 和像素级 DiT，并通过 MM-DiT 融合实现文本和图像 token 之间的联合注意力。 @@ -62,11 +63,11 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' PixelDiT 使用两个模型文件：文本编码器和扩散模型。 - + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 文本编码器 - + pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 扩散模型 diff --git a/zh/tutorials/image/qwen/qwen-image-2512.mdx b/zh/tutorials/image/qwen/qwen-image-2512.mdx index c7e9877ee..1e4b05848 100644 --- a/zh/tutorials/image/qwen/qwen-image-2512.mdx +++ b/zh/tutorials/image/qwen/qwen-image-2512.mdx @@ -2,15 +2,16 @@ title: "Qwen-Image-2512 ComfyUI 原生工作流示例" description: "Qwen-Image-2512 是 Qwen-Image 文生图基础模型的 12 月更新版本，具有增强的人物真实感、更精细的自然细节和改进的文字渲染能力。" sidebarTitle: "Qwen-Image-2512" -translationSourceHash: dd214a11 +translationSourceHash: bdd1c703 translationFrom: tutorials/image/qwen/qwen-image-2512.mdx translationBlockHashes: "_intro": db22951d "Supported Aspect Ratios": 018121b0 - "Qwen-Image-2512 ComfyUI Native Workflow Example": 1221ba89 + "Qwen-Image-2512 ComfyUI Native Workflow Example": 50fbe75c --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Qwen-Image-2512** 是 Qwen-Image 文生图基础模型的 12 月更新版本。与 8 月发布的基础 Qwen-Image 模型相比，Qwen-Image-2512 在图像质量和真实感方面有显著提升。 @@ -55,28 +56,33 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **Text to Image (Qwen-Image 2512)**：标准 50 步生成 - **Text to Image (Qwen-Image 2512 4steps)**：使用 Lightning LoRA 的 4 步加速生成 - -

下载 JSON 工作流

- + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Qwen-Image-2512" in Template Library + + ### 2. 模型下载 **文本编码器** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（可选 - 用于 4 步 Lightning 加速）** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **扩散模型** -- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（推荐大多数用户使用） -- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（如果您有足够的显存并想要更好的质量） +- [qwen_image_2512_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_fp8_e4m3fn.safetensors)（推荐大多数用户使用） +- [qwen_image_2512_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_2512_bf16.safetensors)（如果您有足够的显存并想要更好的质量） **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx index 345d68713..e5a87c4be 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -2,7 +2,7 @@ title: "Qwen-Image-Edit-2511 ComfyUI 原生工作流示例" description: "Qwen-Image-Edit-2511 是 Qwen-Image-Edit 的增强版本，具有改进的角色一致性、多人编辑、集成 LoRA 功能和增强的几何推理能力。" sidebarTitle: "Qwen-Image-Edit-2511" -translationSourceHash: be665671 +translationSourceHash: f1e0ff2e translationFrom: tutorials/image/qwen/qwen-image-edit-2511.mdx --- @@ -33,31 +33,39 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载。 - -

下载 JSON 格式工作流

- + + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Qwen-Image-Edit-2511" in Template Library + + +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) + - -

在 ComfyUI Cloud 上运行

- ### 2. 模型下载 **Text Encoders（文本编码器）** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **LoRA（可选 - 用于 4 步 Lightning 加速）** -- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/resolve/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) +- [Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/Qwen-Image-Edit-2511-Lightning-4steps-V1.0-bf16.safetensors) **Diffusion Models（扩散模型）** -- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) +- [qwen_image_edit_2511_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_2511_bf16.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) **模型存储位置** diff --git a/zh/tutorials/image/qwen/qwen-image-edit.mdx b/zh/tutorials/image/qwen/qwen-image-edit.mdx index 97a4a2e7e..70038b415 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit.mdx @@ -2,15 +2,16 @@ title: "Qwen-Image-Edit ComfyUI 原生工作流示例" description: "Qwen-Image-Edit 是 Qwen-Image 的图像编辑版本，基于20B模型进一步训练，支持精准文字编辑和语义/外观双重编辑能力。" sidebarTitle: "Qwen-Image-Edit" -translationSourceHash: 9c641cf0 +translationSourceHash: 7c27546f translationFrom: tutorials/image/qwen/qwen-image-edit.mdx translationBlockHashes: "_intro": eee08e9a "ComfyOrg Qwen-Image-Edit Live Stream": 6e70a462 - "Qwen-Image-Edit ComfyUI Native Workflow Example": 7df4c813 + "Qwen-Image-Edit ComfyUI Native Workflow Example": 6703460f --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Qwen-Image-Edit** 是 Qwen-Image 的图像编辑版本。它基于20B的Qwen-Image模型进一步训练，成功将Qwen-Image的文本渲染特色能力拓展到编辑任务上，以支持精准的文字编辑。此外，Qwen-Image-Edit将输入图像同时输入到Qwen2.5-VL（获取视觉语义控制）和VAE Encoder（获得视觉外观控制），以同时获得语义/外观双重编辑能力。 @@ -47,13 +48,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) - -

下载 JSON 格式工作流

- + + + Download JSON or search "image_qwen_image_edit" in Template Library + + + Run this workflow on Cloud GPUs with zero setup + + - -

在 ComfyUI Cloud 上运行

- 下载下面的图片作为输入 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) @@ -64,19 +67,19 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Diffusion model** -- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) +- [qwen_image_edit_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_edit_fp8_e4m3fn.safetensors) **LoRA** -- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) +- [Qwen-Image-Lightning-4steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-4steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) Model Storage Location diff --git a/zh/tutorials/image/qwen/qwen-image-layered.mdx b/zh/tutorials/image/qwen/qwen-image-layered.mdx index 49f685ab0..e524ada1f 100644 --- a/zh/tutorials/image/qwen/qwen-image-layered.mdx +++ b/zh/tutorials/image/qwen/qwen-image-layered.mdx @@ -2,17 +2,18 @@ title: "Qwen-Image-Layered ComfyUI 工作流示例" description: "Qwen-Image-Layered 是一个能够将图像分解为多个 RGBA 图层的模型，通过图层分解实现固有的可编辑性。" sidebarTitle: "Qwen-Image-Layered" -translationSourceHash: bf2f2000 +translationSourceHash: 331a3de0 translationFrom: tutorials/image/qwen/qwen-image-layered.mdx translationBlockHashes: "_intro": 19900234 - "Qwen-Image-Layered workflow": ba937275 - "Model links": 98d12555 + "Qwen-Image-Layered workflow": 7a12a7ef + "Model links": 0635af99 "FP8 version": 6bffdd17 - "Workflow settings": b0f81aa2 + "Workflow settings": 098636f1 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Qwen-Image-Layered** 是阿里巴巴通义千问团队开发的模型，能够将图像分解为多个 RGBA 图层。这种分层表示解锁了固有的可编辑性：每个图层都可以独立操作而不影响其他内容。 @@ -30,13 +31,16 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 工作流 - -

下载 JSON 格式工作流

- +| +| +| Download the JSON workflow file +| +| +| +| Run ComfyUI online with zero setup +| +| - -

在 ComfyUI Cloud 上运行

- @@ -44,15 +48,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** -- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) +- [qwen_image_layered_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_layered_bf16.safetensors) **vae** -- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/resolve/main/split_files/vae/qwen_image_layered_vae.safetensors) +- [qwen_image_layered_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-Layered_ComfyUI/blob/main/split_files/vae/qwen_image_layered_vae.safetensors) **模型保存位置** diff --git a/zh/tutorials/image/qwen/qwen-image.mdx b/zh/tutorials/image/qwen/qwen-image.mdx index 86cca15d3..4bb0dc365 100644 --- a/zh/tutorials/image/qwen/qwen-image.mdx +++ b/zh/tutorials/image/qwen/qwen-image.mdx @@ -2,18 +2,19 @@ title: "Qwen-Image ComfyUI原生工作流示例" description: "Qwen-Image 是一个拥有 20B 参数的 MMDiT（多模态扩散变换器）模型，基于 Apache 2.0 许可证开源。" sidebarTitle: "Qwen-Image" -translationSourceHash: f28191a9 +translationSourceHash: 4c682717 translationFrom: tutorials/image/qwen/qwen-image.mdx translationBlockHashes: "_intro": e096d7eb "ComfyOrg Qwen-Image live stream": 6436ac25 - "Qwen-Image Native Workflow Example": 73bab810 - "Qwen Image InstantX ControlNet Workflow": 741d3247 - "Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow": 45d40ad6 - "Qwen Image Union ControlNet LoRA Workflow": 333b4f67 + "Qwen-Image Native Workflow Example": bedb58a8 + "Qwen Image InstantX ControlNet Workflow": c5d39cf3 + "Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow": 6dc30273 + "Qwen Image Union ControlNet LoRA Workflow": a08d8e37 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -61,9 +62,12 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - - 在 Comfy Cloud 上运行 - +| + + + + + 在本篇文档所附工作流中使用的不同模型有三种 1. Qwen-Image 原版模型 fp8_e4m3fn @@ -84,14 +88,9 @@ GPU: RTX4090D 24GB 更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 ![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - -

下载原始版 JSON 格式工作流

- - 蒸馏版 - -

下载蒸馏版JSON 格式工作流

- + + ### 2. 模型下载 **你可以在 ComfyOrg 仓库找到的版本** @@ -104,12 +103,12 @@ GPU: RTX4090D 24GB **Diffusion model** -- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) +- [qwen_image_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors) Qwen_image_distill -- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) -- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) +- [qwen_image_distill_full_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_fp8_e4m3fn.safetensors) +- [qwen_image_distill_full_bf16.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/non_official/diffusion_models/qwen_image_distill_full_bf16.safetensors) - 蒸馏版本原始作者建议在 15 步 cfg 1.0 @@ -118,15 +117,15 @@ Qwen_image_distill **LoRA** -- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/resolve/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) +- [Qwen-Image-Lightning-8steps-V1.0.safetensors](https://huggingface.co/lightx2v/Qwen-Image-Lightning/blob/main/Qwen-Image-Lightning-8steps-V1.0.safetensors) **Text encoder** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **VAE** -- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors) +- [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) 模型保存位置 @@ -165,19 +164,18 @@ Qwen_image_distill 这是一个 ControlNet 模型 - - 在 Comfy Cloud 上运行 - + + + + + + ### 1. 工作流及输入图片 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) - -

下载 JSON 格式工作流

- - 下载下面的图片作为输入 ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) @@ -185,7 +183,7 @@ Qwen_image_distill 1. InstantX Controlnet -下载 [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/resolve/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) 并保存到 `ComfyUI/models/controlnet/` 文件夹下 +下载 [Qwen-Image-InstantX-ControlNet-Union.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-InstantX-ControlNets/blob/main/split_files/controlnet/Qwen-Image-InstantX-ControlNet-Union.safetensors) 并保存到 `ComfyUI/models/controlnet/` 文件夹下 2. **Lotus Depth model** @@ -199,11 +197,11 @@ Qwen_image_distill **Diffusion Model** -- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/resolve/main/lotus-depth-d-v1-1.safetensors) +- [lotus-depth-d-v1-1.safetensors](https://huggingface.co/Comfy-Org/lotus/blob/main/lotus-depth-d-v1-1.safetensors) **VAE Model** -- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors) 或者任意的 SD1.5 的 VAE 都可以使用 +- [vae-ft-mse-840000-ema-pruned.safetensors](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors) 或者任意的 SD1.5 的 VAE 都可以使用 ``` ComfyUI/ @@ -225,9 +223,12 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches 工作流 - - 在 Comfy Cloud 上运行 - + + + + + + 这个模型实际上并不是一个 controlnet，而是一个 Model patch， 支持 canny、depth、inpaint 三种不同的控制模式 @@ -240,10 +241,6 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 下载下面的图片拖入 ComfyUI 中以加载对应的工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - -

下载 JSON 格式工作流

- - 下载下面的图片作为输入图片： ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/input.png) @@ -252,9 +249,9 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 其它模型与 Qwen-Image 基础工作流一致，你只需下载下面的模型并保存到 `ComfyUI/models/model_patches` 文件夹中 -- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) -- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) -- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/resolve/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) +- [qwen_image_canny_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_canny_diffsynth_controlnet.safetensors) +- [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) +- [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) ### 3. 工作流使用说明 @@ -296,9 +293,12 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http ## Qwen Image union ControlNet LoRA 工作流 - - 在 Comfy Cloud 上运行 - + + + + + + 原始模型地址：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 图像结构控制lora 支持 canny、depth、post、lineart、softedge、normal、openpose @@ -307,9 +307,6 @@ Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://h 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -

下载 JSON 格式工作流

- 下载下面的图片作为输入图片 diff --git a/zh/tutorials/image/z-image/z-image-turbo.mdx b/zh/tutorials/image/z-image/z-image-turbo.mdx index e8871670a..cc62fbb37 100644 --- a/zh/tutorials/image/z-image/z-image-turbo.mdx +++ b/zh/tutorials/image/z-image/z-image-turbo.mdx @@ -2,15 +2,16 @@ title: "Z-Image-Turbo ComfyUI 工作流示例" description: "Z-Image-Turbo 是一个蒸馏版 6B 参数高效图像生成模型，可实现亚秒级推理延迟。" sidebarTitle: "Z-Image-Turbo" -translationSourceHash: 6ef05349 +translationSourceHash: b9ae7e8c translationFrom: tutorials/image/z-image/z-image-turbo.mdx translationBlockHashes: "_intro": 062e5fc1 "Z-Image-Turbo text-to-image workflow": 5057b237 - "Z-Image-Turbo Fun Union ControlNet workflow": 18f13b79 + "Z-Image-Turbo Fun Union ControlNet workflow": 34191a15 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -46,15 +47,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### Z-Image-Turbo 模型下载 - + Z-Image-Turbo 文本编码器。 - + Z-Image-Turbo 扩散模型。 - + Z-Image-Turbo VAE。 @@ -81,7 +82,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### ControlNet 所需的额外模型 - + Z-Image-Turbo ControlNet 模型补丁。 diff --git a/zh/tutorials/partner-nodes/anthropic/claude.mdx b/zh/tutorials/partner-nodes/anthropic/claude.mdx index 9a444e45f..9844f4354 100644 --- a/zh/tutorials/partner-nodes/anthropic/claude.mdx +++ b/zh/tutorials/partner-nodes/anthropic/claude.mdx @@ -2,7 +2,7 @@ title: "Anthropic Claude 合作伙伴节点 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Anthropic Claude 合作伙伴节点来完成对话功能" sidebarTitle: "Anthropic Claude" -translationSourceHash: 155df30f +translationSourceHash: b0368d1f translationFrom: tutorials/partner-nodes/anthropic/claude.mdx --- @@ -30,13 +30,16 @@ Anthropic Claude 是 Anthropic 推出的强大 AI 模型系列，以其出色的 ## Anthropic Claude 对话工作流 - - 打开 Comfy Cloud - +Anthropic Claude Chat workflow preview - - 下载 JSON 或在模板库中搜索 "Anthropic Claude" - +Anthropic Claude Chat workflow preview + + + + + + + 在对应模板中，我们构建了一个用于分析图像并生成绘图提示词的角色设定提示词。 diff --git a/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx index e058ee05e..7d18b4589 100644 --- a/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -2,15 +2,16 @@ title: "Beeble SwitchX Partner Nodes ComfyUI 官方示例" description: "本指南介绍如何在 ComfyUI 中使用 Beeble SwitchX Partner Nodes 进行 AI 图像和视频重打光与环境替换" sidebarTitle: "Beeble SwitchX" -translationSourceHash: 3724b3ae +translationSourceHash: 8f168afe translationFrom: tutorials/partner-nodes/beeble/beeble-switchx.mdx translationBlockHashes: "_intro": 470a081a - "Beeble SwitchX: Image Edit": a252a19f - "Beeble SwitchX: Video Edit": 97df1f25 + "Beeble SwitchX: Image Edit": 245e4f8a + "Beeble SwitchX: Video Edit": 166c2e59 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -67,13 +68,12 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; Beeble SwitchX 视频编辑工作流 - - 打开 Comfy Cloud - - - - 下载 JSON 或在模板库中搜索 "Beeble SwitchX: Video Edit" - + + + + + + ### 工作原理 diff --git a/zh/tutorials/partner-nodes/bria/background-removal.mdx b/zh/tutorials/partner-nodes/bria/background-removal.mdx index b16521f0f..ba2e260b0 100644 --- a/zh/tutorials/partner-nodes/bria/background-removal.mdx +++ b/zh/tutorials/partner-nodes/bria/background-removal.mdx @@ -2,15 +2,16 @@ title: "ComfyUI 中使用 Bria 背景移除" description: "学习如何在 ComfyUI 中使用 Bria 合作伙伴节点进行图片和视频背景移除、绿幕抠像和背景替换" sidebarTitle: "Bria 背景移除" -translationSourceHash: c65b0474 +translationSourceHash: dbbb15e9 translationFrom: tutorials/partner-nodes/bria/background-removal.mdx translationBlockHashes: "_intro": f9000433 - "Image Background Removal": 14b7eb56 - "Video Background Processing": 3e621db5 + "Image Background Removal": 988ae876 + "Video Background Processing": aab02df5 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -27,7 +28,7 @@ Bria 的 AI 背景处理模型现已通过合作伙伴节点在 ComfyUI 中使 下载工作流 - + 在 Comfy Cloud 上试试 @@ -44,7 +45,7 @@ Bria 的视频背景处理节点可以移除、替换或为视频背景应用色 下载工作流 - + 在 Comfy Cloud 上试试 @@ -57,7 +58,7 @@ Bria 的视频背景处理节点可以移除、替换或为视频背景应用色 下载工作流 - + 在 Comfy Cloud 上试试 @@ -70,7 +71,7 @@ Bria 的视频背景处理节点可以移除、替换或为视频背景应用色 下载工作流 - + 在 Comfy Cloud 上试试 @@ -83,7 +84,7 @@ Bria 的视频背景处理节点可以移除、替换或为视频背景应用色 下载工作流 - + 在 Comfy Cloud 上试试 diff --git a/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index 518517656..9f4f45751 100644 --- a/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -2,16 +2,17 @@ title: "字节跳动 Seed Audio 1.0 - 通用音频生成" description: "使用 ComfyUI 中的 Seed Audio 1.0，通过单一提示词生成语音、音乐、音效以及多说话人对话，支持声音克隆、预设音色和角色驱动音频。" sidebarTitle: "Seed Audio 1.0" -translationSourceHash: 3fe23cff +translationSourceHash: 86bb1688 translationFrom: tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx translationBlockHashes: - "_intro": 4ce27313 - "Key capabilities": d5c85b17 + "_intro": 7d1fd38c + "Key capabilities": f5c9eb87 "Available workflows": 37e20124 - "How to use Seed Audio 1.0 in ComfyUI": 37aa925d + "How to use Seed Audio 1.0 in ComfyUI": 5630659b "Get started": 1ae5cf45 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 3b9c9c4c5..064e698a4 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -2,17 +2,18 @@ title: "Seedance 2.0 真人支持 - 真人一致性视频生成" description: "在 ComfyUI 中通过一次字节跳动真人活体验证，使用 Seedance 2.0 生成人物一致、音画同步的真人视频" sidebarTitle: "Seedance 2.0 真人支持" -translationSourceHash: 82069c10 +translationSourceHash: 9ea8f507 translationFrom: tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx translationBlockHashes: "_intro": 4890d35b "Video guide": 416e3ec3 "What's different": b67103d8 "How verification works in ComfyUI": ee1e1056 - "Available workflows": bc3d7526 + "Available workflows": 16803a18 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -99,15 +100,51 @@ ComfyUI 提供两套预置的 Seedance 2.0 真人支持模板。两者都会将 使用已验证的人物图片（可选额外参考图片、视频或音频）驱动 Seedance 2.0 生成保持真人身份一致的视频。 - - 获取 Seedance 2.0 真人支持 Reference-to-Video 工作流文件。 +Seedance 2.0 Real Human R2V workflow preview + +Seedance 2.0 Real Human R2V workflow preview + + + + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. + + + + + +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample portrait for verification + +
+Seedance 2.0 Real Human FLF2V workflow preview -### Seedance 2.0 真人支持 首尾帧生成视频（FLF2V） + + + Try the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow instantly on Comfy Cloud. + + + + -提供已验证的起始帧与结束帧，在保持真人身份一致的前提下生成中间视频。 +
+Input materials - - 获取 Seedance 2.0 真人支持 First-Last-Frame-to-Video 工作流文件。 +Download these sample input images to try the workflow: + + + + Download sample first frame + + + Download sample last frame + +
diff --git a/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index 0d6cfc67a..35d08fed0 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -2,17 +2,18 @@ title: "Seedance 2.0 - AI 视频生成" description: "在 ComfyUI 中使用 Seedance 2.0，通过文本、图像、视频和音频生成高质量视频，支持同步音频、角色一致性和电影级摄像机控制" sidebarTitle: "Seedance 2.0" -translationSourceHash: 4c9aec61 +translationSourceHash: 35902650 translationFrom: tutorials/partner-nodes/bytedance/seedance-2-0.mdx translationBlockHashes: "_intro": e1a3013d "Key capabilities": 6ca35ecf - "Available workflows": 4a9917b4 - "Seedance 2.0 Mini": c2cf0c21 + "Available workflows": d232c2a0 + "Seedance 2.0 Mini": db24d1c9 "Using real-person and AI-generated portraits in ComfyUI for Seedance 2.0": d9966fae --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index f8ccfe5b6..582c13235 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -2,17 +2,18 @@ title: "字节跳动 Seedream 5.0 Lite - 智能 AI 图像创作" description: "在 ComfyUI 中使用 Seedream 5.0 lite 进行联网检索和增强指令跟随的图像生成" sidebarTitle: "Seedream 5.0 Lite" -translationSourceHash: ec3a4ae6 +translationSourceHash: 39b0d304 translationFrom: tutorials/partner-nodes/bytedance/seedream-5-lite.mdx translationBlockHashes: "_intro": 1991e54a "What's new in Seedream 5.0 lite": 02ac6a40 - "Seedream 5.0 lite image edit workflow": 44d7f3e4 - "Seedream 5.0 lite text to image workflow": 4fc4c666 + "Seedream 5.0 lite image edit workflow": e90cc958 + "Seedream 5.0 lite text to image workflow": f36785bf "Get started": d14874f4 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -65,13 +66,16 @@ Seedream 5.0 lite 是 BytePlus 最新的图像生成模型。它是 Seedream 系 ## Seedream 5.0 lite 文本到图像工作流 - - 在 Comfy Cloud 中打开 - +Seedream 5.0 Lite Text-to-Image workflow preview - - 下载 JSON 或在模板库中搜索 "Seedream 5.0 lite" - +Seedream 5.0 Lite Text-to-Image workflow preview + + + + + + + ### 世界知识展示 diff --git a/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index 0bea45c57..243f92752 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -2,15 +2,16 @@ title: "字节跳动 Seedream 5.0 Pro - 专业级 AI 图像生成" description: "在 ComfyUI 中使用 Seedream 5.0 Pro，生成高质量图像，具备更强的指令遵循能力、精确编辑和专业级输出" sidebarTitle: "Seedream 5.0 Pro" -translationSourceHash: 125a57d0 +translationSourceHash: c10432af translationFrom: tutorials/partner-nodes/bytedance/seedream-5-pro.mdx translationBlockHashes: "_intro": 0b4a40d7 - "What sets Seedream 5.0 Pro apart": ae41f702 - "Seedream 5.0 Pro text to image workflow": 39d05a61 - "Seedream 5.0 Pro image edit workflow": 7c657db8 + "What sets Seedream 5.0 Pro apart": e1702d4b + "Seedream 5.0 Pro text-to-image workflow": 8dbc58df + "Seedream 5.0 Pro image edit workflow": e2cabdb9 "Get started": 709ac7fd --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -45,13 +46,28 @@ Seedream 5.0 Pro 是字节跳动推出的专业级图像生成模型，作为 Se 通过文本指令编辑现有图像: 更改对象、交换风格、调整颜色或重新构图场景，同时保持原始结构不变。 - +Seedream 5.0 Pro Image Edit workflow preview + +Seedream 5.0 Pro Image Edit workflow preview + + 在 Comfy Cloud 中打开 - + + + + - - 下载 JSON 或在模板库中搜索“Seedream 5.0 Pro” - +
+Input materials + +Download this sample input image to try the workflow: + + + + 下载示例输入图片 + + +
![Seedream 5.0 Pro 图像编辑输入](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/input/interior_college.png) diff --git a/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx index 239b2af8f..d88d8442f 100644 --- a/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -2,16 +2,17 @@ title: "Gemini Omni Flash：对话式视频生成" description: "通过合作节点在 ComfyUI 中使用 Google 的多模态视频模型 Gemini Omni Flash，以自然语言生成和编辑视频" sidebarTitle: "Gemini Omni Flash" -translationSourceHash: 2354813d +translationSourceHash: 6e260d1e translationFrom: tutorials/partner-nodes/google/gemini-omni-flash.mdx translationBlockHashes: "_intro": 3b6973dc "What Gemini Omni Flash offers": 215de783 - "Workflows": d363c40a + "Workflows": 149ee81a "Get started": 64517938 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -41,6 +42,8 @@ Gemini Omni Flash 是 Google DeepMind 推出的高质量、经济高效的视频 +![Gemini Omni Flash 文本转视频工作流预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_t2v-1.webp) + 根据自然语言提示生成电影级视频。将文本描述转换为具有世界感知的运动、光照和声音的视频输出。非常适合社交媒体内容创作、快速视频原型制作以及迭代式视觉叙事。 ### 图像转视频 @@ -60,6 +63,8 @@ Gemini Omni Flash 是 Google DeepMind 推出的高质量、经济高效的视频 +![Gemini Omni Flash 图像转视频工作流预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_i2v-1.webp) + 使用 Gemini Omni Flash 从两张图像生成视频。解释自然语言提示以控制时长和画面比例。非常适合制作简短品牌剪辑、动态社交媒体内容，以及通过对话式提示进行迭代视频编辑。 ### 视频编辑 @@ -76,6 +81,8 @@ Gemini Omni Flash 是 Google DeepMind 推出的高质量、经济高效的视频 +![Gemini Omni Flash 视频编辑工作流预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_gemini_omni_flash_video_edit-1.webp) + 使用 Gemini Omni Flash 以自然语言编辑视频。根据描述性指令将单个输入视频转换为经过编辑的输出。在提示中指定时长和画面比例。非常适合快速社交媒体混剪、电影场景调整以及迭代视频精修。 ## 开始使用 diff --git a/zh/tutorials/partner-nodes/google/gemini.mdx b/zh/tutorials/partner-nodes/google/gemini.mdx index 92c3a0fa7..fa9a0b0f4 100644 --- a/zh/tutorials/partner-nodes/google/gemini.mdx +++ b/zh/tutorials/partner-nodes/google/gemini.mdx @@ -2,7 +2,7 @@ title: " Google Gemini 合作伙伴节点 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Google Gemini 合作伙伴节点来完成对话功能" sidebarTitle: "Google Gemini" -translationSourceHash: ddd327ee +translationSourceHash: 8cd6027c translationFrom: tutorials/partner-nodes/google/gemini.mdx --- @@ -22,13 +22,16 @@ Google Gemini 是 Google 推出的一款强大的 AI 模型，支持对话、文 请下载下面的 Json 文件并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + + Download Json Format Workflow File + + + 在 Comfy Cloud 中打开 + + + ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/partner-nodes/google/nano-banana-2.mdx b/zh/tutorials/partner-nodes/google/nano-banana-2.mdx index 4b378a911..90729ca2f 100644 --- a/zh/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/zh/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -2,17 +2,18 @@ title: "Nano Banana 2 - 快速 AI 图像生成" description: "在 ComfyUI 中使用 Nano Banana 2 以 Flash 速度生成 Pro 级别质量的图像" sidebarTitle: "Nano Banana 2" -translationSourceHash: a146b8b0 +translationSourceHash: a86a55b2 translationFrom: tutorials/partner-nodes/google/nano-banana-2.mdx translationBlockHashes: "_intro": 379a03e0 "What's new in Nano Banana 2": 4048475a - "Nano Banana 2 image edit workflow": 085cc8cb + "Nano Banana 2 image edit workflow": ccade8b6 "Which model should you pick?": f16d5cc7 "Get started": f6189d9e --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/zh/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index 1e1f96635..b5495d33a 100644 --- a/zh/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/zh/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -2,6 +2,8 @@ title: "Grok Imagine Video 1.5 图生视频 ComfyUI 官方示例" description: "本指南介绍如何在 ComfyUI 中使用 Grok Imagine Video 1.5 Partner Node，从图片生成带原生音频的高质量视频" sidebarTitle: "Grok Imagine Video 1.5" +translationSourceHash: 6a40c4e4 +translationFrom: tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx --- import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; diff --git a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 6de18b9a4..321d75da8 100644 --- a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -2,18 +2,19 @@ title: "ComfyUI 中的 HappyHorse 1.0 视频生成" description: "了解如何在 ComfyUI 中通过合作伙伴节点使用 HappyHorse 1.0 进行图生视频、文生视频、参考生成视频和视频编辑，享受电影级美学与多镜头一致性" sidebarTitle: "HappyHorse 1.0" -translationSourceHash: a0fbfb64 +translationSourceHash: 0b14a684 translationFrom: tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx translationBlockHashes: "_intro": 4eba8bbb "Highlights": 376295d6 - "HappyHorse 1.0 image-to-video": 85d1766b - "HappyHorse 1.0 text-to-video": 29f5941d - "HappyHorse 1.0 reference-to-video": fba3d9ea - "HappyHorse 1.0 video edit": 20a4559a + "HappyHorse 1.0 image-to-video": d41d0b77 + "HappyHorse 1.0 text-to-video": 6c594c34 + "HappyHorse 1.0 reference-to-video": b8ae994c + "HappyHorse 1.0 video edit": 5b531078 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -66,13 +67,46 @@ HappyHorse 1.0 聚焦于强美学、多镜头叙事和强大的编辑型工作 src="https://github.com/Comfy-Org/example_workflows/raw/refs/heads/main/api_nodes/happy_horse/1.0/t2v_1.mp4" > - - 获取 HappyHorse 1.0 文生视频工作流文件。 + + + 在 Comfy Cloud 上立即试用文本转视频工作流 - - - 在 Comfy Cloud 上即刻体验文生视频工作流。 + + + +![HappyHorse 1.0 Text-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_t2v-1.webp) + + + 在 Comfy Cloud 上立即试用参考转视频工作流 + + + + +![HappyHorse 1.0 Reference-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_r2v-1.webp) + + + Get the example car reference image. + + + + Get the example person reference image. + + + + Try the Video Edit workflow instantly on Comfy Cloud. + + + +![HappyHorse 1.0 Video Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_video_edit-1.webp) + + + Get the example input image for the video edit workflow. + + + + Get the example input video for the video edit workflow. + ## HappyHorse 1.0 参考生成视频 diff --git a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index 1060dc58d..4b44c86c3 100644 --- a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -2,18 +2,19 @@ title: "在 ComfyUI 中使用 HappyHorse 1.1 生成视频" description: "了解如何通过 ComfyUI 中的合作节点使用 HappyHorse 1.1 进行图生视频、文生视频和参考生视频，具有原生同步音频和电影级多镜头叙事" sidebarTitle: "HappyHorse 1.1" -translationSourceHash: 1c2775c3 +translationSourceHash: 2fadefbd translationFrom: tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx translationBlockHashes: "_intro": 6ae1c8e8 "Highlights": 10f03d9f - "HappyHorse 1.1 text-to-video": 31c81d0f - "HappyHorse 1.1 image-to-video": 2fa5bff9 - "HappyHorse 1.1 reference-to-video": 983e4818 + "HappyHorse 1.1 text-to-video": 3551ba66 + "HappyHorse 1.1 image-to-video": 05ad4488 + "HappyHorse 1.1 reference-to-video": 39830de6 "Getting started": 27bbb438 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index bb2a0f49f..3468b13ce 100644 --- a/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -2,19 +2,20 @@ title: "Hunyuan 3D API 节点模型生成 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Hunyuan 3D 节点的 API 进行 3D 模型生成" sidebarTitle: "Hunyuan 3D 3.0" -translationSourceHash: 8d72daac +translationSourceHash: c92f6932 translationFrom: tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx translationBlockHashes: "_intro": 312a61a4 "Use cases": 60e6d268 "Getting started": 62f4530d - "Text-to-3D workflow": 710fb1e6 - "Image-to-3D workflow": d7a3b87d - "Multi-view-to-3D workflow": 68952799 - "Advanced features": acc7cc22 + "Text-to-3D workflow": 5edf82d4 + "Image-to-3D workflow": ce7efef1 + "Multi-view-to-3D workflow": ff251560 + "Advanced features": 6b37a964 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index 72177ef29..1b795429f 100644 --- a/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -2,18 +2,19 @@ title: "Hunyuan 3D API 节点模型生成 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Hunyuan 3D 节点的 API 进行 3D 模型生成" sidebarTitle: "模型生成" -translationSourceHash: cb7d1627 +translationSourceHash: aa9c5029 translationFrom: tutorials/partner-nodes/hunyuan3d/model-generation.mdx translationBlockHashes: "_intro": 312a61a4 "Use cases": 60e6d268 "Getting started": 62f4530d - "Text-to-3D workflow": 710fb1e6 - "Image-to-3D workflow": d7a3b87d - "Multi-view-to-3D workflow": 68952799 + "Text-to-3D workflow": 5edf82d4 + "Image-to-3D workflow": ce7efef1 + "Multi-view-to-3D workflow": ff251560 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index f89eba2e9..1ead03fb9 100644 --- a/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Ideogram 4.0 节点教程" description: "了解如何在 ComfyUI 中使用 Ideogram 4.0 API Partner Node" sidebarTitle: "Ideogram 4.0" -translationSourceHash: 20fee685 +translationSourceHash: c4f854a7 translationFrom: tutorials/partner-nodes/ideogram/ideogram-v4.mdx translationBlockHashes: "_intro": 9a8e46c4 - "Ideogram 4.0 Partner Node Text-to-Image Workflow": ca6fe17b + "Ideogram 4.0 Partner Node Text-to-Image Workflow": ca41ad51 "Additional Notes": 5cc3cb89 "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx index f685102aa..1fbb77a57 100644 --- a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -2,19 +2,20 @@ title: "Kling 2.6 Motion Control 合作伙伴节点 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Kling 2.6 Motion Control 合作伙伴节点，实现从参考视频到角色图像的精准动作迁移" sidebarTitle: "Kling 2.6 Motion Control" -translationSourceHash: 89919ffc +translationSourceHash: e449a655 translationFrom: tutorials/partner-nodes/kling/kling-motion-control.mdx translationBlockHashes: "_intro": fac1cb1e "Product highlights": a82b89a7 "Character orientation modes": aa2f7dec "Model tiers": 565d97ae - "Kling 2.6 Motion Control workflow": c75b1c17 + "Kling 2.6 Motion Control workflow": 33f2f69c "Input requirements": 0ecbf467 "Tips for better results": b6fdceb5 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -51,9 +52,34 @@ Kling 2.6 Motion Control 是由快手开发的专门多模态模型，能够实 ## Kling 2.6 Motion Control 工作流 - -

下载 Json 格式工作流文件

- + + Kling 2.6 Motion Control workflow preview + +Kling 2.6 Motion Control workflow preview + + + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. + + + + Download the workflow JSON file for local use. + + +
+Input materials + +Download these sample input files to try the workflow: + + + + Download sample reference image + + + Download sample motion reference video + + +
+ ## 输入要求 diff --git a/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx b/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx index f85cacd9e..24feaed8d 100644 --- a/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -2,14 +2,14 @@ title: "Luma Uni-1 使用指南" description: "在 ComfyUI 中使用 Luma Uni-1 合作伙伴节点进行图像创建与编辑。" sidebarTitle: "Luma Uni-1" -translationSourceHash: 16838ef3 +translationSourceHash: 99d39585 translationFrom: tutorials/partner-nodes/luma/luma-uni-1.mdx translationBlockHashes: "_intro": 544c5d43 "What makes Uni-1 different": c0181542 "Strengths": 4dd065dd "The core distinction: Create vs Modify": c3ea208f - "Available workflows": a6cf5659 + "Available workflows": 9b93a043 "Core parameters": b43ac975 "Working with reference images": d20d34ae "Prompting guidelines": ba9ac50d @@ -23,6 +23,7 @@ translationBlockHashes: --- + 在 **ComfyUI** 里，Luma **Uni-1** 以 **合作伙伴 API 节点**形式接入：**Create** 路线根据提示词生成新图（可挂多张参考图）；**Modify** 路线以已有图像为输入做定向编辑。操作上与普通工作流相同——用 **加载图像** / **保存图像** 与 Luma 节点串联，在节点上填写提示词、种子、宽高比与参考图槽位，再在本地排队运行或通过 **Comfy Cloud** 打开下方模板。 Luma 将 Uni-1 表述为非扩散、自回归类的模型，会在成图前对提示词做推理；在画布侧更需要关注的是 **Create / Modify** 选型、参考图角色写清楚，以及用种子做可控迭代。 @@ -74,13 +75,48 @@ Uni-1 在广泛的任务中表现优异： ### Image Create 工作流 - - 在 Comfy Cloud 上即时体验 Image Create 工作流。 +Luma Uni-1 Image Create workflow preview + +Luma Uni-1 Image Create workflow preview + + + + + + + + +
+Input materials + +Download this sample input image to try the workflow: + + + + Download sample style reference + +
+Luma Uni-1 Image Edit workflow preview + + + + + + + + +
+Input materials + +Download this sample input image to try the workflow: - - 下载 JSON 或在模板库中搜索 "Luma UNI-1 Image Create" + + + Download sample source image + +
### Image Edit 工作流 diff --git a/zh/tutorials/partner-nodes/meshy/meshy-6.mdx b/zh/tutorials/partner-nodes/meshy/meshy-6.mdx index 912c09c06..a7c7dc87c 100644 --- a/zh/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/zh/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -2,10 +2,17 @@ title: "Meshy 6 API 节点 3D 模型生成 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Meshy 6 节点的 API 进行 3D 模型生成" sidebarTitle: "Meshy 6" -translationSourceHash: 302cd348 +translationSourceHash: f3036351 translationFrom: tutorials/partner-nodes/meshy/meshy-6.mdx +translationBlockHashes: + "_intro": 2dd46cd7 + "About Meshy 6": bc605135 + "Text-to-Model Workflow": f800f061 + "Image-to-Model Workflow": c148120d + "Multi-view to Model Workflow": 0cf6bb73 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -29,21 +36,37 @@ Meshy 6 是 Meshy 最新一代的 3D 模型生成技术，在几何质量、纹 使用 Meshy 6 直接从文本描述生成 3D 模型。 - - 在 Comfy Cloud 上即时运行文本到模型工作流。 - + + Run the text-to-model workflow instantly on Comfy Cloud. + +![Meshy 6 Text-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_text_to_model-1.webp) - - 下载工作流 JSON 文件以供本地使用。 - + + Run the image-to-model workflow instantly on Comfy Cloud. + +![Meshy 6 Image-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_image_to_model-1.webp) -## 图像到模型工作流 + + 获取此工作流的示例输入图片 + -使用 Meshy 6 的图像到 3D 功能将 2D 图像转换为详细的 3D 模型。 + + Run the multi-view workflow instantly on Comfy Cloud. + - - 在 Comfy Cloud 上即时运行图像到模型工作流。 +![Meshy 6 Multi-view to Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_multi_image_to_model-1.webp) + + + + Get the example front view image. + + + Get the example back view image. + + + Get the example side view image. + 下载工作流 JSON 文件以供本地使用。 diff --git a/zh/tutorials/partner-nodes/openai/chat.mdx b/zh/tutorials/partner-nodes/openai/chat.mdx index 10107c75e..3c142ef69 100644 --- a/zh/tutorials/partner-nodes/openai/chat.mdx +++ b/zh/tutorials/partner-nodes/openai/chat.mdx @@ -2,7 +2,7 @@ title: " OpenAI Chat 合作伙伴节点 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 OpenAI Chat 合作伙伴节点来完成对话功能" sidebarTitle: "OpenAI Chat" -translationSourceHash: e4c0b73e +translationSourceHash: 77289d08 translationFrom: tutorials/partner-nodes/openai/chat.mdx --- @@ -22,13 +22,16 @@ OpenAI 是一家专注于生成式 AI 的科技公司，提供强大的对话功 请下载下面的 Json 文件并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + + 在 Comfy Cloud 中打开 + + + Download the JSON format workflow file. + + + ### 2. 按步骤完成工作流的运行 diff --git a/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx b/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx index cad988dc1..68a0eb502 100644 --- a/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -3,18 +3,19 @@ title: "OpenAI GPT-Image-2 节点" description: "了解如何在 ComfyUI 中使用 OpenAI GPT-Image-2 合作伙伴节点生成图像" sidebarTitle: "GPT-Image-2" icon: "image" -translationSourceHash: b99cdc88 +translationSourceHash: ede8405e translationFrom: tutorials/partner-nodes/openai/gpt-image-2.mdx translationBlockHashes: "_intro": 9f80a1fc "Node Overview": 84ac2d88 "Getting Started": ab5a2231 - "Available workflows": 08afb96c + "Available workflows": d687898b "Key Capabilities": df71dcca "Hybrid Pipelines": 9c6479f0 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/pricing.mdx b/zh/tutorials/partner-nodes/pricing.mdx index 170ffe066..2d64a6830 100644 --- a/zh/tutorials/partner-nodes/pricing.mdx +++ b/zh/tutorials/partner-nodes/pricing.mdx @@ -3,7 +3,7 @@ title: "定价" description: "本文列出了当前合作节点的价格。所有价格均以积分计价（211 积分 = 1 美元）。" sidebarTitle: "定价" mode: wide -translationSourceHash: c864b96d +translationSourceHash: 907be779 translationFrom: tutorials/partner-nodes/pricing.mdx translationBlockHashes: "_intro": 845592d4 @@ -16,7 +16,9 @@ translationBlockHashes: "Magnific": 4000e7e1 "Google": 21dc3165 "HappyHorse": bceb584a - "Hitpaw": e83827b3 + "Hitpaw": 320c0537 + "Ideogram": 9410bf72 + "Krea": fb20c257 "Kling": cdf2c899 "Lightricks": 4870964f "Luma": 655207a5 @@ -52,6 +54,7 @@ translationBlockHashes: + 下表列出了当前合作节点的价格。所有价格均以积分计算。 211 积分 = 1 美元 diff --git a/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx b/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx index d8b33193d..351c7319f 100644 --- a/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -2,17 +2,18 @@ title: "Recraft V4 图像与矢量生成教程" description: "在 ComfyUI 中使用 Recraft V4 生成专业图像和可用于生产的矢量图" sidebarTitle: "Recraft V4" -translationSourceHash: fe0eaf6f +translationSourceHash: da066915 translationFrom: tutorials/partner-nodes/recraft/recraft-v4.mdx translationBlockHashes: "_intro": 8e346040 "What's new in V4": 6bdb9842 - "Recraft V4 text to image workflow": 7273fda3 - "Recraft V4 text to vector workflow": d322ba5f + "Recraft V4 text to image workflow": 2ff998b0 + "Recraft V4 text to vector workflow": c3255bfc "Additional notes": d5f3f109 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/rodin/model-generation.mdx b/zh/tutorials/partner-nodes/rodin/model-generation.mdx index 051d1f871..7b64890af 100644 --- a/zh/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/zh/tutorials/partner-nodes/rodin/model-generation.mdx @@ -2,16 +2,17 @@ title: " Rodin 合作伙伴节点模型生成 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Rodin 节点的 API 来进行模型生成" sidebarTitle: "模型生成" -translationSourceHash: 8d58740b +translationSourceHash: 738362ca translationFrom: tutorials/partner-nodes/rodin/model-generation.mdx translationBlockHashes: "_intro": a35f46e8 - "Single-view Model Generation Workflow": d982caac - "Multi-view Model Generation Workflow": 8a727575 + "Single-view Model Generation Workflow": d7303db4 + "Multi-view Model Generation Workflow": 5c5d3550 "Other Related Nodes": c885ce31 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -32,13 +33,14 @@ Hyper3D Rodin (hyper3d.ai) 是一个专注于通过人工智能快速生成高 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + Single-view Model Generation (Json Format) + + + 在 Comfy Cloud 中打开 + + 下载下面的图片作为输入图片 @@ -66,13 +68,11 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + 下载示例输入图片 + + 下载下面的图片作为输入图片 diff --git a/zh/tutorials/partner-nodes/topaz/astra-2.mdx b/zh/tutorials/partner-nodes/topaz/astra-2.mdx index 15826ca94..7c36435f6 100644 --- a/zh/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/zh/tutorials/partner-nodes/topaz/astra-2.mdx @@ -2,7 +2,7 @@ title: "Astra 2 - 创意扩散视频放大" description: "Astra 2 模型在 ComfyUI 中的用途，以及对细节和风格的可控创意放大。" sidebarTitle: "Astra 2" -translationSourceHash: 2840b2ae +translationSourceHash: d28f921a translationFrom: tutorials/partner-nodes/topaz/astra-2.mdx --- diff --git a/zh/tutorials/partner-nodes/tripo/model-generation.mdx b/zh/tutorials/partner-nodes/tripo/model-generation.mdx index dca3da2b4..be5a0b82f 100644 --- a/zh/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/zh/tutorials/partner-nodes/tripo/model-generation.mdx @@ -2,17 +2,18 @@ title: " Tripo 合作伙伴节点模型生成 ComfyUI 官方示例" description: "本文将介绍如何在 ComfyUI 中使用 Tripo 节点的 API 来进行模型生成" sidebarTitle: "模型生成" -translationSourceHash: 91cfb63d +translationSourceHash: 9fa47d10 translationFrom: tutorials/partner-nodes/tripo/model-generation.mdx translationBlockHashes: "_intro": b329659d - "Text-to-Model Workflow": 1f82cb18 - "Image-to-Model Workflow": 20fac619 - "Multi-view Model Generation Workflow": b0e0a897 + "Text-to-Model Workflow": ae8e8384 + "Image-to-Model Workflow": 7463fca4 + "Multi-view Model Generation Workflow": c0a49c52 "Subsequent Task Processing for the Same Task": d24582e4 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -35,9 +36,14 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + Try the Text-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + ### 2. 按步骤完成工作流的运行 @@ -58,9 +64,14 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + Try the Image-to-Model workflow instantly on Comfy Cloud. + + + Download JSON to run locally in ComfyUI. + + 下载下面的图片作为输入图片 @@ -86,9 +97,11 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - -

下载 Json 格式工作流文件

- + + + 下载示例输入图片 + + 下载下面的图片作为输入图片 diff --git a/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx index 31d15fe6a..b3d3a21d9 100644 --- a/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -2,16 +2,17 @@ title: "Tripo 3.1 — 高细节 3D 资产生成 ComfyUI 官方指南" description: "了解如何在 ComfyUI 中通过合作伙伴节点使用 Tripo 3.1 生成高细节 3D 资产，具备高密度几何和 PBR 就绪材质，适用于生产级场景。" sidebarTitle: "Tripo 3.1" -translationSourceHash: c520daa5 +translationSourceHash: dbbf32c8 translationFrom: tutorials/partner-nodes/tripo/tripo-3-1.mdx translationBlockHashes: "_intro": ff0ef375 "Capabilities": 6187e278 "Use Cases": 10ad3fe3 - "Available Workflows": c68e812a + "Available Workflows": 5ec1d9c5 --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -40,13 +41,38 @@ Tripo 3.1 是通过 ComfyUI Tripo 合作伙伴节点提供的最新模型版本 ### 文本转模型 - - 在 Comfy Cloud 上立即尝试文本转模型工作流。 - +Tripo 3.1 Text-to-Model workflow preview - - 下载 JSON 或在模板库中搜索 "Tripo 3.1 Text-to-Model"。 - +Generate a high-detail 3D model from a text prompt using Tripo 3.1. + +Tripo 3.1 Text-to-Model workflow preview + + + + + + + +Generate a high-detail 3D model from an image input using Tripo 3.1. + +Tripo 3.1 Image-to-Model workflow preview + + + + + + + +Generate a high-detail 3D model from multiple view images using Tripo 3.1. + +Tripo 3.1 Multiview-to-Model workflow preview + + + + + + + ### 图片转模型 diff --git a/zh/tutorials/partner-nodes/wan/wan2-7.mdx b/zh/tutorials/partner-nodes/wan/wan2-7.mdx index 092f070a1..755408c95 100644 --- a/zh/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/zh/tutorials/partner-nodes/wan/wan2-7.mdx @@ -2,19 +2,20 @@ title: "ComfyUI 中的 Wan2.7 视频生成" description: "了解如何在 ComfyUI 中通过合作伙伴节点使用 Wan2.7 进行图生视频、文生视频、参考生成视频、视频续写和视频编辑" sidebarTitle: "Wan2.7" -translationSourceHash: 3bb39265 +translationSourceHash: 42f5b589 translationFrom: tutorials/partner-nodes/wan/wan2-7.mdx translationBlockHashes: "_intro": 611e27ab "Key features": 6ca0965f "Highlights": cec936d1 - "Wan2.7 image-to-video": 29a49b02 - "Wan2.7 text-to-video": 25cfcacf - "Wan2.7 reference-to-video": cf70840e - "Wan2.7 video edit": ad198739 + "Wan2.7 image-to-video": af05fd5c + "Wan2.7 text-to-video": c9f23ab0 + "Wan2.7 reference-to-video": fe373e60 + "Wan2.7 video edit": 0d2511bb --- + import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -44,13 +45,40 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 从图像输入生成视频。支持首帧、首尾帧和音频驱动生成模式。 - - 获取 Wan2.7 图生视频工作流文件。 - - - - 在 Comfy Cloud 上即刻体验图生视频工作流。 - +Wan2.7 I2V workflow preview + +Wan2.7 I2V workflow preview + + + + + + + +Wan2.7 T2V workflow preview + + + + + + + +Wan2.7 R2V workflow preview + + + + + + + +Wan2.7 Video Edit workflow preview + + + + + + + ## Wan2.7 文生视频 diff --git a/zh/tutorials/utility/moge.mdx b/zh/tutorials/utility/moge.mdx index 0c6c7223d..f4e05e4cf 100644 --- a/zh/tutorials/utility/moge.mdx +++ b/zh/tutorials/utility/moge.mdx @@ -2,19 +2,20 @@ title: "ComfyUI MoGe 使用示例" description: "本指南演示如何在 ComfyUI 中使用 MoGe 进行单目几何估计：公尺度点云、深度图、法线图和网格生成。" sidebarTitle: "MoGe" -translationSourceHash: 7ae1c247 +translationSourceHash: ff036dcc translationFrom: tutorials/utility/moge.mdx translationBlockHashes: "_intro": f3974fd4 "Model Installation": fb228472 "Example Workflows": cf2f84c1 - "1. Depth Estimation": 0c7bafb8 - "2. Perspective to Mesh": 03608c5b - "3. Panorama to Mesh": bdbb850d + "1. Depth Estimation": 030e1494 + "2. Perspective to Mesh": 4fde2470 + "3. Panorama to Mesh": b273460c "Community Resources": d9fcd8cd --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' # ComfyUI MoGe 介绍 @@ -50,8 +51,8 @@ ComfyUI 现已原生支持 MoGe 节点。开始前请确保已更新到最新版 下载 MoGe 检查点并保存到相应的 ComfyUI 文件夹： -- **MoGe-2（推荐）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1（基线版本）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2（推荐）**：[moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1（基线版本）**：[moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/bytedance/bernini-r.mdx b/zh/tutorials/video/bytedance/bernini-r.mdx index 8506530ac..6f8bdf07d 100644 --- a/zh/tutorials/video/bytedance/bernini-r.mdx +++ b/zh/tutorials/video/bytedance/bernini-r.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Bernini-R 官方示例" description: "了解如何在 ComfyUI 中使用 Bernini-R 进行图像和视频编辑——重光照、风格转换、主体插入等。" sidebarTitle: "Bernini-R" -translationSourceHash: 84ecde34 +translationSourceHash: e0b08f10 translationFrom: tutorials/video/bytedance/bernini-r.mdx translationBlockHashes: - "_intro": efe73774 - "Model Installation": 84eedf7b + "_intro": b791048b + "Model Installation": 286fbd69 "Example Workflows": cf2f84c1 - "1. Image Editing": cd75c4ca - "2. Video Editing": 9e90382c - "Community Resources": 57f652a5 + "1. Image Editing": 0c53cda8 + "2. Video Editing": ef878ac4 + "Community Resources": 9b4a0aaf --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' # ComfyUI Bernini-R 简介 @@ -49,16 +50,16 @@ ComfyUI 现已原生支持 Bernini-R 节点。开始前请确保已更新到最 下载所需的模型权重并将其保存到对应的 ComfyUI 文件夹： **text_encoders:** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **vae:** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors?download=true) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors?download=true) **loras:** -- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) +- [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors?download=true) **diffusion_models:** -- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/resolve/main/wan2.2_bernini_r_fp16.safetensors) +- [wan2.2_bernini_r_fp16.safetensors](https://huggingface.co/Comfy-Org/Bernini-R/blob/main/wan2.2_bernini_r_fp16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index a61a79890..aea227b02 100644 --- a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -2,7 +2,7 @@ title: "Cosmos Predict2 视频生成 ComfyUI 官方示例" description: "本文介绍了如何在 ComfyUI 中完成 Cosmos-Predict2 文生视频及图生视频的工作流" sidebarTitle: "Cosmos-Predict2" -translationSourceHash: 7c20daa6 +translationSourceHash: 3386fb94 translationFrom: tutorials/video/cosmos/cosmos-predict2-video2world.mdx --- @@ -37,9 +37,14 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/cosmos_predict2_2B_video2world_480p_16fps.mp4" > - -

下载 Json 格式工作流文件

- + + + Cosmos-Predict2 source code and documentation + + + Cosmos-Predict2 model collection + + 请下载下面的图片作为输入文件： @@ -50,17 +55,17 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- **Diffusion model** -- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) +- [cosmos_predict2_2B_video2world_480p_16fps.safetensors](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/blob/main/cosmos_predict2_2B_video2world_480p_16fps.safetensors) 其它权重请访问 [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) 进行下载 **Text encoder** -[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) +[oldt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/blob/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors) **VAE** -[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +[wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) 文件保存位置 diff --git a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 6d4a77890..1b9048a36 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -2,10 +2,18 @@ title: "混元视频 1.5 ComfyUI 教程" description: "了解如何使用混元视频 1.5，一个轻量级的 8.3B 参数模型，可在消费级 GPU 上生成高质量视频" sidebarTitle: "混元视频 1.5" -translationSourceHash: 0b15fd89 +translationSourceHash: 4379c5f7 translationFrom: tutorials/video/hunyuan/hunyuan-video-1-5.mdx +translationBlockHashes: + "_intro": 74d9d0ad + "Model highlights": a18b0a64 + "Common models for all workflows": 87181049 + "Hunyuan Video 1.5 Text-to-Video Workflow": 73a4afe7 + "Hunyuan Video 1.5 Image-to-Video Workflow": 42a0612d + "Super-resolution upscaler": 89338328 --- + import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; 混元视频 1.5是由腾讯混元团队开发的轻量级 8.3B 参数模型。它可在消费级 GPU（24GB 显存）上提供旗舰级质量的视频生成，大幅降低了使用门槛，同时不影响质量。 @@ -24,17 +32,17 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; **text_encoders** -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) **diffusion_models** -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) +- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) **vae** -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/resolve/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) Model Storage Location @@ -56,4 +64,4 @@ Model Storage Location [video_hunyuan_video_1.5_720p_i2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_i2v.json) -[video_hunyuan_video_1.5_720p_t2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_t2v.json) \ No newline at end of file +[video_hunyuan_video_1.5_720p_t2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_t2v.json) diff --git a/zh/tutorials/video/hunyuan/hunyuan-video.mdx b/zh/tutorials/video/hunyuan/hunyuan-video.mdx index 6f53ebd0a..244df58a0 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video.mdx @@ -2,17 +2,18 @@ title: "ComfyUI 混元视频示例" description: "本文介绍了如何在 ComfyUI 中完成混元文生视频及图生视频的工作流" sidebarTitle: "混元视频" -translationSourceHash: 045f4b68 +translationSourceHash: 2cbcac4c translationFrom: tutorials/video/hunyuan/hunyuan-video.mdx translationBlockHashes: "_intro": 1af8ce94 - "Common Models for All Workflows": 3e4807f2 - "Hunyuan Text-to-Video Workflow": 8c3a1c81 - "Hunyuan Image-to-Video Workflow": ef169a18 - "Try it yourself": d4ee74b3 + "Common Models for All Workflows": 004bda25 + "Hunyuan Text-to-Video Workflow": fdc74d6f + "Hunyuan Image-to-Video Workflow": fef3b225 + "Try It Yourself": 076fb43f --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - -

下载 Json 格式工作流文件

- + + + Wan2.1 Fun Camera 1.3B diffusion model + + + Wan2.1 Fun Camera 14B diffusion model + + 如果你想使用 14B 版本，只需要将模型文件替换为 14B 版本即可，但请注意显存要求。 @@ -119,9 +125,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B.mp4" > - -

下载 Json 格式工作流文件

- + + + Full precision text encoder + + + FP8 quantized text encoder (recommended for lower VRAM) + + **输入图片** ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B_input.jpg) diff --git a/zh/tutorials/video/wan/fun-control.mdx b/zh/tutorials/video/wan/fun-control.mdx index a45cec2a1..0f6572dde 100644 --- a/zh/tutorials/video/wan/fun-control.mdx +++ b/zh/tutorials/video/wan/fun-control.mdx @@ -2,18 +2,19 @@ title: "ComfyUI Wan2.1 Fun Control 视频示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.1 Fun Control 使用控制视频来完成视频生成的示例" sidebarTitle: "Wan2.1 Fun Control" -translationSourceHash: 88043153 +translationSourceHash: f6e4d92d translationFrom: tutorials/video/wan/fun-control.mdx translationBlockHashes: "_intro": 9efc0241 - "About Wan2.1-Fun-Control": 19f05e50 - "Model Installation": aee3a183 - "ComfyUI Native Workflow": 13eb7888 - "Workflow Using Custom Nodes": cf408ff0 - "Usage Tips": ae173b3b + "About Wan2.1-Fun-Control": 2e562079 + "Model Installation": ef08dc73 + "ComfyUI Native Workflow": d64a3fa4 + "Workflow Using Custom Nodes": ca1f733a + "Usage Tips": 9e74c201 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 Wan2.1-Fun-Control @@ -54,18 +55,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 点击对应链接进行下载，如果你之前使用过 Wan 相关的工作流，那么你仅需要下载 **Diffusino models** **Diffusion models** 选择 1.3B 或 14B, 14B 的文件体积更大（32GB）但是对于运行显存要求也较高， -- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) +- [wan2.1_fun_control_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_control_1.3B_bf16.safetensors?download=true) - [Wan2.1-Fun-14B-Control](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-Control/blob/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-Control.safetensors` **Text encoders** 选择下面两个模型中的一个，fp16 精度体积较大对性能要求高 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/fun-inp.mdx b/zh/tutorials/video/wan/fun-inp.mdx index d0e8741aa..842454283 100644 --- a/zh/tutorials/video/wan/fun-inp.mdx +++ b/zh/tutorials/video/wan/fun-inp.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Wan2.1 Fun InP 视频示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.1 Fun InP 视频首尾帧视频生成示例" sidebarTitle: "Wan2.1 Fun InP" -translationSourceHash: db96fa7d +translationSourceHash: 866f2586 translationFrom: tutorials/video/wan/fun-inp.mdx translationBlockHashes: "_intro": 9efc0241 - "About Wan2.1-Fun-InP": ab94df56 - "Wan2.1 Fun InP Workflow": 33ad43c2 + "About Wan2.1-Fun-InP": f67d80ee + "Wan2.1 Fun InP Workflow": 140545b2 "Other Wan2.1 Fun InP or video-related custom node packages": 4c96c810 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 Wan2.1-Fun-InP @@ -54,18 +55,18 @@ Wan-Fun InP 是阿里巴巴推出的开源视频生成模型，属于 ​​Wan2 下面的模型你可以在 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 和 [Wan2.1-Fun](https://huggingface.co/collections/alibaba-pai/wan21-fun-67e4fb3b76ca01241eb7e334) 找到 **Diffusion models** 选择 1.3B 或 14B, 14B 的文件体积更大（32GB）但是对于运行显存要求也较高， -- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) -- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/resolve/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-InP.safetensors` +- [wan2.1_fun_inp_1.3B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_fun_inp_1.3B_bf16.safetensors?download=true) +- [Wan2.1-Fun-14B-InP](https://huggingface.co/alibaba-pai/Wan2.1-Fun-14B-InP/blob/main/diffusion_pytorch_model.safetensors?download=true)： 建议下载后重命名为 `Wan2.1-Fun-14B-InP.safetensors` **Text encoders** 选择下面两个模型中的一个，fp16 精度体积较大对性能要求高 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/vace.mdx b/zh/tutorials/video/wan/vace.mdx index 4d6fc967f..e650ee9c1 100644 --- a/zh/tutorials/video/wan/vace.mdx +++ b/zh/tutorials/video/wan/vace.mdx @@ -2,20 +2,22 @@ title: "ComfyUI Wan2.1 VACE 视频示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.1 VACE 视频生成示例" sidebarTitle: "Wan2.1 VACE" -translationSourceHash: 9b04cff5 +translationSourceHash: 882bb348 translationFrom: tutorials/video/wan/vace.mdx translationBlockHashes: - "_intro": f6c2c850 - "About VACE": 4fa63090 - "Model Download and Loading in Workflows": 61bc1fa6 - "VACE Text-to-Video Workflow": 0275c575 - "VACE Image-to-Video Workflow": 6dbc84f2 - "VACE Video-to-Video Workflow": 0b58d1d2 - "VACE Video Outpainting Workflow": 22da2682 - "VACE First-Last Frame Video Generation": 0f2de66b + "_intro": 4e003b58 + "About VACE": 3d2b2490 + "Model Download and Loading in Workflows": 0a6b52f4 + "1. VACE Text-to-Video": 1715a0ab + "2. VACE Image-to-Video": 0ade4d39 + "3. VACE Video-to-Video": 4642a1ae + "4. VACE Inpainting": 6535794d + "5. VACE Video Outpainting": 24f1f2ff + "6. VACE First-Last Frame Video Generation": e58f2ea1 --- + import CancelBypass from '/snippets/zh/interface/cancel-bypass.mdx' import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -59,19 +61,19 @@ VACE 14B 是阿里通义万相团队推出的开源视频编辑统一模型。 ### 模型下载 **diffusion_models** -[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) -[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) +[wan2.1_vace_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_14B_fp16.safetensors) +[wan2.1_vace_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_vace_1.3B_fp16.safetensors) 如果你之前使用过 Wan Video 相关的工作流，下面的模型文件你已经下载过了。 **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) 从**Text encoders** 选择一个版本进行下载 -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/wan-alpha.mdx b/zh/tutorials/video/wan/wan-alpha.mdx index c6ae3abd2..deb0828d1 100644 --- a/zh/tutorials/video/wan/wan-alpha.mdx +++ b/zh/tutorials/video/wan/wan-alpha.mdx @@ -2,10 +2,17 @@ title: "Wan-Alpha 教程" description: "学习如何在 ComfyUI 中使用 Wan-Alpha 生成带有 Alpha 通道透明度的视频" sidebarTitle: "Wan-Alpha" -translationSourceHash: 82295242 +translationSourceHash: 775fde1b translationFrom: tutorials/video/wan/wan-alpha.mdx +translationBlockHashes: + "_intro": a57abe06 + "Resources": afef8889 + "Wan-Alpha Text-to-Video Workflow (14B)": 32a5a973 --- + + +import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; Wan-Alpha 是一个专门的文本生成视频模型,可以生成带有 Alpha 通道透明度的高质量视频。它基于 Wan2.1-14B-T2V 基础模型构建,能够创建具有透明背景和半透明物体的视频,非常适合合成工作流程。 该模型擅长生成透明背景、半透明物体(气泡、玻璃、水)、发光效果以及具有适当 Alpha 通道的精细细节(头发、烟雾、粒子)。 diff --git a/zh/tutorials/video/wan/wan-ati.mdx b/zh/tutorials/video/wan/wan-ati.mdx index 240fdb0fd..c62b965ed 100644 --- a/zh/tutorials/video/wan/wan-ati.mdx +++ b/zh/tutorials/video/wan/wan-ati.mdx @@ -2,15 +2,16 @@ title: "Wan ATI ComfyUI 原生工作流教程" description: "使用轨迹控制视频生成。" sidebarTitle: "Wan2.1 ATI" -translationSourceHash: a8716511 +translationSourceHash: 39439781 translationFrom: tutorials/video/wan/wan-ati.mdx translationBlockHashes: - "_intro": 247f391d + "_intro": 52f7de29 "Key Features": 67ce28af - "WAN ATI Trajectory Control Workflow Example": 01ea7d04 + "WAN ATI Trajectory Control Workflow Example": f704f2c5 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -47,17 +48,17 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 如果你没有成功下载工作流中的模型文件，可以尝试使用下面的链接手动下载 **Diffusion Model** -- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) +- [Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1-I2V-ATI-14B_fp8_e4m3fn.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **Text encoders** Chose one of following model -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) File save location diff --git a/zh/tutorials/video/wan/wan-causal-forcing.mdx b/zh/tutorials/video/wan/wan-causal-forcing.mdx index 62324a6a1..f45014dbb 100644 --- a/zh/tutorials/video/wan/wan-causal-forcing.mdx +++ b/zh/tutorials/video/wan/wan-causal-forcing.mdx @@ -2,8 +2,17 @@ title: "Causal Forcing 图生视频 ComfyUI 工作流示例" description: "使用 Wan2.1 的 Causal Forcing 或 Causal Forcing++ 从图片生成视频——只需 1 步推理即可获得流畅、时序一致的视频。" sidebarTitle: "Causal Forcing I2V" +translationSourceHash: 4f300f64 +translationFrom: tutorials/video/wan/wan-causal-forcing.mdx +translationBlockHashes: + "_intro": dbdb8d89 + "How it works": ea3eaaf0 + "Using the workflow": 5e6245a2 + "Steps to run": 3202924b + "Model downloads": c6ce079e --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Causal Forcing（因果强制）** 是一种视频生成技术，在推理过程中应用 **循环条件控制**：每一帧生成后都会作为输入反馈给模型，用于预测下一帧。通过这种方式，仅需 **1 到 4 步推理**，就能从单张起始图片生成流畅、时序一致的视频。 @@ -83,10 +92,10 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### Wan2.1 I2V - + wan2.1_i2v_480p_14B_fp16.safetensors — Wan2.1 I2V 14B 检查点 - + wan2.1_t2v_1.3B_fp16.safetensors — Wan2.1 1.3B 检查点（最低 8GB 显存） @@ -94,10 +103,10 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### CLIP 和 VAE - + google-bert/bert-base-uncased — CLIP 文本编码器 - + Wan2.1_VAE_bf16.safetensors — Wan2.1 VAE diff --git a/zh/tutorials/video/wan/wan-dancer.mdx b/zh/tutorials/video/wan/wan-dancer.mdx index 21ae06fa7..7aa876cb3 100644 --- a/zh/tutorials/video/wan/wan-dancer.mdx +++ b/zh/tutorials/video/wan/wan-dancer.mdx @@ -2,16 +2,17 @@ title: "Wan Dancer：从音乐生成舞蹈视频" description: "使用基于 Wan 2.2 构建的分层音频驱动框架 Wan Dancer，从音乐生成分钟级连贯的舞蹈视频。输入参考图像和音频，即可生成同步的舞蹈视频。" sidebarTitle: "Wan Dancer" -translationSourceHash: ea7690e9 +translationSourceHash: 73c80f3f translationFrom: tutorials/video/wan/wan-dancer.mdx translationBlockHashes: "_intro": c890f065 - "Model Highlights": 06703fea + "Model Highlights": 0826f4bf "Workflow Overview": 0fa086f1 - "Wan Dancer Workflow": fbc54f05 - "Model Information": 97e02e88 + "Wan Dancer Workflow": 2251f138 + "Model Information": bc2b5961 "Report Issues": 271203b5 --- + import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" Wan Dancer 是一款基于 Wan万相 2.2 架构的音频驱动舞蹈视频生成模型。它采用分层框架，包含全局专家模型和局部专家模型，能够生成与输入音乐同步的连贯、富有表现力的舞蹈视频。 @@ -61,20 +62,20 @@ Wan万相 Dancer 工作流接受两个输入：角色参考图像和一个音频 ### 3. 手动下载模型 **扩散模型** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **文本编码器** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP 视觉** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan-flf.mdx b/zh/tutorials/video/wan/wan-flf.mdx index a624005e3..df72b9d14 100644 --- a/zh/tutorials/video/wan/wan-flf.mdx +++ b/zh/tutorials/video/wan/wan-flf.mdx @@ -2,7 +2,7 @@ title: "ComfyUI Wan2.1 FLF2V 原生示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.1 FLF2V 视频生成示例" sidebarTitle: "Wan2.1 FLF2V 首尾帧" -translationSourceHash: 9b049266 +translationSourceHash: 5b7a25cb translationFrom: tutorials/video/wan/wan-flf.mdx --- @@ -56,7 +56,7 @@ Wan FLF2V（首尾帧视频生成）是由阿里通义万相团队推出的开 **diffusion_models** 根据你的硬件情况选择一个版本进行下载，FP8 版本对显存要求低一些 -- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) +- FP16:[wan2.1_flf2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors?download=true) - FP8:[wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn.safetensors) @@ -64,14 +64,14 @@ Wan FLF2V（首尾帧视频生成）是由阿里通义万相团队推出的开 从**Text encoders** 选择一个版本进行下载， -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 ``` diff --git a/zh/tutorials/video/wan/wan-move.mdx b/zh/tutorials/video/wan/wan-move.mdx index 93a027390..dd2c1945d 100644 --- a/zh/tutorials/video/wan/wan-move.mdx +++ b/zh/tutorials/video/wan/wan-move.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan-Move 工作流示例" description: "Wan-Move 是一个通过潜在轨迹引导实现运动可控视频生成的模型，支持图像到视频生成的精细点级运动控制。" sidebarTitle: "Wan-Move" -translationSourceHash: 1b11ad45 +translationSourceHash: dda86ac3 translationFrom: tutorials/video/wan/wan-move.mdx translationBlockHashes: "_intro": 67a26359 - "Wan-Move image-to-video workflow": ec83b0e4 - "Model links": 25e1a28d + "Wan-Move image-to-video workflow": b77082ab + "Model links": 8e0ef166 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Wan-Move** 是阿里巴巴通义实验室开发的运动可控视频生成框架。它允许用户通过在输入图像上指定点轨迹来控制生成视频中的物体运动，使图生视频更加精确可控。 @@ -28,13 +29,49 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Wan-Move 图生视频工作流 - -

下载 JSON 工作流文件

- - - -

在 ComfyUI Cloud 上运行

- + + Preview the workflow output: +![Wan-Move preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_wanmove_480p-1.webp) + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Wan-Move Motion-Control" in Template Library + + +### Input Image + + + Download the default input image, or use your own image as the starting frame. + + + +## Model links + + + Place in ComfyUI/models/diffusion_models/ + + + Place in ComfyUI/models/loras/ + + + Place in ComfyUI/models/text_encoders/ + + + Place in ComfyUI/models/clip_vision/ + + + Place in ComfyUI/models/vae/ + + + + + + + Download the default input image, or use your own image as the starting frame. + + @@ -42,23 +79,23 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **diffusion_models** -- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) +- [Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/WanMove/Wan21-WanMove_fp8_scaled_e4m3fn_KJ.safetensors) **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **模型存放位置** diff --git a/zh/tutorials/video/wan/wan-video.mdx b/zh/tutorials/video/wan/wan-video.mdx index 5ee12bf95..d3ade7c5e 100644 --- a/zh/tutorials/video/wan/wan-video.mdx +++ b/zh/tutorials/video/wan/wan-video.mdx @@ -2,17 +2,18 @@ title: ComfyUI Wan2.1 Video 示例 description: "本文介绍了如何在 ComfyUI 中完成 Wan2.1 Video 视频首尾帧视频生成示例" sidebarTitle: Wan2.1 -translationSourceHash: e0c0f714 +translationSourceHash: c11f515a translationFrom: tutorials/video/wan/wan-video.mdx translationBlockHashes: - "_intro": 228a7575 + "_intro": 376d341e "Wan2.1 ComfyUI Native Workflow Examples": b092012f - "Model Installation": 0bd3e4c2 - "Wan2.1 Text-to-Video Workflow": 611b9bf9 - "Wan2.1 Image-to-Video Workflow": bee5fc35 + "Model Installation": 21d2cbd6 + "Wan2.1 Text-to-Video Workflow (1.3B)": 268e4e8b + "Wan2.1 Image-to-Video Workflow (14B)": dfb19590 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' Wan2.1 Video 系列为阿里巴巴于 2025年2月开源的视频生成模型，其开源协议为 [Apache 2.0](https://github.com/Wan-Video/Wan2.1?tab=Apache-2.0-1-ov-file)，提供 14B（140亿参数）和 1.3B（13亿参数）两个版本，覆盖文生视频（T2V）、图生视频（I2V）等多项任务。 @@ -34,14 +35,14 @@ Wan2.1 Video 系列为阿里巴巴于 2025年2月开源的视频生成模型， 本篇指南涉及的所有模型你都可以在[这里](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files)找到, 下面是本篇示例中将会使用到的共用的模型，你可以提前进行下载： 从**Text encoders** 选择一个版本进行下载， -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors?download=true) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors?download=true) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors?download=true) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors?download=true) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) 文件保存位置 @@ -64,7 +65,7 @@ ComfyUI/ ## Wan2.1 文生视频工作流 -在开始工作流前请下载 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下。 +在开始工作流前请下载 [wan2.1_t2v_1.3B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下。 > 如果你需要其它的 t2v 精度版本，请访问[这里](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models)进行下载 @@ -101,7 +102,7 @@ ComfyUI/ ![Wan2.1 图生视频工作流 14B 480P Workflow 输入图片示例](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1/input/flux_dev_example.png) #### 2. 模型下载 -请下载[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 +请下载[wan2.1_i2v_480p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_480p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 #### 3. 按步骤完成工作流的运行 @@ -129,7 +130,7 @@ ComfyUI/ #### 2. 模型下载 -请下载[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 +请下载[wan2.1_i2v_720p_14B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_i2v_720p_14B_fp16.safetensors?download=true)，并保存到 `ComfyUI/models/diffusion_models/` 目录下 #### 3. 按步骤完成工作流的运行 diff --git a/zh/tutorials/video/wan/wan2-2-animate.mdx b/zh/tutorials/video/wan/wan2-2-animate.mdx index 23407f486..a6495ee7d 100644 --- a/zh/tutorials/video/wan/wan2-2-animate.mdx +++ b/zh/tutorials/video/wan/wan2-2-animate.mdx @@ -2,16 +2,16 @@ title: "Wan2.2 Animate ComfyUI 原生工作流" description: "统一的人物动画和替换框架，具有精确的运动和表情复制。" sidebarTitle: "Wan2.2 Animate" -translationSourceHash: 8d8cb435 +translationSourceHash: 07849135 translationFrom: tutorials/video/wan/wan2-2-animate.mdx translationBlockHashes: "_intro": 96eb0424 "Model Highlights": fc4e7349 "ComfyOrg Wan2.2 Animate stream replay": a43ba31b - "About Wan2.2 Animate workflow": f3713999 - "Wan2.2 Anmate ComfyUI native workflow(without custom nodes)": fc4b2e18 + "About Wan2.2 Animate workflow": ea4c1d9a --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架。 @@ -54,13 +54,49 @@ Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架 下载以下工作流文件并将其拖入 ComfyUI 以加载工作流。 - -

下载工作流

- - - -

在 Comfy 云上运行

- + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Wan2.2 Animate" in Template Library + + + + + All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files). + +**Diffusion Models** + + + + Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors: Scaled FP8 version from Kijai's repo + + + wan2.2_animate_14B_bf16.safetensors: Original bf16 model weight + + + +**CLIP Vision** + + + clip_vision_h.safetensors: CLIP Vision encoder + + +**LoRAs** + + + lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4-step acceleration LoRA + +**VAE** + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding + +**Text Encoders** + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + + 下载以下素材作为输入： @@ -77,19 +113,19 @@ Wan-Animate 是由 WAN 团队开发的一个统一的人物动画和替换框架 ### 2. 模型链接 **diffusion_models** -- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/resolve/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) 这是来自 Kijai 仓库的模型 -- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 原始模型权重 +- [Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/Wan22Animate/Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors) 这是来自 Kijai 仓库的模型 +- [wan2.2_animate_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_animate_14B_bf16.safetensors) 原始模型权重 **clip_visions** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 这是一个 4 步的加速 lora +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) 这是一个 4 步的加速 lora **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan2-2-fun-camera.mdx b/zh/tutorials/video/wan/wan2-2-fun-camera.mdx index 8702967cb..d3fdc93c4 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -2,7 +2,7 @@ title: "ComfyUI Wan2.2 Fun Camera Control 相机控制视频生成工作流示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.2 Fun Camera Control 使用相机控制来完成视频生成的示例" sidebarTitle: "Wan2.2 Fun Camera" -translationSourceHash: c2df0664 +translationSourceHash: d7f3bd26 translationFrom: tutorials/video/wan/wan2-2-fun-camera.mdx --- @@ -48,9 +48,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_camera/wan2.2_14B_fun_camera.mp4" > - -

下载 JSON 格式工作流

- + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Wan2.2 Fun Camera" in Template Library + + 请下载下面的图片，我们将作为输入。 @@ -61,18 +66,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面的模型你可以在 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 找到 **Diffusion Model** -- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_camera_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_camera_low_noise_14B_fp8_scaled.safetensors) **Wan2.2-Lightning LoRA (可选，用于加速)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) File save location diff --git a/zh/tutorials/video/wan/wan2-2-fun-control.mdx b/zh/tutorials/video/wan/wan2-2-fun-control.mdx index 1447efb01..d738bc2ba 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-control.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan2.2 Fun Control 视频控制生成示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.2 Fun Control 使用控制视频来完成视频生成的示例" sidebarTitle: "Wan2.2 Fun Control" -translationSourceHash: 59b97fdc +translationSourceHash: f1c35d6e translationFrom: tutorials/video/wan/wan2-2-fun-control.mdx translationBlockHashes: "_intro": 239e24b6 "ComfyOrg Wan2.2 Fun InP & Control Youtube Live Stream Replay": 22557a2e - "Wan2.2 Fun Control Video Generation Workflow Example": 68f53061 + "Wan2.2 Fun Control Video Generation Workflow Example": cbbb7456 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Wan2.2-Fun-Control** 是 Alibaba PAI 团队推出的新一代视频生成与控制模型，通过引入创新性的控制代码（Control Codes）机制，结合深度学习和多模态条件输入，能够生成高质量且符合预设控制条件的视频。该模型采用 **Apache 2.0 许可协议**发布，支持商业使用。 @@ -66,9 +67,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/wan2.2_14B_fun_inp.mp4" > - -

下载 JSON 格式工作流

- + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Wan2.2 Fun Control" in Template Library + + 请下载下面的图片及视频，我们将作为输入。 @@ -87,18 +93,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下面的模型你可以在 [Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 找到 **Diffusion Model** -- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_control_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_control_low_noise_14B_fp8_scaled.safetensors) ** Wan2.2-Lightning LoRA (可选，用于加速)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) File save location diff --git a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx index 50dd30fa8..80cbea3c4 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -2,15 +2,16 @@ title: "ComfyUI Wan2.2 Fun Inp 首尾帧视频生成示例" description: "本文介绍了如何在 ComfyUI 中完成 Wan2.2 Fun Inp 首尾帧视频生成示例" sidebarTitle: "Wan2.2 Fun Inp" -translationSourceHash: 9d81688f +translationSourceHash: 714331ba translationFrom: tutorials/video/wan/wan2-2-fun-inp.mdx translationBlockHashes: - "_intro": 75061370 + "_intro": 7a3595d7 "ComfyOrg Wan2.2 Fun InP & Control Youtube Live Stream Replay": 22557a2e - "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 39f39063 + "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 692b42a8 --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Wan2.2-Fun-Inp** 是 Alibaba pai团队推出的首尾帧控制视频生成模型，支持输入**首帧和尾帧图像**，生成中间过渡视频，为创作者带来更强的创意控制力。该模型采用 **Apache 2.0 许可协议**发布，支持商业使用。 @@ -61,13 +62,60 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 或者，在更新 ComfyUI 至最新版本后，下载下面的工作流并拖入 ComfyUI 中加载。 - -

下载 JSON 格式工作流

- - -

在 Comfy 云上运行

- + + Wan2.2 Fun Inp workflow + +Wan2.2 Fun Inp workflow + + + Download JSON or search "Wan2.2 Fun Inp" in Template Library + + + 在 Comfy Cloud 中打开 + + +### Input materials + + + + Start frame for video generation. Download and use this image, or replace with your own. + + + End frame for video generation. Download and use this image, or replace with your own. + + +All models involved in this guide can be found [here](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files). + +**Diffusion Models** + + + + wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors: High noise diffusion model for start-end frame inpainting + + + wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors: Low noise diffusion model for start-end frame inpainting + + + + + + wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors: 4-step acceleration LoRA for high noise model + + + wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors: 4-step acceleration LoRA for low noise model + + + + wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding + + +**Text Encoder** + + + umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder + + 使用下面的素材作为首尾帧 @@ -77,18 +125,18 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ### 2. 手动下载模型 **Diffusion Model** -- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) -- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_inpaint_high_noise_14B_fp8_scaled.safetensors) +- [wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_fun_inpaint_low_noise_14B_fp8_scaled.safetensors) **Lightning LoRA (可选，用于加速)** -- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) -- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors) +- [wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/loras/wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors) **VAE** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **Text Encoder** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` ComfyUI/ diff --git a/zh/tutorials/video/wan/wan2-2-s2v.mdx b/zh/tutorials/video/wan/wan2-2-s2v.mdx index 447201527..47076544f 100644 --- a/zh/tutorials/video/wan/wan2-2-s2v.mdx +++ b/zh/tutorials/video/wan/wan2-2-s2v.mdx @@ -2,7 +2,7 @@ title: Wan2.2-S2V 音频驱动视频生成 ComfyUI 原生工作流示例 description: 这是一个基于 ComfyUI 的 Wan2.2-S2V 音频驱动视频生成原生工作流示例。 sidebarTitle: "Wan2.2 S2V" -translationSourceHash: 8d1b0bf2 +translationSourceHash: 10604cda translationFrom: tutorials/video/wan/wan2-2-s2v.mdx --- @@ -33,37 +33,54 @@ Wan2.2 S2V 模型仓库：[Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > - -

Download JSON Workflow

- - - -

Run on Comfy Cloud

- + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Wan2.2 S2V" in Template Library + + + + + + Download the default input image, or use your own image. + + + Download the default input audio, or use your own audio. + + 下载下面的图片及音频作为输入: ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - -

下载输入音频

- + + + + FP8 scaled diffusion model + + + BF16 diffusion model + + + ### 2. 模型链接 你可以在 [我们的仓库](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 中找到所有模型。 **diffusion_models** -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) **audio_encoders** -- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) +- [wav2vec2_large_english_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/audio_encoders/wav2vec2_large_english_fp16.safetensors) **vae** -- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) +- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) **text_encoders** -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) ``` @@ -92,8 +109,8 @@ ComfyUI/ 你可以在 [这里](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) 找到两种模型： -- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) -- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) +- [wan2.2_s2v_14B_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_fp8_scaled.safetensors) +- [wan2.2_s2v_14B_bf16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/wan2.2_s2v_14B_bf16.safetensors) 本模板使用 `wan2.2_s2v_14B_fp8_scaled.safetensors`，它需要更少的显存。但你可以尝试 `wan2.2_s2v_14B_bf16.safetensors` 来减少质量损失。 diff --git a/zh/tutorials/video/wan/wan2_2.mdx b/zh/tutorials/video/wan/wan2_2.mdx index 3c2588525..44406d6cb 100644 --- a/zh/tutorials/video/wan/wan2_2.mdx +++ b/zh/tutorials/video/wan/wan2_2.mdx @@ -2,21 +2,22 @@ title: "Wan2.2 视频生成ComfyUI 官方原生工作流示例" description: "阿里云通义万相2.2视频生成模型在ComfyUI中的官方使用指南" sidebarTitle: Wan2.2 -translationSourceHash: 70e49209 +translationSourceHash: a75c822c translationFrom: tutorials/video/wan/wan2_2.mdx translationBlockHashes: "_intro": 3d02f6d3 "Model Highlights": fc7121c1 "Wan2.2 Open Source Model Versions": 7ed2d913 "ComfyOrg Wan2.2 Live Streams": d380a603 - "Wan2.2 TI2V 5B Hybrid Version Workflow Example": b0e108bd - "Wan2.2 14B T2V Text-to-Video Workflow Example": 9eaf9107 - "Wan2.2 14B I2V Image-to-Video Workflow Example": a006f8aa - "Wan2.2 14B FLF2V Workflow Example": 525e0946 + "Wan2.2 TI2V 5B Hybrid Version Workflow Example": a9d53e08 + "Wan2.2 14B T2V Text-to-Video Workflow Example": 6a6398c7 + "Wan2.2 14B I2V Image-to-Video Workflow Example": 356d6b7a + "Wan2.2 14B FLF2V Workflow Example": ca3b4239 "Community Resources": 7463b48b --- + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' -[ワークフローをダウンロード](https://github.com/Comfy-Org/workflows/blob/main/tutorial_workflows/Get_Comfy_With_Comfy_Wan_Alpha.json) - ## リソース - [Wan-Alpha GitHub](https://github.com/WeChatCV/Wan-Alpha) - [Hugging Face モデル](https://huggingface.co/htdong/Wan-Alpha) - [ComfyUI 版](https://huggingface.co/htdong/Wan-Alpha_ComfyUI) - [研究論文](https://arxiv.org/pdf/2509.24979) + + + +## Wan-Alpha テキストから動画ワークフロー (14B) + +### 1. ワークフローをダウンロード + +ComfyUI を最新版に更新し、ワークフローファイルをダウンロードして ComfyUI にドラッグするか、テンプレートライブラリの `Workflow` → `Browse Templates` → `Video` で "Wan 2.1 Alpha T2V 14B" を検索してください。 + +Wan-Alpha T2V ワークフロー + + + + Comfy Cloud で開く + + + JSON をダウンロード、またはテンプレートライブラリで "Wan 2.1 Alpha T2V 14B" を検索 + + + +### 2. モデルをインストール + +関連するすべてのモデルは [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) で入手できます。 + +**拡散モデル**: いずれかのバージョンを選択： + + + + FP8 スケール済み拡散モデル。ComfyUI/models/diffusion_models/ に保存 + + + BF16 拡散モデル。ComfyUI/models/diffusion_models/ に保存 + + + +**テキストエンコーダー** + + + + FP8 テキストエンコーダー。ComfyUI/models/text_encoders/ に保存 + + + +**VAE モデル** + + + + RGB チャンネル VAE。ComfyUI/models/vae/ に保存 + + + アルファチャンネル VAE。ComfyUI/models/vae/ に保存 + + + +**LoRA モデル** + + + + アルファ生成 LoRA。ComfyUI/models/loras/ に保存 + + + 高速推論用 Lightning LoRA。ComfyUI/models/loras/ に保存 + + + +ファイル保存場所： + +``` +ComfyUI/ +├── models/ +│ ├── diffusion_models/ +│ │ ├── wan2.1_t2v_14B_fp8_scaled.safetensors +│ │ └── wan2.1_t2v_14B_bf16.safetensors +│ ├── text_encoders/ +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── vae/ +│ │ ├── wan_alpha_2.1_vae_rgb_channel.safetensors +│ │ └── wan_alpha_2.1_vae_alpha_channel.safetensors +│ └── loras/ +│ ├── wan_alpha_2.1_rgba_lora.safetensors +│ └── lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors +``` + +### 3. ワークフローをステップごとに完了 + +1. `Load Diffusion Model` ノードに正しい拡散モデルが読み込まれていることを確認 +2. `Load CLIP` ノードに `umt5_xxl_fp8_e4m3fn_scaled.safetensors` が読み込まれていることを確認 +3. RGB 用の `Load VAE` ノードに `wan_alpha_2.1_vae_rgb_channel.safetensors` が読み込まれていることを確認 +4. アルファ用の 2 つ目の `Load VAE` ノードに `wan_alpha_2.1_vae_alpha_channel.safetensors` が読み込まれていることを確認 +5. `LoRA Loader` ノードに `wan_alpha_2.1_rgba_lora.safetensors` が読み込まれていることを確認 +6. （オプション）Lightning LoRA ノードに `lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors` を読み込んで高速生成 +7. `CLIP Text Encode (Positive Prompt)` ノードでポジティブプロンプトを設定 +8. （オプション）`EmptyHunyuanLatentVideo` ノードで動画サイズを変更 +9. `Run` ボタンをクリックするか `Ctrl(Cmd) + Enter` で生成を実行 diff --git a/ja/tutorials/video/wan/wan-causal-forcing.mdx b/ja/tutorials/video/wan/wan-causal-forcing.mdx index 3ec115972..74008a7ce 100644 --- a/ja/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ja/tutorials/video/wan/wan-causal-forcing.mdx @@ -43,7 +43,6 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' このワークフローはサブグラフノードを使用してモジュール化された処理を行います。サブグラフのドキュメントを参照して、ワークフローをカスタマイズおよび拡張する方法を学んでください。 - ### Causal Forcing と Causal Forcing++ | モード | 説明 | diff --git a/ja/tutorials/video/wan/wan2_2.mdx b/ja/tutorials/video/wan/wan2_2.mdx index ed58201ae..b75e586ba 100644 --- a/ja/tutorials/video/wan/wan2_2.mdx +++ b/ja/tutorials/video/wan/wan2_2.mdx @@ -107,7 +107,7 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows Download JSON or search "Wan2.2 5B" in Template Library - + Comfy Cloud で開く @@ -125,7 +125,7 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows Download JSON or search "Wan2.2 5B" in Template Library - + Comfy Cloud で開く @@ -181,7 +181,7 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows Download JSON or search "Wan2.2 14B T2V" in Template Library - + Comfy Cloud で開く @@ -202,7 +202,7 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows Download JSON or search "Wan2.2 5B" in Template Library - + Comfy Cloud で開く @@ -260,7 +260,7 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows Download JSON or search "Wan2.2 14B I2V" in Template Library - + Comfy Cloud で開く @@ -281,7 +281,7 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows Download JSON or search "Wan2.2 5B" in Template Library - + Comfy Cloud で開く @@ -341,7 +341,7 @@ ComfyUI/ Download JSON or search "Wan2.2 14B FLF2V" in Template Library - + Comfy Cloud で開く @@ -350,7 +350,7 @@ ComfyUI/ Download JSON or search "Wan2.2 5B" in Template Library - + Comfy Cloud で開く diff --git a/ja/tutorials/video/zai/scail2.mdx b/ja/tutorials/video/zai/scail2.mdx index ee2be2d4e..515be3201 100644 --- a/ja/tutorials/video/zai/scail2.mdx +++ b/ja/tutorials/video/zai/scail2.mdx @@ -54,7 +54,6 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' このワークフローはモジュール処理に Subgraph ノードを使用します。ワークフローのカスタマイズと拡張については、サブグラフのドキュメントをご覧ください。 - ### 長尺ビデオ 長いビデオの場合は、セグメント数を計算します：`ceil(total_frames / 76)`。最初のセグメント以外はすべて Extend サブグラフを使用します。さらにセグメントを追加するには Extend ノードを複製し、`previous_frames` 出力を連鎖させ、`segment_index` を増やします。 diff --git a/ko/tutorials/3d/hunyuan3D-2.mdx b/ko/tutorials/3d/hunyuan3D-2.mdx index 906e29e40..d67fada7b 100644 --- a/ko/tutorials/3d/hunyuan3D-2.mdx +++ b/ko/tutorials/3d/hunyuan3D-2.mdx @@ -59,7 +59,7 @@ Hunyuan3D-2mv 워크플로우에서는 다중뷰 이미지를 사용해 3D 모 - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 + 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 워크플로우 JSON 파일 다운로드 @@ -107,7 +107,7 @@ ComfyUI/ Hunyuan3D-2mv-turbo 워크플로우에서는 Hunyuan3D-2mv-turbo 모델을 사용해 3D 모델을 생성합니다. 이 모델은 Hunyuan3D-2mv의 단계 증류 버전으로, 더 빠른 3D 모델 생성을 가능하게 합니다. 이번 버전의 워크플로우에서는 `cfg`를 1.0으로 설정하고, `flux guidance` 노드를 추가해 `증류된 cfg` 생성을 제어합니다. - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 + 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 워크플로우 JSON 파일 다운로드 @@ -150,7 +150,7 @@ ComfyUI/ Hunyuan3D-2 워크플로우에서는 Hunyuan3D-2 모델을 사용해 3D 모델을 생성합니다. 이 모델은 다중뷰 모델이 아닙니다. 이번 워크플로우에서는 `Hunyuan3Dv2ConditioningMultiView` 노드 대신 `Hunyuan3Dv2Conditioning` 노드를 사용합니다. - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 + 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 워크플로우 JSON 파일 다운로드 diff --git a/ko/tutorials/3d/triposplat.mdx b/ko/tutorials/3d/triposplat.mdx index af537eef4..26224ebc6 100644 --- a/ko/tutorials/3d/triposplat.mdx +++ b/ko/tutorials/3d/triposplat.mdx @@ -2,7 +2,7 @@ title: "TripoSplat 이미지에서 가우시안 스플래트로의 ComfyUI 워크플로우 예시" description: "TripoSplat을 사용해 단일 2D 이미지로부터 고품질 3D 가우시안 스플래트 표현을 생성하며, 렌더링을 위한 밀도와 예산을 제어할 수 있습니다." sidebarTitle: "TripoSplat" -translationSourceHash: 95c7decf +translationSourceHash: 0cba7aeb translationFrom: tutorials/3d/triposplat.mdx translationBlockHashes: "_intro": 77f8628a @@ -10,11 +10,10 @@ translationBlockHashes: "Workflow node guide": 83cd15a0 "Steps to run": e4ad7fa4 "Output options": 9a1cc408 - "Model downloads": c6ba2321 + "Model downloads": 44894623 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" **TripoSplat**는 단일 2D 이미지에서 직접 **3D 가우시안 스플래트** 표현을 생성하는 오픈소스 모델입니다. VAST-AI가 개발했으며 오픈소스 라이선스로 공개되었습니다. @@ -26,23 +25,108 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 + + 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 + - + JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "TripoSplat" 검색하세요 + +## 작동 방식 + +TripoSplat은 단일 RGB 이미지를 입력으로 받아 3D 가우시안 프리미티브 세트를 직접 예측하는 **피드포워드 아키텍처**를 사용합니다. 파이프라인은 다음과 같은 단계로 이루어집니다: + +1. **이미지 인코딩**: 입력 이미지는 비전 인코더(DINOv2)로 처리됩니다. +2. **삼평면 생성**: 특징들이 삼평면 표현으로 디코드됩니다. +3. **가우시안 예측**: 삼평면을 샘플링해 가우시안 매개변수(위치, 스케일, 회전, 투명도, 색상)를 생성합니다. +4. **렌더링**: 가우시안을 다양한 시점에서 미분 가능한 스플래팅을 이용해 렌더링합니다. + + + 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 알아보세요. + +## 워크플로우 노드 안내 + +### LoadImage +- 입력 이미지(PNG/JPG)를 로드합니다. +- 샘플 이미지: `white-hotel-on-rocky-island.png` (템플릿 라이브러리에서 사용 가능) + +### TripoSplat (서브그래프) + +주요 서브그래프 노드는 이미지를 처리해 3D 가우시안 스플래트를 생성합니다. 노출된 매개변수: + +| 매개변수 | 기본값 | 설명 | +|---|---|---| +| `switch` | — | 서브그래프 활성화/비활성화 | +| `num_gaussians` | — | 생성할 가우시안 프리미티브의 수 (품질/성능 제어) | +| `seed` | — | 재현성을 위한 난수 시드 | +| `unet_name` | — | TripoSplat 확산 모델 체크포인트 | +| `clip_name` | — | CLIP 비전 인코더 모델 | +| `vae_name` | — | VAE 인코딩/디코딩용 (메인 VAE와 인코더용 두 개 항목) | +| `bg_removal_name` | — | 배경 제거 모델 | + +### CreateCameraInfo +- 결과를 렌더링하기 위한 카메라 궤도를 정의합니다. +- 매개변수: 궤도 유형, 각도, 거리, 시야각 등. +- 기본값: 35° 고도, 30 거리, 2.5 줌 + +### RenderSplat +- 정의된 카메라 각도에서 가우시안 스플래트를 2D 이미지로 렌더링합니다. +- 매개변수: 출력 해상도(기본 1024×1024), 이미지 품질 설정 + +### SplatToMesh +- 가우시안 스플래트를 메쉬로 변환합니다(선택사항). +- 매개변수: 메쉬 밀도, 평탄화, 간소화 + +### SaveGLB +- 결과를 GLB 3D 파일로 저장합니다. + +### SaveVideo +- 렌더링된 3D 장면의 동영상을 저장합니다. + +### SplatToFile3D +- 가우시안 스플래트를 SPZ 형식으로 보냅니다. + +### CreateVideo +- 렌더링된 프레임들로 동영상을 생성합니다. + +## 실행 단계 + +1. **이미지 로드**: **LoadImage** 노드를 사용해 단일 2D 이미지를 로드하세요. +2. **TripoSplat 서브그래프 실행**: 모델이 가우시안 스플래트 표현을 생성합니다. +3. **출력 형식 선택**: GLB, SPZ, 동영상으로 보내거나 메쉬로 렌더링하세요. +4. **결과 보기**: 생성된 3D 파일이나 렌더링된 미리보기를 활용하세요. + +## 출력 옵션 + +| 노드 | 형식 | 활용 사례 | +|---|---|---| +| **SaveGLB** | `.glb` | 표준 3D 파일 형식, 3D 소프트웨어로 불러올 수 있음 | +| **SplatToFile3D** | `.spz` | 압축된 가우시안 스플래트 형식, 효율적 저장용 | +| **RenderSplat** | 2D 이미지 | 임의 각도에서의 결과 미리보기 | +| **SplatToMesh** | 메쉬 | 전통적인 메쉬로 변환해 추가 편집 가능 | + +## 모델 다운로드 + +TripoSplat 모델과 필요한 파일을 다운로드하세요. 해당 `models/` 하위 디렉토리에 배치하세요. + + + + triposplat_fp16.safetensors: TripoSplat 확산 모델 체크포인트 + + + triposplat_vae_decoder_fp16.safetensors: VAE 디코더 + - flux2-vae.safetensors — Flux.2 VAE, 잠재적 인코딩용 + flux2-vae.safetensors: Flux.2 VAE, 잠재적 인코딩용 - - dino_v3_vit_h.safetensors — CLIP 비전 인코더 (DINOv2) + dino_v3_vit_h.safetensors: CLIP 비전 인코더 (DINOv2) - - birefnet.safetensors — 전처리용 배경 제거 모델 + birefnet.safetensors: 전처리용 배경 제거 모델 - + ### 모델 저장 위치 ``` @@ -57,3 +141,4 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" │ │ └── dino_v3_vit_h.safetensors │ └── 📂 background_removal/ │ └── birefnet.safetensors +``` diff --git a/ko/tutorials/flux/flux-1-controlnet.mdx b/ko/tutorials/flux/flux-1-controlnet.mdx index f1db957d5..e3d0d9c95 100644 --- a/ko/tutorials/flux/flux-1-controlnet.mdx +++ b/ko/tutorials/flux/flux-1-controlnet.mdx @@ -52,7 +52,7 @@ Depth 버전은 깊이 맵 추출 기법을 통해 원본 이미지의 공간적 JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Canny" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 ### 1. 워크플로우 및 자산 @@ -120,7 +120,7 @@ ComfyUI/ JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Depth LoRA" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 LoRA 버전 워크플로우는 완전한 버전을 기반으로 LoRA 모델을 추가한 것입니다. [Flux 워크플로우의 전체 버전](/ko/tutorials/flux/flux-1-text-to-image)과 비교해, 해당 LoRA 모델을 로드하고 사용하는 노드가 추가되었습니다. diff --git a/ko/tutorials/flux/flux-1-fill-dev.mdx b/ko/tutorials/flux/flux-1-fill-dev.mdx index 3794cfc4c..b8e20cd31 100644 --- a/ko/tutorials/flux/flux-1-fill-dev.mdx +++ b/ko/tutorials/flux/flux-1-fill-dev.mdx @@ -65,7 +65,7 @@ ComfyUI/ JSON 다운로드 또는 템플릿 라이브러리에서 "flux_fill_inpaint" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 diff --git a/ko/tutorials/flux/flux-1-kontext-dev.mdx b/ko/tutorials/flux/flux-1-kontext-dev.mdx index f8f5e98f3..24962af73 100644 --- a/ko/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ko/tutorials/flux/flux-1-kontext-dev.mdx @@ -83,7 +83,7 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 JSON 다운로드 또는 템플릿 라이브러리에서 "Flux Kontext Dev" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 이 워크플로우는 `Load Image(from output)` 노드를 사용해 편집할 이미지를 불러오므로, 여러 차례의 편집을 위해 편집된 이미지에 더욱 편리하게 접근할 수 있습니다. diff --git a/ko/tutorials/flux/flux-1-text-to-image.mdx b/ko/tutorials/flux/flux-1-text-to-image.mdx index cafd71ae4..9cb3e38c6 100644 --- a/ko/tutorials/flux/flux-1-text-to-image.mdx +++ b/ko/tutorials/flux/flux-1-text-to-image.mdx @@ -55,7 +55,7 @@ Flux는 뛰어난 이미지 품질과 유연성으로 고화질의 다양한 이 ![Flux Dev 원본 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - + 이 워크플로우를 Comfy Cloud에서 실행하세요 @@ -113,7 +113,7 @@ Flux의 뛰어난 프롬프트 추종 능력 덕분에 부정적인 프롬프트 ![Flux Schnell 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) - + 이 워크플로우를 Comfy Cloud에서 실행하세요 @@ -166,7 +166,7 @@ fp8 버전은 원본 Flux.1 fp16 버전의 양자화된 버전입니다. ### Flux.1 Dev - + 이 워크플로우를 Comfy Cloud에서 실행하세요 @@ -185,7 +185,7 @@ fp8 버전은 원본 Flux.1 fp16 버전의 양자화된 버전입니다. ### Flux.1 Schnell - + 이 워크플로우를 Comfy Cloud에서 실행하세요 diff --git a/ko/tutorials/flux/flux-1-uso.mdx b/ko/tutorials/flux/flux-1-uso.mdx index ff8cebd9b..06979597f 100644 --- a/ko/tutorials/flux/flux-1-uso.mdx +++ b/ko/tutorials/flux/flux-1-uso.mdx @@ -35,7 +35,7 @@ USO는 세 가지 주요 방식을 지원합니다: Download the workflow JSON and drag it into ComfyUI - 이 워크플로우를 Comfy Cloud에서 실행하세요 + 이 워크플로우를 Comfy Cloud에서 실행하세요 diff --git a/ko/tutorials/flux/flux-2-dev.mdx b/ko/tutorials/flux/flux-2-dev.mdx index 617406994..1696ffdb8 100644 --- a/ko/tutorials/flux/flux-2-dev.mdx +++ b/ko/tutorials/flux/flux-2-dev.mdx @@ -41,25 +41,27 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" FLUX.2 Dev를 사용하여 단일 이미지를 생성하는 기본적인 텍스트-이미지 워크플로우입니다. - + + 이 워크플로우를 Comfy Cloud에서 바로 열기 - - + 로컬에서 사용하기 위해 JSON 워크플로우 파일 다운로드 + ## 다중 이미지 참조 워크플로우 2개 이미지 참조 워크플로우 예제입니다. 이 구현을 확장해 더 많은 참조 이미지를 지원할 수 있습니다. - + + 이 워크플로우를 Comfy Cloud에서 바로 열기 - - + 로컬에서 사용하기 위해 JSON 워크플로우 파일 다운로드 + ## 모델 링크 diff --git a/ko/tutorials/flux/flux-2-klein.mdx b/ko/tutorials/flux/flux-2-klein.mdx index bacd3dde6..da2f98a4b 100644 --- a/ko/tutorials/flux/flux-2-klein.mdx +++ b/ko/tutorials/flux/flux-2-klein.mdx @@ -32,36 +32,33 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 ## Flux.2 Klein 4B 워크플로우 - + + Flux.2 Klein 4B용 텍스트 기반 이미지 생성 워크플로우를 다운로드하세요. - - + 4B 베이스 모델을 사용한 이미지 편집 워크플로우를 다운로드하세요. - - + 빠른 정제된 4B 이미지 편집 워크플로우를 다운로드하세요. - + ## Flux.2 Klein 4B 모델 다운로드 + 4B 모델용 텍스트 인코더입니다. - 확산 모델(4B 베이스). - 확산 모델(4B 정제). - 4B 모델용 VAE입니다. - + **4B 모델 저장 위치** ``` @@ -78,40 +75,37 @@ FLUX.2 [Klein]은 Flux 제품군 중 가장 빠른 모델로, 텍스트 기반 ## Flux.2 Klein 9B 워크플로우 - + + Flux.2 Klein 9B용 텍스트 기반 이미지 생성 워크플로우를 다운로드하세요. - - + 9B 베이스 모델을 사용한 이미지 편집 워크플로우를 다운로드하세요. - - + 빠른 정제된 9B 이미지 편집 워크플로우를 다운로드하세요. - + ## Flux.2 Klein 9B 모델 다운로드 확산 모델의 경우 BFL의 리포지토리를 방문해 계약을 수락한 후 모델을 다운로드하세요. + 확산 모델(9B 베이스). - 확산 모델(9B 정제). - 9B 모델용 텍스트 인코더입니다. - 9B 모델용 VAE입니다. - + **9B 모델 저장 위치** ``` diff --git a/ko/tutorials/flux/flux1-krea-dev.mdx b/ko/tutorials/flux/flux1-krea-dev.mdx index 8940a7003..a005f4cf7 100644 --- a/ko/tutorials/flux/flux1-krea-dev.mdx +++ b/ko/tutorials/flux/flux1-krea-dev.mdx @@ -32,7 +32,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ![Flux Krea Dev 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - 이 워크플로우를 Comfy Cloud에서 실행하세요 + 이 워크플로우를 Comfy Cloud에서 실행하세요 JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Krea Dev" 검색 diff --git a/ko/tutorials/image/hidream/hidream-i1.mdx b/ko/tutorials/image/hidream/hidream-i1.mdx index 64a823e2e..30484cb72 100644 --- a/ko/tutorials/image/hidream/hidream-i1.mdx +++ b/ko/tutorials/image/hidream/hidream-i1.mdx @@ -100,7 +100,7 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 풀버전 워크플로우 - + Run this workflow on Comfy Cloud with zero setup 워크플로우 JSON 파일 다운로드 @@ -141,7 +141,7 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Dev 버전 워크플로우 - + Run this workflow on Comfy Cloud with zero setup 워크플로우 JSON 파일 다운로드 @@ -180,7 +180,7 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 ### HiDream-I1 Fast 버전 워크플로우 - + Run this workflow on Comfy Cloud with zero setup 워크플로우 JSON 파일 다운로드 diff --git a/ko/tutorials/image/omnigen/omnigen2.mdx b/ko/tutorials/image/omnigen/omnigen2.mdx index 93d7e3ed6..1c5473e0b 100644 --- a/ko/tutorials/image/omnigen/omnigen2.mdx +++ b/ko/tutorials/image/omnigen/omnigen2.mdx @@ -68,7 +68,7 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 ### 1. 워크플로우 파일 다운로드 - + Open and run this workflow directly in Comfy Cloud. @@ -98,7 +98,7 @@ OmniGen2는 풍부한 이미지 편집 기능을 갖추고 있으며, 이미지 ### 1. 워크플로우 파일 다운로드 - + Open and run this workflow directly in Comfy Cloud. diff --git a/ko/tutorials/image/pixeldit/pixeldit.mdx b/ko/tutorials/image/pixeldit/pixeldit.mdx index 17a0cdd5b..0a371da8f 100644 --- a/ko/tutorials/image/pixeldit/pixeldit.mdx +++ b/ko/tutorials/image/pixeldit/pixeldit.mdx @@ -34,7 +34,6 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" JSON을 다운로드하거나 템플릿 라이브러리에서 "PixelDiT"를 검색하세요 - 워크플로우는 세 가지 주요 노드로 구성됩니다: @@ -46,7 +45,6 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인하여 워크플로우를 맞춤화하고 확장하는 방법을 배워보세요. - ### 워크플로우 컨트롤 **텍스트 기반 이미지 생성 (PixelDiT)** 서브그래프 노드에 노출된 컨트롤은 다음과 같습니다: @@ -63,14 +61,14 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" PixelDiT는 두 개의 모델 파일을 사용합니다: 텍스트 인코더와 확산 모델입니다. + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 텍스트 인코더 - pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 확산 모델 - + ### 모델 저장 위치 ``` diff --git a/ko/tutorials/image/qwen/qwen-image-2512.mdx b/ko/tutorials/image/qwen/qwen-image-2512.mdx index 2509ec437..ffca6338b 100644 --- a/ko/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ko/tutorials/image/qwen/qwen-image-2512.mdx @@ -45,7 +45,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 JSON 다운로드 또는 템플릿 라이브러리에서 "Qwen-Image-2512" 검색 diff --git a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx index b8562daa8..08acf49bd 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -34,7 +34,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그하여 불러올 수 있습니다. - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 Download JSON or search "Qwen-Image-Edit-2511" in Template Library diff --git a/ko/tutorials/image/qwen/qwen-image-edit.mdx b/ko/tutorials/image/qwen/qwen-image-edit.mdx index 211ddfc83..96e5e01b7 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit.mdx @@ -50,7 +50,7 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 JSON 다운로드 또는 템플릿 라이브러리에서 "image_qwen_image_edit" 검색 - + Run this workflow on Cloud GPUs with zero setup diff --git a/ko/tutorials/image/qwen/qwen-image-layered.mdx b/ko/tutorials/image/qwen/qwen-image-layered.mdx index 1029408d2..513808568 100644 --- a/ko/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ko/tutorials/image/qwen/qwen-image-layered.mdx @@ -32,11 +32,11 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 워크플로우 | -| +| | Download the JSON workflow file | | -| +| | Run ComfyUI online with zero setup | | diff --git a/ko/tutorials/image/qwen/qwen-image.mdx b/ko/tutorials/image/qwen/qwen-image.mdx index 040c81323..b5a2b3b9b 100644 --- a/ko/tutorials/image/qwen/qwen-image.mdx +++ b/ko/tutorials/image/qwen/qwen-image.mdx @@ -60,7 +60,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - + @@ -163,7 +163,7 @@ Qwen_image_distill 이것은 ControlNet 모델이므로 일반 ControlNet처럼 사용할 수 있습니다. - + @@ -218,7 +218,7 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets 모델 패치 워크플로우 - + @@ -289,7 +289,7 @@ ControlNet 관련 워크플로우를 처음 사용한다면, 제어 이미지는 ## Qwen Image Union ControlNet LoRA 워크플로우 - + diff --git a/ko/tutorials/image/z-image/z-image-turbo.mdx b/ko/tutorials/image/z-image/z-image-turbo.mdx index 039474d00..05e9548bd 100644 --- a/ko/tutorials/image/z-image/z-image-turbo.mdx +++ b/ko/tutorials/image/z-image/z-image-turbo.mdx @@ -35,30 +35,29 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ## Z-Image-Turbo 텍스트-이미지 변환 워크플로우 - + + Z-Image-Turbo 텍스트-이미지 변환 워크플로우 JSON 파일을 다운로드하세요. - 이 워크플로우를 ComfyUI Cloud에서 바로 실행하세요. - + ### Z-Image-Turbo 모델 다운로드 + Z-Image-Turbo용 텍스트 인코더입니다. - Z-Image-Turbo용 디퓨전 모델입니다. - Z-Image-Turbo용 VAE입니다. - + **Z-Image-Turbo 모델 저장 위치** ``` @@ -76,16 +75,14 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" 이 워크플로우는 Z-Image-Turbo Fun Union ControlNet 모델을 사용하여 ControlNet 가이드를 기반으로 이미지를 생성합니다. 참조 이미지에 Canny 에지 감지를 적용하고 ControlNet을 통해 생성 과정을 제어합니다. - + Z-Image-Turbo Fun Union ControlNet 워크플로우 JSON 파일을 다운로드하세요. - ### ControlNet용 추가 모델 Z-Image-Turbo용 ControlNet 모델 패치입니다. - **모델 저장 위치** ``` diff --git a/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 363ee621f..bdc045829 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -103,7 +103,7 @@ ComfyUI에는 두 가지 사전 구축된 Seedance 2.0 리얼 휴먼 템플릿 Seedance 2.0 Real Human R2V workflow preview - + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. diff --git a/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index fd51b563a..53fd63a09 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -41,7 +41,7 @@ Seedance 2.0은 ByteDance의 차세대 다중 모달 동영상 생성 모델로, Seedance 2.0 Text-to-Video workflow preview - + @@ -53,7 +53,7 @@ Seedance 2.0은 ByteDance의 차세대 다중 모달 동영상 생성 모델로, Seedance 2.0 Reference-to-Video workflow preview - + @@ -65,7 +65,7 @@ Seedance 2.0은 ByteDance의 차세대 다중 모달 동영상 생성 모델로, Seedance 2.0 FLF2V workflow preview - + diff --git a/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index 23af4bd7a..b9f81d8b7 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -35,7 +35,6 @@ Seedream 5.0 Pro는 ByteDance의 전문가급 이미지 생성 모델로, Seedre Comfy Cloud에서 열기 - Comfy Cloud에서 열기 diff --git a/ko/tutorials/partner-nodes/google/nano-banana-2.mdx b/ko/tutorials/partner-nodes/google/nano-banana-2.mdx index 57b6d3e05..1bad8129b 100644 --- a/ko/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/ko/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -32,20 +32,19 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 나노 바나나 2 이미지 편집 워크플로우 + Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "나노 바나나 2" 검색 - + ![Nano Banana 2 workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_nano_banana2_image_edit-1.webp) 이 워크플로의 예제 입력 이미지 가져오기 - ### 프로 수준 품질 샘플 ![품질 비교](https://substack-post-media.s3.amazonaws.com/public/images/1f92ae2e-14d8-4a3b-9ed7-e57dacd2584f_1825x1696.png) diff --git a/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index f17dbacce..0531182e4 100644 --- a/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/ko/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -37,7 +37,6 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 이 워크플로의 예제 입력 이미지를 가져오세요. - ### 워크플로 개요 이 워크플로우는 세 개의 노드를 사용합니다: diff --git a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index c0f1a7a32..61ca9f379 100644 --- a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -50,7 +50,7 @@ HappyHorse 1.0은 강력한 미학, 다중 샷 시퀀싱, 그리고 강력한 - + Try the Image-to-Video workflow instantly on Comfy Cloud. @@ -68,7 +68,7 @@ HappyHorse 1.0은 강력한 미학, 다중 샷 시퀀싱, 그리고 강력한 > - + Try the Text-to-Video workflow instantly on Comfy Cloud. @@ -80,7 +80,7 @@ HappyHorse 1.0은 강력한 미학, 다중 샷 시퀀싱, 그리고 강력한 참조 피사체를 활용해 동영상 생성을 유도하며, 영화 같은 다중 샷 시퀀스에서도 정체성을 유지합니다. - + Try the Reference-to-Video workflow instantly on Comfy Cloud. @@ -92,7 +92,7 @@ HappyHorse 1.0은 강력한 미학, 다중 샷 시퀀싱, 그리고 강력한 기존 영상을 변형하거나 피사체를 교체/삽입하면서 모션과 구성을 유지하는 V2V 및 SV2V 편집 워크플로우를 활용하세요. - + Try the Video Edit workflow instantly on Comfy Cloud. diff --git a/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index ef6f12e16..51d95b1d0 100644 --- a/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -61,7 +61,6 @@ ComfyUI는 현재 해당 Hunyuan 3D API를 기본적으로 통합하여, ComfyUI Comfy Cloud에서 실행 - ## 이미지 기반 3D 생성 워크플로우 하나 이상의 이미지를 업로드해 고품질 3D 모델을 생성합니다. 2~4개의 멀티뷰 이미지 지원으로 지오메트리와 재질의 정밀도가 향상됩니다. @@ -69,7 +68,6 @@ ComfyUI는 현재 해당 Hunyuan 3D API를 기본적으로 통합하여, ComfyUI Comfy Cloud에서 실행 - ## 멀티뷰 기반 3D 모델 생성 워크플로우 앞면, 뒷면, 옆면 등 여러 각도의 이미지를 제공해 더욱 정확하고 세부적인 3D 모델을 생성합니다. 이는 이미지 기반 3D 생성과 동일한 워크플로우로, 서로 다른 각도의 이미지 2~4장을 업로드하기만 하면 됩니다. @@ -77,7 +75,6 @@ ComfyUI는 현재 해당 Hunyuan 3D API를 기본적으로 통합하여, ComfyUI Comfy Cloud에서 실행 - ## 고급 기능 [초기 HY 3D 3.0 통합](https://blog.comfy.org/p/hunyuan-3d-30-in-comfyui-state-of) 이후, Hunyuan 3D의 고급 처리 기능들이 파트너 노드를 통해 이용 가능해졌습니다. 이러한 워크플로우는 주요 후처리 단계를 ComfyUI로 가져옴으로써 생성과 생산 간의 격차를 줄여줍니다. @@ -86,34 +83,35 @@ ComfyUI는 현재 해당 Hunyuan 3D API를 기본적으로 통합하여, ComfyUI 완전한 3D 모델을 갑옷 조각, 액세서리, 바퀴 등 의미 있는 구조적 부품으로 나눕니다. 이를 통해 자산의 특정 영역을 쉽게 편집하고, 부품을 교체해 다양한 변형을 만들며, 모듈식 워크플로우, 애니메이션 또는 하위 조립을 위한 모델을 준비할 수 있습니다. - + + 워크플로우 실행 - JSON 파일 받기 - + ### UV 언랩핑 지원되는 3D 모델에 대해 자동으로 UV 레이아웃을 생성해 원시 지오메트리를 훨씬 더 쉽게 텍스처링할 수 있는 자산으로 변환합니다. 수작업으로 이음새를 자르고 UV 섬을 정리하는 대신, 창작자는 보다 깔끔한 시작점에서 페인팅, 베이킹 및 재질 작업으로 빠르게 진행할 수 있습니다. - + + 워크플로우 실행 - JSON 파일 받기 - + ### 스마트 토폴로지 밀도 높은 지오메트리를 보다 깔끔한 메쉬로 변환해 에지 흐름을 정리하며, 생성된 모델을 실제 생산 파이프라인에서 더 쉽게 최적화하고 재사용할 수 있도록 돕습니다. 특히 게임 엔진, 실시간 렌더링 또는 밀도가 낮고 구조가 더 좋은 지오메트리를 활용하는 모든 워크플로우에 유용합니다. - + + 워크플로우 실행 - JSON 파일 받기 + \ No newline at end of file diff --git a/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index 80807e285..af4d3202f 100644 --- a/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -60,7 +60,6 @@ ComfyUI는 현재 해당 Hunyuan 3D API를 기본적으로 통합하여, ComfyUI Comfy Cloud에서 실행 - ## 이미지 기반 3D 생성 워크플로우 하나 이상의 이미지를 업로드해 고품질 3D 모델을 생성하세요. 2~4개의 멀티뷰 이미지 지원으로 지오메트리와 재질의 정밀도가 향상됩니다. @@ -68,11 +67,10 @@ ComfyUI는 현재 해당 Hunyuan 3D API를 기본적으로 통합하여, ComfyUI Comfy Cloud에서 실행 - ## 멀티뷰 기반 3D 모델 생성 워크플로우 앞면, 뒷면, 옆면 등 여러 각도의 이미지를 제공해 더욱 정확하고 상세한 3D 모델을 생성하세요. 이 작업은 이미지 기반 3D 생성과 동일한 워크플로우를 사용하며, 서로 다른 각도에서 2~4장의 이미지를 업로드하기만 하면 됩니다. Comfy Cloud에서 실행 - + \ No newline at end of file diff --git a/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 827c26d0b..88c0bb015 100644 --- a/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -25,14 +25,14 @@ Ideogram 4.0은 Ideogram의 최신 텍스트 기반 이미지 생성 모델로, Ideogram 4.0 Text-to-Image workflow preview + Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "Ideogram v4: Text to Image (API)" 검색 - + ![Ideogram 4.0 예시 출력](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_ideogram_v4_t2i.png) *Ideogram 4.0 API의 예시 출력* diff --git a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx index 44a16ee92..ce4f0354d 100644 --- a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -54,14 +54,14 @@ Kling 2.6 모션 컨트롤은 Kuaishou가 개발한 특수 다중모달 모델 Kling 2.6 Motion Control workflow preview + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. - Download the workflow JSON file for local use. - +
Input materials diff --git a/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx b/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx index 8f5dd261e..a8afd2bdc 100644 --- a/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -78,7 +78,7 @@ Uni-1의 모든 것은 한 가지 질문으로 시작합니다: **새로운 것 Luma Uni-1 Image Create workflow preview - + @@ -88,7 +88,7 @@ Uni-1의 모든 것은 한 가지 질문으로 시작합니다: **새로운 것 Luma Uni-1 Image Edit workflow preview - + 워크플로는 간단합니다: **프롬프트 → 평가 → 다듬기**. 탐색 단계에서는 시드를 비워 두세요. 마음에 드는 결과를 찾으면 시드를 고정하고 거기서부터 반복하세요. diff --git a/ko/tutorials/partner-nodes/meshy/meshy-6.mdx b/ko/tutorials/partner-nodes/meshy/meshy-6.mdx index f4c91fc0c..1cd382114 100644 --- a/ko/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/ko/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -36,34 +36,33 @@ Meshy 6는 Meshy의 최신 3D 모델 생성 기술로, 지오메트리 품질, Meshy 6를 사용해 텍스트 설명으로부터 바로 3D 모델을 생성하세요. - + + Comfy Cloud에서 즉시 텍스트 기반 모델 워크플로우를 실행하세요. - 로컬에서 사용할 수 있도록 워크플로우 JSON 파일을 다운로드하세요. - + ## 이미지 기반 모델 워크플로우 Meshy 6의 이미지 기반 3D 생성 기능을 활용해 2D 이미지를 세부적인 3D 모델로 변환하세요. - + + Comfy Cloud에서 즉시 이미지 기반 모델 워크플로우를 실행하세요. - 로컬에서 사용할 수 있도록 워크플로우 JSON 파일을 다운로드하세요. - + ## 다중 뷰 기반 모델 워크플로우 여러 뷰 이미지로부터 3D 모델을 생성해 더욱 정확한 지오메트리와 텍스처 재구성을 실현하세요. - + Run the multi-view workflow instantly on Comfy Cloud. - ![Meshy 6 Multi-view to Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_multi_image_to_model-1.webp) diff --git a/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx b/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx index e0f109c94..4d31d2521 100644 --- a/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -45,28 +45,28 @@ GPT-Image-2는 노드 라이브러리의 **OpenAI GPT Image 1.5** 노드에서 ` GPT-Image-2의 추론 기반 구성으로 텍스트 프롬프트를 통해 이미지를 생성하세요. - + + Comfy Cloud에서 텍스트 → 이미지 워크플로를 즉시 시험해 보세요. - 워크플로 JSON을 다운로드하세요. - + ![GPT-Image-2 텍스트 → 이미지 예시](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_t2i_1.jpg) ### 이미지 편집 최대 2K 해상도에서 높은 구조적 정확도로 입력 이미지를 편집하세요. - + + Comfy Cloud에서 이미지 편집 워크플로를 즉시 시험해 보세요. - 워크플로 JSON을 다운로드하세요. - + ![GPT-Image-2 이미지 → 이미지 예시](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_i2i_1.jpg) ![GPT-Image-2 이미지 편집 예시 1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_image_edit_1.jpg) diff --git a/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx b/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx index 32540aee1..9c03e84ac 100644 --- a/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -34,14 +34,14 @@ Recraft V4는 전문 디자인 작업을 위해 개발된 새로운 이미지 ## Recraft V4 텍스트 기반 이미지 생성 워크플로우 + Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "Recraft V4 텍스트 기반 이미지 생성" 검색 - + ![Recraft V4 Text to Image workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_t2i-1.webp) ### 워크플로우 실행 단계 @@ -67,14 +67,14 @@ Recraft V4는 전문 디자인 작업을 위해 개발된 새로운 이미지 Recraft V4는 직접 제작 준비가 완료된 SVG 벡터 출력을 생성할 수 있습니다. 이는 로고, 아이콘, 브랜드 자산 등 확장성이 필요한 경우 유용합니다. SVG 출력은 일러스트레이터, 피그마, 스케치와 호환됩니다. + Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "Recraft V4 텍스트 기반 벡터 생성" 검색 - + ![Recraft V4 Text to Vector workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_text_to_vector-1.webp) ### 벡터 쇼케이스 diff --git a/ko/tutorials/partner-nodes/rodin/model-generation.mdx b/ko/tutorials/partner-nodes/rodin/model-generation.mdx index 82121b3b4..a7388b30a 100644 --- a/ko/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ko/tutorials/partner-nodes/rodin/model-generation.mdx @@ -37,7 +37,7 @@ Generate a 3D model from a single image input with Rodin. Single-view Model Generation (Json Format) - + Comfy Cloud에서 열기 @@ -80,7 +80,7 @@ Download this sample input image to try the workflow: Multi-view Model Generation (Json Format) - Comfy Cloud에서 열기 + Comfy Cloud에서 열기
diff --git a/ko/tutorials/partner-nodes/topaz/astra-2.mdx b/ko/tutorials/partner-nodes/topaz/astra-2.mdx index 51dbdc03b..f37569f2a 100644 --- a/ko/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/ko/tutorials/partner-nodes/topaz/astra-2.mdx @@ -40,14 +40,14 @@ Astra 1을 기반으로 하며, 세부사항과 스타일화가 어떻게 생성 Astra 2 workflow preview - + + Comfy Cloud에서 바로 워크플로우를 열어보세요. - 로컬 ComfyUI용 워크플로우 JSON을 다운로드하세요. - +
입력 자료 diff --git a/ko/tutorials/partner-nodes/tripo/model-generation.mdx b/ko/tutorials/partner-nodes/tripo/model-generation.mdx index 8c129071a..8d336dfc3 100644 --- a/ko/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ko/tutorials/partner-nodes/tripo/model-generation.mdx @@ -37,7 +37,7 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - + Try the Text-to-Model workflow instantly on Comfy Cloud. @@ -65,7 +65,7 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 아래 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. - + Try the Image-to-Model workflow instantly on Comfy Cloud. @@ -102,7 +102,7 @@ Generate a 3D model from multiple view images for enhanced accuracy. Tripo Multi-view Model Generation workflow preview - + Try the Multiview-to-Model workflow instantly on Comfy Cloud. diff --git a/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx index 38758feb5..6152aaaac 100644 --- a/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -46,7 +46,7 @@ Generate a high-detail 3D model from a text prompt using Tripo 3.1. Tripo 3.1 Text-to-Model workflow preview - + @@ -58,7 +58,7 @@ Generate a high-detail 3D model from an image input using Tripo 3.1. Tripo 3.1 Image-to-Model workflow preview - + @@ -70,7 +70,7 @@ Generate a high-detail 3D model from multiple view images using Tripo 3.1. Tripo 3.1 Multiview-to-Model workflow preview - + diff --git a/ko/tutorials/utility/moge.mdx b/ko/tutorials/utility/moge.mdx index 2611f89f9..4876243f0 100644 --- a/ko/tutorials/utility/moge.mdx +++ b/ko/tutorials/utility/moge.mdx @@ -72,14 +72,14 @@ ComfyUI/ MoGe는 이미지로부터 카메라의 시야각(FOV)도 추정하며, 이를 정확성을 더욱 높이기 위해 실제값으로 덮어쓸 수도 있습니다. + JSON 다운로드하거나 템플릿 라이브러리에서 "MoGe 깊이 추정" 검색 - 이 워크플로우의 예제 입력 이미지 받기 - +
깊이 추정 컬러 미리보기 깊이 추정 원본 미리보기 @@ -98,14 +98,14 @@ MoGe는 이미지로부터 카메라의 시야각(FOV)도 추정하며, 이를 **작동 방식**: 단일 투시 사진을 질감이 적용된 GLB 메시로 변환하며, 법선 및 깊이 미리보기를 함께 제공합니다. MoGe는 보이는 장면에서 점 지도, 깊이 및 법선을 추정한 후 이를 메시로 변환합니다. 이는 **단안 기하학 추정**으로, 가려진 부분과 객체 뒷면은 누락되거나 조각나게 됩니다. 빠른 장면 프로토타입, 참조 기하학 또는 깊이와 법선을 메시로 시각화하는 데 유용하지만, 다중뷰 3D 재구성의 대체물은 아닙니다. + JSON 다운로드하거나 템플릿 라이브러리에서 "3D MoGe 투시도에서 메시 생성" 검색 - 이 워크플로우의 예제 입력 이미지 받기 - + ![투시도에서 메시 생성 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_perspective_to_mesh-1.webp) ### 2.1 실행 단계 @@ -121,14 +121,14 @@ MoGe는 이미지로부터 카메라의 시야각(FOV)도 추정하며, 이를 **작동 방식**: 등거리(360°) 파노라마를 질감이 적용된 GLB 메시로 변환합니다. 워크플로우는 `MoGePanoramaInference`를 사용해 파노라마를 12개의 투시도로 분할한 후 각각의 투시도에서 단안 기하학 추정을 독립적으로 수행한 후 이를 하나의 메시로 합칩니다. 각 세그먼트는 여전히 단일뷰 추정이므로 결과는 대략적인 장면 재구성입니다. 360° 장면의 공간적 개요를 얻는 데 유용하지만, 가려진 부분과 표면 뒤쪽의 기하학은 누락되거나 조각나게 됩니다. + JSON 다운로드하거나 템플릿 라이브러리에서 "3D MoGe 파노라마에서 메시 생성" 검색 - 이 워크플로우의 예제 입력 이미지 받기 - + ![파노라마에서 메시 생성 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_panorama_to_mesh-1.webp) ### 3.1 실행 단계 diff --git a/ko/tutorials/video/bytedance/bernini-r.mdx b/ko/tutorials/video/bytedance/bernini-r.mdx index 6c9c0b56b..649485dbb 100644 --- a/ko/tutorials/video/bytedance/bernini-r.mdx +++ b/ko/tutorials/video/bytedance/bernini-r.mdx @@ -108,7 +108,6 @@ ComfyUI/ 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 사용자 지정 및 확장에 대한 자세한 내용은 서브그래프 문서를 확인하세요. - --- ## 2. 비디오 편집 @@ -139,7 +138,6 @@ ComfyUI/ 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 사용자 지정 및 확장에 대한 자세한 내용은 서브그래프 문서를 확인하세요. - ## 커뮤니티 리소스 - [Bernini GitHub (bytedance/Bernini)](https://github.com/bytedance/Bernini) — 연구 논문 및 작업 문서 diff --git a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 7a75aa8d7..2bcdd7102 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -28,39 +28,165 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; - **시네마틱 품질**: 기본 720p 출력(1080p까지 스케일링 가능)으로 전문적인 미학을 자랑합니다. - **풍부한 기능**: 다양한 스타일(사실적, 애니메이션, 3D)과 비디오 내 텍스트 렌더링(중국어/영어)을 지원합니다. -## 워크플로우 템플릿 +## 모든 워크플로에 공통으로 사용하는 모델 -[video_hunyuan_video_1.5_720p_i2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_i2v.json) +다음 모델은 텍스트-투-비디오 및 이미지-투-비디오 워크플로 모두에서 사용됩니다. 다운로드하여 지정된 디렉터리에 저장하세요. -[video_hunyuan_video_1.5_720p_t2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_t2v.json) + + + qwen_2.5_vl_7b_fp8_scaled.safetensors. ComfyUI/models/text_encoders/에 저장 + + + byt5_small_glyphxl_fp16.safetensors. ComfyUI/models/text_encoders/에 저장 + + + hunyuanvideo15_vae_fp16.safetensors. ComfyUI/models/vae/에 저장 + + -## 모델 링크 -**텍스트 인코더** +#### 저장 위치 -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +``` +ComfyUI/ +├── 📂 models/ +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors +│ │ └── byt5_small_glyphxl_fp16.safetensors +│ └── 📂 vae/ +│ └── hunyuanvideo15_vae_fp16.safetensors +``` + +## Hunyuan Video 1.5 텍스트-투-비디오 워크플로 + +HunyuanVideo 1.5 텍스트-투-비디오는 자연어 설명으로 5~10초 비디오를 생성하며, 품질을 높이면서 VRAM 요구량을 줄입니다. + +![ComfyUI 워크플로 - Hunyuan Video 1.5 T2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_t2v-1.webp) + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Hunyuan Video 1.5 T2V" 검색 + + + +### 모델 다운로드 + + + + hunyuanvideo1.5_720p_t2v_fp16.safetensors. ComfyUI/models/diffusion_models/에 저장 + + + +#### 모델 저장 위치 + +``` +ComfyUI/ +├── 📂 models/ +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // 공유 모델 +│ │ └── byt5_small_glyphxl_fp16.safetensors // 공유 모델 +│ ├── 📂 vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors // 공유 모델 +│ └── 📂 diffusion_models/ +│ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V 모델 +``` + +### 워크플로 실행 단계 -**디퓨전 모델** +1. `DualCLIPLoader` 노드에 다음 모델이 로드되었는지 확인: + - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` + - clip_name2: `byt5_small_glyphxl_fp16.safetensors` +2. `Load Diffusion Model` 노드에 `hunyuanvideo1.5_720p_t2v_fp16.safetensors`가 로드되었는지 확인 +3. `Load VAE` 노드에 `hunyuanvideo15_vae_fp16.safetensors`가 로드되었는지 확인 +4. `Queue` 버튼을 클릭하거나 `Ctrl(Cmd) + Enter` 단축키로 워크플로 실행 + + +워크플로에는 초해상도 업스케일러 노드가 포함되어 있습니다. 활성화하면 `hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors`를 사용해 출력을 1080p로 업스케일합니다. + + +## Hunyuan Video 1.5 이미지-투-비디오 워크플로 + +HunyuanVideo 1.5 이미지-투-비디오는 정지 이미지를 부드럽고 고품질의 비디오로 변환하며, 일관성과 모션 역학을 개선합니다. + +![ComfyUI 워크플로 - Hunyuan Video 1.5 I2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_i2v-1.webp) + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Hunyuan Video 1.5 I2V" 검색 + + + +#### 입력 소재 + + + + 기본 입력 이미지를 다운로드하거나, 자신의 이미지를 시작 프레임으로 사용하세요. + + + +### 모델 다운로드 + + + + sigclip_vision_patch14_384.safetensors. ComfyUI/models/clip_vision/에 저장 + + + hunyuanvideo1.5_720p_i2v_fp16.safetensors. ComfyUI/models/diffusion_models/에 저장 + + + +#### 모델 저장 위치 + +``` +ComfyUI/ +├── 📂 models/ +│ ├── 📂 clip_vision/ +│ │ └── sigclip_vision_patch14_384.safetensors // I2V 비전 인코더 +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // 공유 모델 +│ │ └── byt5_small_glyphxl_fp16.safetensors // 공유 모델 +│ ├── 📂 vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors // 공유 모델 +│ └── 📂 diffusion_models/ +│ └── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V 모델 +``` -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +### 워크플로 실행 단계 -**VAE** +1. `DualCLIPLoader` 노드에 다음 모델이 로드되었는지 확인: + - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` + - clip_name2: `byt5_small_glyphxl_fp16.safetensors` +2. `CLIPVisionLoader` 노드에 `sigclip_vision_patch14_384.safetensors`가 로드되었는지 확인 +3. `Load Diffusion Model` 노드에 `hunyuanvideo1.5_720p_i2v_fp16.safetensors`가 로드되었는지 확인 +4. `Load VAE` 노드에 `hunyuanvideo15_vae_fp16.safetensors`가 로드되었는지 확인 +5. `Queue` 버튼을 클릭하거나 `Ctrl(Cmd) + Enter` 단축키로 워크플로 실행 -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +## 초해상도 업스케일러 +두 워크플로 모두 720p 출력 비디오를 1080p로 업스케일할 수 있는 초해상도 노드를 포함합니다. 이 선택적 업스케일러는 증류 모델을 사용해 효율적인 고해상도 출력을 제공합니다. -모델 저장 위치 + + hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors. ComfyUI/models/diffusion_models/에 저장 + +공유 모델과 워크플로별 모델 전체 목록: ``` -:open_file_folder: ComfyUI/ -├── :open_file_folder: models/ -│ ├── :open_file_folder: text_encoders/ -│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors -│ │ └── byt5_small_glyphxl_fp16.safetensors -│ ├── :open_file_folder: diffusion_models/ -│ │ ├── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors -│ │ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors -│ └── :open_file_folder: vae/ -│ └── hunyuanvideo15_vae_fp16.safetensors +ComfyUI/ +├── 📂 models/ +│ ├── 📂 clip_vision/ +│ │ └── sigclip_vision_patch14_384.safetensors +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors +│ │ └── byt5_small_glyphxl_fp16.safetensors +│ ├── 📂 vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors +│ └── 📂 diffusion_models/ +│ ├── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V 모델 +│ ├── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V 모델 +│ └── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors // 업스케일러 ``` diff --git a/ko/tutorials/video/kandinsky/kandinsky-5.mdx b/ko/tutorials/video/kandinsky/kandinsky-5.mdx index 2553e5a67..bb8b68c94 100644 --- a/ko/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/ko/tutorials/video/kandinsky/kandinsky-5.mdx @@ -52,13 +52,13 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ComfyUI를 최신 버전으로 업데이트해 주시고, 메뉴 `워크플로우` -> `템플릿 탐색` -> `비디오`를 통해 "칸딘스키 5.0 T2V"를 찾아 워크플로우를 로드해 주세요. -![Kandinsky 5.0 T2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_t2v-1.webp) +![Kandinsky 5.0 T2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_t2v-1.webp) - + Download the T2V workflow to use locally - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 ### 2. 모델 수동 다운로드 @@ -91,13 +91,13 @@ ComfyUI/ ComfyUI를 최신 버전으로 업데이트해 주시고, 메뉴 `워크플로우` -> `템플릿 탐색` -> `비디오`를 통해 "칸딘스키 5.0 I2V"를 찾아 워크플로우를 로드해 주세요. -![Kandinsky 5.0 I2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_i2v-1.webp) +![Kandinsky 5.0 I2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_i2v-1.webp) - + Download the I2V workflow to use locally - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 ### 2. 모델 수동 다운로드 diff --git a/ko/tutorials/video/ltx/ltx-2-3.mdx b/ko/tutorials/video/ltx/ltx-2-3.mdx index 2d44aa928..e7b5d9ddc 100644 --- a/ko/tutorials/video/ltx/ltx-2-3.mdx +++ b/ko/tutorials/video/ltx/ltx-2-3.mdx @@ -300,7 +300,6 @@ ComfyUI/ 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인하여 워크플로우를 사용자 정의하고 확장하는 방법을 알아보세요. - #### 입력 자료 diff --git a/ko/tutorials/video/ltxv.mdx b/ko/tutorials/video/ltxv.mdx index 589857753..2c4a28824 100644 --- a/ko/tutorials/video/ltxv.mdx +++ b/ko/tutorials/video/ltxv.mdx @@ -34,7 +34,7 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 첫 번째 [프레임 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)를 통해 비디오를 제어할 수 있습니다. - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 JSON 다운로드 또는 템플릿 라이브러리에서 "LTX-Video" 검색 diff --git a/ko/tutorials/video/wan/fun-camera.mdx b/ko/tutorials/video/wan/fun-camera.mdx index 28982ce14..ac4bb8765 100644 --- a/ko/tutorials/video/wan/fun-camera.mdx +++ b/ko/tutorials/video/wan/fun-camera.mdx @@ -42,10 +42,10 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 1.3B 또는 14B 중 하나를 선택하세요: - + Wan2.1 Fun Camera 1.3B 디퓨전 모델 - + Wan2.1 Fun Camera 14B 디퓨전 모델 @@ -57,10 +57,10 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 다음 중 하나를 선택하세요: - + 전체 정밀도 텍스트 인코더 - + FP8 양자화 텍스트 인코더 (낮은 VRAM에 권장) @@ -68,7 +68,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ### VAE - + Wan2.1 VAE 모델 @@ -76,7 +76,7 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ### CLIP 비전 - + CLIP 비전 인코더 diff --git a/ko/tutorials/video/wan/wan-alpha.mdx b/ko/tutorials/video/wan/wan-alpha.mdx index c2ffa8ec1..55423d5ca 100644 --- a/ko/tutorials/video/wan/wan-alpha.mdx +++ b/ko/tutorials/video/wan/wan-alpha.mdx @@ -11,6 +11,8 @@ translationBlockHashes: --- +import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; + Wan-Alpha는 알파 채널 투명도가 적용된 고품질 동영상을 생성하는 특수한 텍스트 기반 비디오 생성 모델입니다. Wan2.1-14B-T2V 베이스 모델을 기반으로 하며, 투명한 배경과 반투명 객체를 포함한 동영상을 만들어 합성 작업에 완벽합니다. 이 모델은 투명한 배경, 반투명 객체(기포, 유리, 물), 빛나는 효과 및 정확한 알파 채널을 갖춘 세밀한 디테일(머리카락, 연기, 입자)을 잘 생성합니다. @@ -27,11 +29,103 @@ Wan-Alpha는 알파 채널 투명도가 적용된 고품질 동영상을 생성 allowfullscreen > -[워크플로 다운로드](https://github.com/Comfy-Org/workflows/blob/main/tutorial_workflows/Get_Comfy_With_Comfy_Wan_Alpha.json) - ## 리소스 - [Wan-Alpha GitHub](https://github.com/WeChatCV/Wan-Alpha) - [Hugging Face 모델](https://huggingface.co/htdong/Wan-Alpha) - [ComfyUI 버전](https://huggingface.co/htdong/Wan-Alpha_ComfyUI) - [연구 논문](https://arxiv.org/pdf/2509.24979) + + + +## Wan-Alpha 텍스트-투-비디오 워크플로 (14B) + +### 1. 워크플로 다운로드 + +ComfyUI를 최신 버전으로 업데이트한 뒤 워크플로 파일을 다운로드해 ComfyUI에 드래그하거나, 템플릿 라이브러리의 `Workflow` → `Browse Templates` → `Video`에서 "Wan 2.1 Alpha T2V 14B"를 검색하세요. + +Wan-Alpha T2V 워크플로 + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan 2.1 Alpha T2V 14B" 검색 + + + +### 2. 모델 설치 + +관련 모델은 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged)에서 찾을 수 있습니다. + +**디퓨전 모델**: 버전 중 하나를 선택하세요. + + + + FP8 스케일 디퓨전 모델. ComfyUI/models/diffusion_models/에 저장 + + + BF16 디퓨전 모델. ComfyUI/models/diffusion_models/에 저장 + + + +**텍스트 인코더** + + + + FP8 텍스트 인코더. ComfyUI/models/text_encoders/에 저장 + + + +**VAE 모델** + + + + RGB 채널 VAE. ComfyUI/models/vae/에 저장 + + + 알파 채널 VAE. ComfyUI/models/vae/에 저장 + + + +**LoRA 모델** + + + + 알파 생성 LoRA. ComfyUI/models/loras/에 저장 + + + 빠른 추론용 Lightning LoRA. ComfyUI/models/loras/에 저장 + + + +파일 저장 위치: + +``` +ComfyUI/ +├── models/ +│ ├── diffusion_models/ +│ │ ├── wan2.1_t2v_14B_fp8_scaled.safetensors +│ │ └── wan2.1_t2v_14B_bf16.safetensors +│ ├── text_encoders/ +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── vae/ +│ │ ├── wan_alpha_2.1_vae_rgb_channel.safetensors +│ │ └── wan_alpha_2.1_vae_alpha_channel.safetensors +│ └── loras/ +│ ├── wan_alpha_2.1_rgba_lora.safetensors +│ └── lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors +``` + +### 3. 워크플로 단계별 완료 + +1. `Load Diffusion Model` 노드에 올바른 디퓨전 모델이 로드되었는지 확인 +2. `Load CLIP` 노드에 `umt5_xxl_fp8_e4m3fn_scaled.safetensors`가 로드되었는지 확인 +3. RGB용 `Load VAE` 노드에 `wan_alpha_2.1_vae_rgb_channel.safetensors`가 로드되었는지 확인 +4. 알파용 두 번째 `Load VAE` 노드에 `wan_alpha_2.1_vae_alpha_channel.safetensors`가 로드되었는지 확인 +5. `LoRA Loader` 노드에 `wan_alpha_2.1_rgba_lora.safetensors`가 로드되었는지 확인 +6. (선택) Lightning LoRA 노드에 `lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors`를 로드해 더 빠르게 생성 +7. `CLIP Text Encode (Positive Prompt)` 노드에서 긍정 프롬프트 설정 +8. (선택) `EmptyHunyuanLatentVideo` 노드에서 동영상 크기 수정 +9. `Run` 버튼을 클릭하거나 `Ctrl(Cmd) + Enter`로 생성 실행 diff --git a/ko/tutorials/video/wan/wan-causal-forcing.mdx b/ko/tutorials/video/wan/wan-causal-forcing.mdx index 8c0970a51..6ff065a2e 100644 --- a/ko/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ko/tutorials/video/wan/wan-causal-forcing.mdx @@ -43,7 +43,6 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인하여 워크플로우를 사용자 정의하고 확장하는 방법을 알아보세요. - ### Causal Forcing vs Causal Forcing++ | 모드 | 설명 | diff --git a/ko/tutorials/video/wan/wan2_2.mdx b/ko/tutorials/video/wan/wan2_2.mdx index 9ad610bac..7c85fffa8 100644 --- a/ko/tutorials/video/wan/wan2_2.mdx +++ b/ko/tutorials/video/wan/wan2_2.mdx @@ -104,7 +104,7 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 5B" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 @@ -157,7 +157,7 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 14B T2V" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 @@ -212,7 +212,7 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 14B I2V" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 @@ -269,7 +269,7 @@ ComfyUI/ JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 14B FLF2V" 검색 - Comfy Cloud에서 열기 + Comfy Cloud에서 열기 diff --git a/ko/tutorials/video/zai/scail2.mdx b/ko/tutorials/video/zai/scail2.mdx index 3a2183f37..8bd4b33b6 100644 --- a/ko/tutorials/video/zai/scail2.mdx +++ b/ko/tutorials/video/zai/scail2.mdx @@ -54,7 +54,6 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 이 워크플로우는 모듈식 처리를 위해 Subgraph 노드를 사용합니다. 서브그래프 문서에서 워크플로우를 사용자 정의하고 확장하는 방법을 알아보세요. - ### 긴 비디오 긴 비디오의 경우 세그먼트 수를 계산합니다: `ceil(total_frames / 76)`. 첫 번째 세그먼트를 제외한 모든 세그먼트는 Extend 서브그래프를 사용합니다. 더 많은 세그먼트를 추가하려면 Extend 노드를 복제하고 `previous_frames` 출력을 연결한 다음 `segment_index`를 증가시킵니다. diff --git a/tutorials/3d/hunyuan3D-2.mdx b/tutorials/3d/hunyuan3D-2.mdx index aca45d0fd..3e371ce92 100644 --- a/tutorials/3d/hunyuan3D-2.mdx +++ b/tutorials/3d/hunyuan3D-2.mdx @@ -48,7 +48,7 @@ In the Hunyuan3D-2mv workflow, we'll use multi-view images to generate a 3D mode - + Run this workflow instantly on Comfy Cloud @@ -100,7 +100,7 @@ If you need to add more views, make sure to load other view images in the `Hunyu In the Hunyuan3D-2mv-turbo workflow, we'll use the Hunyuan3D-2mv-turbo model to generate 3D models. This model is a step distillation version of Hunyuan3D-2mv, allowing for faster 3D model generation. In this version of the workflow, we set `cfg` to 1.0 and add a `flux guidance` node to control the `distilled cfg` generation. - + Run this workflow instantly on Comfy Cloud @@ -147,7 +147,7 @@ ComfyUI/ In the Hunyuan3D-2 workflow, we'll use the Hunyuan3D-2 model to generate 3D models. This model is not a multi-view model. In this workflow, we use the `Hunyuan3Dv2Conditioning` node instead of the `Hunyuan3Dv2ConditioningMultiView` node. - + Run this workflow instantly on Comfy Cloud diff --git a/tutorials/3d/triposplat.mdx b/tutorials/3d/triposplat.mdx index 5af78a21a..0e56235b8 100644 --- a/tutorials/3d/triposplat.mdx +++ b/tutorials/3d/triposplat.mdx @@ -8,14 +8,14 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' **TripoSplat** is an open-source model that generates **3D Gaussian splat** representations directly from a single 2D image. It was developed by VAST-AI and released under an open-source license. -Unlike traditional 3D reconstruction methods that require multiple views or generate meshes as the primary output, TripoSplat creates **Gaussian splat** representations — a rendering technique where thousands of colored 3D Gaussians are placed in space to represent a scene. This approach enables fast, high-quality rendering with controllable density and budget. +Unlike traditional 3D reconstruction methods that require multiple views or generate meshes as the primary output, TripoSplat creates **Gaussian splat** representations: a rendering technique where thousands of colored 3D Gaussians are placed in space to represent a scene. This approach enables fast, high-quality rendering with controllable density and budget. TripoSplat workflow - + Run this workflow instantly on Comfy Cloud @@ -27,15 +27,14 @@ Unlike traditional 3D reconstruction methods that require multiple views or gene TripoSplat uses a **feed-forward architecture** that takes a single RGB image and directly predicts a set of 3D Gaussian primitives. The pipeline involves: -1. **Image encoding** — the input image is processed by a vision encoder (DINOv2) -2. **Triplane generation** — features are decoded into a triplane representation -3. **Gaussian prediction** — the triplane is sampled to produce Gaussian parameters (position, scale, rotation, opacity, color) -4. **Rendering** — Gaussians are rendered from arbitrary viewpoints using differentiable splatting +1. **Image encoding**: the input image is processed by a vision encoder (DINOv2) +2. **Triplane generation**: features are decoded into a triplane representation +3. **Gaussian prediction**: the triplane is sampled to produce Gaussian parameters (position, scale, rotation, opacity, color) +4. **Rendering**: Gaussians are rendered from arbitrary viewpoints using differentiable splatting This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. - ## Workflow node guide ### LoadImage @@ -83,10 +82,10 @@ The main subgraph node processes the image and generates the 3D Gaussian splat. ## Steps to run -1. **Load an image** — use the **LoadImage** node to load a single 2D image -2. **Run the TripoSplat subgraph** — the model will generate a Gaussian splat representation -3. **Choose output format** — export as GLB, SPZ, video, or render to mesh -4. **View results** — use the created 3D file or rendered preview +1. **Load an image**: use the **LoadImage** node to load a single 2D image +2. **Run the TripoSplat subgraph**: the model will generate a Gaussian splat representation +3. **Choose output format**: export as GLB, SPZ, video, or render to mesh +4. **View results**: use the created 3D file or rendered preview ## Output options @@ -101,26 +100,23 @@ The main subgraph node processes the image and generates the 3D Gaussian splat. Download the TripoSplat model and required files. Place them in the corresponding `models/` subdirectories. - - triposplat_fp16.safetensors — TripoSplat diffusion model checkpoint + + + triposplat_fp16.safetensors: TripoSplat diffusion model checkpoint - - - triposplat_vae_decoder_fp16.safetensors — VAE decoder + + triposplat_vae_decoder_fp16.safetensors: VAE decoder - - - flux2-vae.safetensors — Flux.2 VAE for latent encoding + + flux2-vae.safetensors: Flux.2 VAE for latent encoding - - - dino_v3_vit_h.safetensors — CLIP vision encoder (DINOv2) + + dino_v3_vit_h.safetensors: CLIP vision encoder (DINOv2) - - - birefnet.safetensors — Background removal model for preprocessing + + birefnet.safetensors: Background removal model for preprocessing - + ### Model storage location ``` diff --git a/tutorials/flux/flux-1-controlnet.mdx b/tutorials/flux/flux-1-controlnet.mdx index 19d176506..301e186d3 100644 --- a/tutorials/flux/flux-1-controlnet.mdx +++ b/tutorials/flux/flux-1-controlnet.mdx @@ -44,7 +44,7 @@ For image preprocessors, you can use the following custom nodes to complete imag Download JSON or search "Flux.1 Canny" in Template Library - + Open in Comfy Cloud @@ -68,10 +68,10 @@ Since you need to first agree to the terms of [black-forest-labs/FLUX.1-Canny-de Complete model list: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/resolve/main/flux1-canny-dev.safetensors?download=true) (Please ensure you have agreed to the corresponding repo's terms) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true) (Please ensure you have agreed to the corresponding repo's terms) File storage location: ``` @@ -115,7 +115,7 @@ Or use the following custom nodes to complete image preprocessing: Download JSON or search "Flux.1 Depth LoRA" in Template Library - + Open in Comfy Cloud @@ -139,11 +139,11 @@ If you have previously used the [complete version of Flux related workflows](/tu Complete model list: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true) -- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/resolve/main/flux1-depth-dev-lora.safetensors?download=true) +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) +- [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) File storage location: ``` diff --git a/tutorials/flux/flux-1-fill-dev.mdx b/tutorials/flux/flux-1-fill-dev.mdx index fb4ed5dc9..c613cea72 100644 --- a/tutorials/flux/flux-1-fill-dev.mdx +++ b/tutorials/flux/flux-1-fill-dev.mdx @@ -57,7 +57,7 @@ ComfyUI/ Download JSON or search "flux_fill_inpaint" in Template Library - + Open in Comfy Cloud @@ -93,7 +93,7 @@ If you want to draw your own mask, please [click here](https://raw.githubusercon Download JSON or search "flux_fill_outpaint" in Template Library - + Open in Comfy Cloud diff --git a/tutorials/flux/flux-1-kontext-dev.mdx b/tutorials/flux/flux-1-kontext-dev.mdx index 96d35ba6e..c7dfbf460 100644 --- a/tutorials/flux/flux-1-kontext-dev.mdx +++ b/tutorials/flux/flux-1-kontext-dev.mdx @@ -76,7 +76,7 @@ Model save location Download JSON or search "Flux Kontext Dev" in Template Library - + Open in Comfy Cloud diff --git a/tutorials/flux/flux-1-text-to-image.mdx b/tutorials/flux/flux-1-text-to-image.mdx index 1d0a0bde0..d0eeee46b 100644 --- a/tutorials/flux/flux-1-text-to-image.mdx +++ b/tutorials/flux/flux-1-text-to-image.mdx @@ -44,7 +44,7 @@ If you can't download models from [black-forest-labs/FLUX.1-dev](https://hugging #### 1. Workflow File - + Run this workflow on Comfy Cloud @@ -102,7 +102,7 @@ Thanks to Flux's excellent prompt following capability, we don't need any negati #### 1. Workflow File - + Run this workflow on Comfy Cloud @@ -161,7 +161,7 @@ but it also requires less VRAM, and you only need to install one model file to t ### Flux.1 Dev - + Run this workflow on Comfy Cloud @@ -180,7 +180,7 @@ Ensure that the corresponding `Load Checkpoint` node loads `flux1-dev-fp8.safete ### Flux.1 Schnell - + Run this workflow on Comfy Cloud diff --git a/tutorials/flux/flux-1-uso.mdx b/tutorials/flux/flux-1-uso.mdx index 8bcd8201a..bec397bc0 100644 --- a/tutorials/flux/flux-1-uso.mdx +++ b/tutorials/flux/flux-1-uso.mdx @@ -33,7 +33,7 @@ Download the image below and drag it into ComfyUI to load the corresponding work Download the workflow JSON and drag it into ComfyUI - + Run this workflow on Comfy Cloud diff --git a/tutorials/flux/flux-2-dev.mdx b/tutorials/flux/flux-2-dev.mdx index 810c2d5ae..0cb271140 100644 --- a/tutorials/flux/flux-2-dev.mdx +++ b/tutorials/flux/flux-2-dev.mdx @@ -32,39 +32,41 @@ We are using quantized weights in this workflow. The original FLUX.2 repository Basic text-to-image workflow for generating single images with FLUX.2 Dev. - + + Open this workflow directly in Comfy Cloud - - + Download the JSON workflow file for local use + ## Multi-image reference workflow A 2-image reference workflow example. You can extend this implementation to support more reference images. - + + Open this workflow directly in Comfy Cloud - - + Download the JSON workflow file for local use + ## Model links **text_encoders** -- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) +- [mistral_3_small_flux2_bf16.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/text_encoders/mistral_3_small_flux2_bf16.safetensors) **diffusion_models** -- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) +- [flux2_dev_fp8mixed.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/diffusion_models/flux2_dev_fp8mixed.safetensors) **vae** -- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/resolve/main/split_files/vae/flux2-vae.safetensors) +- [flux2-vae.safetensors](https://huggingface.co/Comfy-Org/flux2-dev/blob/main/split_files/vae/flux2-vae.safetensors) **Model Storage Location** diff --git a/tutorials/flux/flux-2-klein.mdx b/tutorials/flux/flux-2-klein.mdx index 5e1f25059..660502295 100644 --- a/tutorials/flux/flux-2-klein.mdx +++ b/tutorials/flux/flux-2-klein.mdx @@ -10,7 +10,7 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## About FLUX.2 [klein] -FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image and image editing in one compact architecture. It’s designed for interactive workflows, immediate previews, and latency-critical applications, with distilled variants delivering end-to-end inference around one second while keeping strong quality for single- and multi-reference editing. +FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image and image editing in one compact architecture. It's designed for interactive workflows, immediate previews, and latency-critical applications, with distilled variants delivering end-to-end inference around one second while keeping strong quality for single- and multi-reference editing. **Model highlights:** - Two 4B types: Base (undistilled) for maximum flexibility and fine-tuning; Distilled (4-step) for speed-first deployments @@ -21,48 +21,49 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a ## Flux.2 Klein 4B Workflows - + + Run this workflow on Comfy Cloud - - + Download the text-to-image workflow for Flux.2 Klein 4B. + - + + Run this workflow on Comfy Cloud - - + Download the image editing workflow using the 4B base model. + - + + Run this workflow on Comfy Cloud - - + Download the fast distilled 4B image editing workflow. + ## Flux.2 Klein 4B Model Downloads - + + Text encoder for 4B models. - - + Diffusion model (4B base). - - + Diffusion model (4B distilled). - - + VAE for 4B models. - + **4B Model Storage Location** ``` @@ -79,29 +80,32 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a ## Flux.2 Klein 9B Workflows - + + Run this workflow on Comfy Cloud - - + Download the text-to-image workflow for Flux.2 Klein 9B. + - + + Run this workflow on Comfy Cloud - - + Download the image editing workflow using the 9B base model. + - + + Run this workflow on Comfy Cloud - - + Download the fast distilled 9B image editing workflow. + ## Flux.2 Klein 9B Model Downloads @@ -109,22 +113,20 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a For diffusion models, please visit BFL's repo, accept the agreement, and then download the models. + Diffusion model (9B base). - Diffusion model (9B distilled). - - + Text encoder for 9B models. - - + VAE for 9B models. - + **9B Model Storage Location** ``` diff --git a/tutorials/flux/flux1-krea-dev.mdx b/tutorials/flux/flux1-krea-dev.mdx index 924f47547..6923ca645 100644 --- a/tutorials/flux/flux1-krea-dev.mdx +++ b/tutorials/flux/flux1-krea-dev.mdx @@ -30,7 +30,7 @@ Download the image or JSON below and drag it into ComfyUI to load the correspond ![Flux Krea Dev Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - + Run this workflow on Comfy Cloud diff --git a/tutorials/image/hidream/hidream-i1.mdx b/tutorials/image/hidream/hidream-i1.mdx index 685fdaf21..8984a8f8a 100644 --- a/tutorials/image/hidream/hidream-i1.mdx +++ b/tutorials/image/hidream/hidream-i1.mdx @@ -95,7 +95,7 @@ Model file save location ### HiDream-I1 Full Version Workflow - + Run this workflow on Comfy Cloud with zero setup @@ -138,7 +138,7 @@ Complete the workflow execution step by step ### HiDream-I1 Dev Version Workflow - + Run this workflow on Comfy Cloud with zero setup @@ -179,7 +179,7 @@ Complete the workflow execution step by step ### HiDream-I1 Fast Version Workflow - + Run this workflow on Comfy Cloud with zero setup diff --git a/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index ea45a64bf..703c71d71 100644 --- a/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -37,16 +37,16 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' **text_encoders** -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) +- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) **diffusion_models** -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/resolve/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) +- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **Model Storage Location** diff --git a/tutorials/image/omnigen/omnigen2.mdx b/tutorials/image/omnigen/omnigen2.mdx index 3a3f8d10c..b99d1d4a4 100644 --- a/tutorials/image/omnigen/omnigen2.mdx +++ b/tutorials/image/omnigen/omnigen2.mdx @@ -33,13 +33,13 @@ OmniGen2 is a powerful and efficient unified multimodal generation model with ap Since this article involves different workflows, the corresponding model files and installation locations are as follows. The download information for model files is also included in the corresponding workflows: **Diffusion Models** -- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/diffusion_models/omnigen2_fp16.safetensors) +- [omnigen2_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/diffusion_models/omnigen2_fp16.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/blob/main/split_files/vae/ae.safetensors) **Text Encoders** -- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/resolve/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) +- [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) File save location: ``` @@ -58,7 +58,7 @@ File save location: ### 1. Download Workflow File - + Open and run this workflow directly in Comfy Cloud. @@ -88,7 +88,7 @@ OmniGen2 has rich image editing capabilities and supports adding text to images ### 1. Download Workflow File - + Open and run this workflow directly in Comfy Cloud. diff --git a/tutorials/image/ovis/ovis-image.mdx b/tutorials/image/ovis/ovis-image.mdx index b65093002..2a2b546c0 100644 --- a/tutorials/image/ovis/ovis-image.mdx +++ b/tutorials/image/ovis/ovis-image.mdx @@ -35,15 +35,15 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' **text_encoders** -- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/text_encoders/ovis_2.5.safetensors) +- [ovis_2.5.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/text_encoders/ovis_2.5.safetensors) **diffusion_models** -- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/resolve/main/split_files/diffusion_models/ovis_image_bf16.safetensors) +- [ovis_image_bf16.safetensors](https://huggingface.co/Comfy-Org/Ovis-Image/blob/main/split_files/diffusion_models/ovis_image_bf16.safetensors) **vae** -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/resolve/main/split_files/vae/ae.safetensors) +- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) **Model Storage Location** diff --git a/tutorials/image/pixeldit/pixeldit.mdx b/tutorials/image/pixeldit/pixeldit.mdx index 04913bbf6..fcb0ec176 100644 --- a/tutorials/image/pixeldit/pixeldit.mdx +++ b/tutorials/image/pixeldit/pixeldit.mdx @@ -1,18 +1,18 @@ --- title: "PixelDiT ComfyUI Workflow Example" -description: "PixelDiT is NVIDIA's pixel-space diffusion transformer for 1024px text-to-image generation. It operates directly in pixel space — no VAE encode/decode required." +description: "PixelDiT is NVIDIA's pixel-space diffusion transformer for 1024px text-to-image generation. It operates directly in pixel space: no VAE encode/decode required." sidebarTitle: "PixelDiT" --- import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' -**PixelDiT** is NVIDIA's pixel-space diffusion transformer for 1024px text-to-image generation. Unlike traditional diffusion models that operate in latent space, PixelDiT generates images directly in pixel space using a dual-level DiT architecture — a patch-level DiT combined with a pixel-level DiT — with MM-DiT fusion for joint attention between text and image tokens. +**PixelDiT** is NVIDIA's pixel-space diffusion transformer for 1024px text-to-image generation. Unlike traditional diffusion models that operate in latent space, PixelDiT generates images directly in pixel space using a dual-level DiT architecture: a patch-level DiT combined with a pixel-level DiT: with MM-DiT fusion for joint attention between text and image tokens. **Model Highlights**: -- **VAE-free** — generates directly in pixel space; no traditional VAE encode/decode -- **Dual-level DiT** — patch-level DiT + pixel-level DiT for high-quality generation -- **Multi aspect ratio** — 1024px base resolution with support for several aspect ratios -- **~1.3B parameters** — efficient enough for consumer GPUs +- **VAE-free**: generates directly in pixel space; no traditional VAE encode/decode +- **Dual-level DiT**: patch-level DiT + pixel-level DiT for high-quality generation +- **Multi aspect ratio**: 1024px base resolution with support for several aspect ratios +- **~1.3B parameters**: efficient enough for consumer GPUs - **License**: NSCLv1 (non-commercial research/evaluation only) **Related Links**: @@ -23,26 +23,26 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' PixelDiT text-to-image workflow - + + Download JSON or search "PixelDiT" in Template Library - - + Open in cloud + The workflow consists of three main nodes: -1. **ResolutionSelector** — choose your desired output resolution -2. **Text to Image (PixelDiT) subgraph** — the core generation node with exposed controls for prompt, seed, model selection and resolution -3. **SaveImage** — saves the generated image +1. **ResolutionSelector**: choose your desired output resolution +2. **Text to Image (PixelDiT) subgraph**: the core generation node with exposed controls for prompt, seed, model selection and resolution +3. **SaveImage**: saves the generated image This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. - ### Workflow controls The exposed controls on the **Text to Image (PixelDiT)** subgraph node include: @@ -59,14 +59,14 @@ The exposed controls on the **Text to Image (PixelDiT)** subgraph node include: PixelDiT uses two model files: a text encoder and the diffusion model. - - gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT text encoder + + + gemma_2_2b_it_elm_bf16.safetensors: Gemma-2-2B-IT text encoder - - - pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px diffusion model + + pixeldit_1300m_1024px_bf16.safetensors: PixelDiT 1300M 1024px diffusion model - + ### Model storage location ``` diff --git a/tutorials/image/qwen/qwen-image-2512.mdx b/tutorials/image/qwen/qwen-image-2512.mdx index 3fd6f2940..24c3b16e9 100644 --- a/tutorials/image/qwen/qwen-image-2512.mdx +++ b/tutorials/image/qwen/qwen-image-2512.mdx @@ -37,7 +37,7 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - + Open in Comfy Cloud diff --git a/tutorials/image/qwen/qwen-image-edit-2511.mdx b/tutorials/image/qwen/qwen-image-edit-2511.mdx index 906df8a3c..0691c6a5f 100644 --- a/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -32,7 +32,7 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow below into ComfyUI to load it. - + Open in Comfy Cloud diff --git a/tutorials/image/qwen/qwen-image-edit.mdx b/tutorials/image/qwen/qwen-image-edit.mdx index 6cbc89c1f..ffd046d2d 100644 --- a/tutorials/image/qwen/qwen-image-edit.mdx +++ b/tutorials/image/qwen/qwen-image-edit.mdx @@ -44,7 +44,7 @@ After updating ComfyUI, you can find the workflow file from the templates, or dr Download JSON or search "image_qwen_image_edit" in Template Library - + Run this workflow on Cloud GPUs with zero setup diff --git a/tutorials/image/qwen/qwen-image-layered.mdx b/tutorials/image/qwen/qwen-image-layered.mdx index 945c965da..c337dc653 100644 --- a/tutorials/image/qwen/qwen-image-layered.mdx +++ b/tutorials/image/qwen/qwen-image-layered.mdx @@ -22,11 +22,11 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Qwen-Image-Layered workflow | -| +| | Download the JSON workflow file | | -| +| | Run ComfyUI online with zero setup | | diff --git a/tutorials/image/qwen/qwen-image.mdx b/tutorials/image/qwen/qwen-image.mdx index 82e081193..1dc15807a 100644 --- a/tutorials/image/qwen/qwen-image.mdx +++ b/tutorials/image/qwen/qwen-image.mdx @@ -49,7 +49,7 @@ Currently Qwen-Image has multiple ControlNet support options available: - + @@ -78,7 +78,6 @@ After updating ComfyUI, you can find the workflow file in the templates, or drag Distilled version - ### 2. Model Download **Available Models in ComfyUI** @@ -153,7 +152,7 @@ Qwen_image_distill This is a ControlNet model, so you can use it as normal ControlNet. - + @@ -207,7 +206,7 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow - + @@ -277,7 +276,7 @@ For the Inpaint model, it requires using the [Mask Editor](/interface/maskeditor ## Qwen Image Union ControlNet LoRA Workflow - + diff --git a/tutorials/image/z-image/z-image-turbo.mdx b/tutorials/image/z-image/z-image-turbo.mdx index 264770ce4..4757e5f6c 100644 --- a/tutorials/image/z-image/z-image-turbo.mdx +++ b/tutorials/image/z-image/z-image-turbo.mdx @@ -27,30 +27,30 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Z-Image-Turbo text-to-image workflow - + + Download the Z-Image-Turbo text-to-image workflow JSON file. - - + Run this workflow directly on ComfyUI Cloud. + ### Z-Image-Turbo model downloads - + + Text encoder for Z-Image-Turbo. - - + Diffusion model for Z-Image-Turbo. - - + VAE for Z-Image-Turbo. - + **Z-Image-Turbo Model Storage Location** ``` @@ -68,20 +68,19 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' This workflow uses the Z-Image-Turbo Fun Union ControlNet model to generate images with ControlNet guidance. It applies Canny edge detection to a reference image and uses the ControlNet to guide the generation process. - + + Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file. - Run this workflow directly on ComfyUI Cloud. - + ### Additional model for ControlNet - + ControlNet model patch for Z-Image-Turbo. - **Model Storage Location** ``` diff --git a/tutorials/partner-nodes/anthropic/claude.mdx b/tutorials/partner-nodes/anthropic/claude.mdx index 142b74e64..58d1ab986 100644 --- a/tutorials/partner-nodes/anthropic/claude.mdx +++ b/tutorials/partner-nodes/anthropic/claude.mdx @@ -44,7 +44,7 @@ In the corresponding template, we have built a prompt for analyzing and generati Refer to the numbered steps in the workflow image to complete the basic workflow execution: -1. In the `Load Image` node, load the image you need AI to interpret (optional — skip for text-only queries) +1. In the `Load Image` node, load the image you need AI to interpret (optional: skip for text-only queries) 2. In the `Anthropic Claude` node, modify `prompt` to set your conversation prompt, or change `model` to select a different Claude model 3. (Optional) Set the `system_prompt` to define the model's behavioral instructions 4. Click `Run` or press `Ctrl(cmd) + Enter` to execute the conversation @@ -53,4 +53,4 @@ Refer to the numbered steps in the workflow image to complete the basic workflow ### Additional Notes - Adjust `max_tokens` (up to 32000) and `temperature` (0.0 to 1.0) in the node's advanced options -- The `seed` parameter controls whether the node re-runs — results are non-deterministic regardless of seed value +- The `seed` parameter controls whether the node re-runs: results are non-deterministic regardless of seed value diff --git a/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 8a1bc5ea5..569670d5c 100644 --- a/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -32,20 +32,20 @@ Replace the background environment and relight your subject with a reference ima ### How It Works This workflow takes two input images: -- **Subject image** — your character or main subject (with background) -- **Reference environment** — the new scene, lighting, or style you want to apply +- **Subject image**: your character or main subject (with background) +- **Reference environment**: the new scene, lighting, or style you want to apply The node generates an output where the subject is naturally composited into the new environment with proper lighting, shadows, and perspective. ### Steps to Run -1. **Upload subject image** — use the first **LoadImage** node to load your subject image (e.g., a person wearing headphones) -2. **Upload reference image** — use the second **LoadImage** node to load the target environment or style -3. **Enter your prompt** — describe the scene in the **BeebleSwitchXImageEdit** node -4. **Set resolution** — choose output resolution -5. **Adjust seed** — modify the seed for different results -6. **Click Queue** — press `Ctrl+Enter` to generate -7. **View results** — the edited image appears in **SaveImage**, and a masked cutout is saved separately +1. **Upload subject image**: use the first **LoadImage** node to load your subject image (e.g., a person wearing headphones) +2. **Upload reference image**: use the second **LoadImage** node to load the target environment or style +3. **Enter your prompt**: describe the scene in the **BeebleSwitchXImageEdit** node +4. **Set resolution**: choose output resolution +5. **Adjust seed**: modify the seed for different results +6. **Click Queue**: press `Ctrl+Enter` to generate +7. **View results**: the edited image appears in **SaveImage**, and a masked cutout is saved separately ### Node Parameters @@ -76,14 +76,14 @@ This workflow takes a source video and a reference image, then applies the refer ### Steps to Run -1. **Upload source video** — use the **LoadVideo** node to load your input video -2. **Upload reference image** — use the **LoadImage** node to load the target environment/style -3. **Set video range** — the **Video Slice** node controls which frames to process (start, end, inclusive) -4. **Enter your prompt** — describe the desired environment and lighting in the **BeebleSwitchXVideoEdit** node -5. **Set resolution** — choose output resolution -6. **Adjust seed** — modify the seed for different results -7. **Click Queue** — press `Ctrl+Enter` -8. **View results** — the edited video is saved, and a separate alpha matte video is generated for compositing +1. **Upload source video**: use the **LoadVideo** node to load your input video +2. **Upload reference image**: use the **LoadImage** node to load the target environment/style +3. **Set video range**: the **Video Slice** node controls which frames to process (start, end, inclusive) +4. **Enter your prompt**: describe the desired environment and lighting in the **BeebleSwitchXVideoEdit** node +5. **Set resolution**: choose output resolution +6. **Adjust seed**: modify the seed for different results +7. **Click Queue**: press `Ctrl+Enter` +8. **View results**: the edited video is saved, and a separate alpha matte video is generated for compositing ### Outputs diff --git a/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 4920828af..ad9369233 100644 --- a/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -9,7 +9,7 @@ import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; ![Seedance 2.0 Real Human Cover](/images/tutorial/api_nodes/bytedance/seedance2_0/cover.png) -**Seedance 2.0 Real Human** is available directly inside ComfyUI. You can upload a portrait through a ComfyUI workflow, complete a one-time verification, and then generate videos featuring that real person with a consistent identity and authentic facial expressions — without leaving ComfyUI. +**Seedance 2.0 Real Human** is available directly inside ComfyUI. You can upload a portrait through a ComfyUI workflow, complete a one-time verification, and then generate videos featuring that real person with a consistent identity and authentic facial expressions: without leaving ComfyUI. With Real Human support in ComfyUI, you are no longer limited to virtual characters. Seedance 2.0's controllability and native audio-video sync now extend to real people, with identity and performance held together across a full clip, using the same node-based workflows you already use for other Seedance 2.0 generations. @@ -28,46 +28,46 @@ Watch the full walkthrough covering verification and both Seedance 2.0 Real Huma ## What's different -- **Real people, cinematic control** — Direct camera moves, lighting, and scene changes while keeping the subject believable -- **Identity consistency** — Facial features and overall appearance remain stable across motion, camera moves, and transitions -- **Native audio-video sync** — Video and audio (dialogue, ambient, music) are generated together with more reliable lip-sync and expressions -- **Multi-reference directing** — Mix text with up to **9 images**, **3 videos**, and **3 audio clips** to lock in look, motion, and rhythm +- **Real people, cinematic control**: Direct camera moves, lighting, and scene changes while keeping the subject believable +- **Identity consistency**: Facial features and overall appearance remain stable across motion, camera moves, and transitions +- **Native audio-video sync**: Video and audio (dialogue, ambient, music) are generated together with more reliable lip-sync and expressions +- **Multi-reference directing**: Mix text with up to **9 images**, **3 videos**, and **3 audio clips** to lock in look, motion, and rhythm ## How verification works in ComfyUI -Verification is required before any real person can appear in a Seedance 2.0 generation. It prevents impersonation and unauthorized use of someone's likeness and is handled by ByteDance, which complies with emerging AI transparency regulations. The entire flow is driven from ComfyUI through the **ByteDance Create Image/Video Asset** node — you only need to leave ComfyUI briefly to complete the liveness check on your phone or browser. +Verification is required before any real person can appear in a Seedance 2.0 generation. It prevents impersonation and unauthorized use of someone's likeness and is handled by ByteDance, which complies with emerging AI transparency regulations. The entire flow is driven from ComfyUI through the **ByteDance Create Image/Video Asset** node: you only need to leave ComfyUI briefly to complete the liveness check on your phone or browser. ![ByteDance Create Image/Video Asset nodes used to generate Group ID and Asset ID](/images/tutorial/api_nodes/bytedance/seedance2_0/create_asset_nodes.png) ### First run (verification required) 1. In ComfyUI, add the **ByteDance Create Image/Video Asset** node to your workflow and upload a portrait image -2. Run the workflow — ComfyUI returns a verification link from ByteDance +2. Run the workflow: ComfyUI returns a verification link from ByteDance 3. Open the link on your phone or browser and complete the liveness check (less than 1 minute) -4. Once verification passes, return to ComfyUI — the node output now contains the verified asset +4. Once verification passes, return to ComfyUI: the node output now contains the verified asset #### Verification flow screenshots -**Step 1 — Upload and run to start verification** +**Step 1: Upload and run to start verification** ![Upload your portrait and run to start liveness verification](/images/tutorial/api_nodes/bytedance/seedance2_0/workflow_1.png) -**Step 2 — Get the verification link** +**Step 2: Get the verification link** ![Workflow run returns a verification link to open in browser](/images/tutorial/api_nodes/bytedance/seedance2_0/workflow_2.png) -**Step 3 — Complete liveness check in the H5 page** +**Step 3: Complete liveness check in the H5 page** ![Complete identity authentication in the H5 verification page](/images/tutorial/api_nodes/bytedance/seedance2_0/workflow_3.png) -**Step 4 — Return to ComfyUI and save IDs (no extra credits)** +**Step 4: Return to ComfyUI and save IDs (no extra credits)** ![After verification, ComfyUI outputs Asset ID and Group ID to save for reuse](/images/tutorial/api_nodes/bytedance/seedance2_0/workflow_4.png) After real human verification passes, the node exposes two IDs you can use in downstream Seedance 2.0 video nodes: -- **Group ID** — represents the verified person. Save this so you can reuse the same identity later without re-verifying -- **Asset ID** — represents this specific image. Wire it into a Seedance 2.0 video generation node to produce a video of the verified person; it can be reused across multiple workflows +- **Group ID**: represents the verified person. Save this so you can reuse the same identity later without re-verifying +- **Asset ID**: represents this specific image. Wire it into a Seedance 2.0 video generation node to produce a video of the verified person; it can be reused across multiple workflows ### Future uploads (no verification) @@ -93,7 +93,7 @@ Use a verified portrait (plus optional additional reference images, videos, or a Seedance 2.0 Real Human R2V workflow preview - + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. @@ -120,7 +120,7 @@ Provide a verified starting frame and ending frame to generate the video between Seedance 2.0 Real Human FLF2V workflow preview - + Try the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow instantly on Comfy Cloud. diff --git a/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index c785caaa9..6eeacb22d 100644 --- a/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -11,11 +11,11 @@ Seedance 2.0 is ByteDance's next-generation multimodal video generation model, n ## Key capabilities -- **Multimodal in** — Prompt with text plus images, video, and audio -- **Audio-video sync** — Generate video and audio together -- **Directing control** — Camera moves and shot pacing stay controllable -- **Consistency** — Keep characters and scenes stable across a clip -- **Editing + extend** — Edit footage or extend clips without starting over +- **Multimodal in**: Prompt with text plus images, video, and audio +- **Audio-video sync**: Generate video and audio together +- **Directing control**: Camera moves and shot pacing stay controllable +- **Consistency**: Keep characters and scenes stable across a clip +- **Editing + extend**: Edit footage or extend clips without starting over With these capabilities, Seedance 2.0 fits a wide range of use cases including advertising, short-form drama, film previsualization, product showcases, virtual character content, and assets for games and animation. @@ -31,7 +31,7 @@ Generate a video from a text prompt, with Seedance 2.0 handling scene, motion, a Seedance 2.0 Text-to-Video workflow preview - + Try the Text-to-Video workflow instantly on Comfy Cloud. @@ -46,7 +46,7 @@ Use reference images, video, or audio to guide look, motion, and rhythm while ke Seedance 2.0 Reference-to-Video workflow preview - + Try the Reference-to-Video workflow instantly on Comfy Cloud. @@ -82,7 +82,7 @@ Provide a starting frame and ending frame, and Seedance 2.0 generates the motion Seedance 2.0 FLF2V workflow preview - + Try the First-Last-Frame-to-Video workflow instantly on Comfy Cloud. @@ -167,7 +167,7 @@ Seedance 2.0 **supports both** **real-person portraits** (imagery of actual peop Use the Seedance 2.0 node like any other generator: upload your **AI-generated character image** (or an equivalent reference), wire it up, and run. -After submission the service classifies the asset. **If it is flagged as protected real-person-related content**, the node returns an error — switch to the **real-person portrait** workflow below instead. Once the backend treats it as an AI portrait and the check clears, use it directly as reference with no extra verification steps. +After submission the service classifies the asset. **If it is flagged as protected real-person-related content**, the node returns an error: switch to the **real-person portrait** workflow below instead. Once the backend treats it as an AI portrait and the check clears, use it directly as reference with no extra verification steps. ### Real-person portrait @@ -176,11 +176,11 @@ Use the **Real Human** workflow (requires **ByteDance Create Image/Video Asset** **First use (verification required)** 1. Upload portrait imagery via **ByteDance Create Image/Video Asset**. -2. Run the workflow — a verification link is generated. +2. Run the workflow: a verification link is generated. 3. Open the link on your phone or browser and finish liveness (typically under ~30 seconds). 4. After verification succeeds you receive two IDs: - - **Group ID** — the verified individual; retain it when uploading the same person again. - - **Asset ID** — bound to this portrait; connect it for video generation and reuse across runs. + - **Group ID**: the verified individual; retain it when uploading the same person again. + - **Asset ID**: bound to this portrait; connect it for video generation and reuse across runs. **New photo or video of the same person (no repeat liveness)** diff --git a/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index e660c7c28..0683d1737 100644 --- a/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -14,10 +14,10 @@ Seedream 5.0 lite is the latest image generation model from BytePlus. It is the ## What's new in Seedream 5.0 lite -- **Web-connected retrieval** — The model pulls from live web data during generation, allowing creations to be tied to real-world trends, current events, and up-to-date knowledge -- **Smarter instruction following** — Handles complex, multi-step instructional prompts far more reliably. Multi-image inputs, local editing instructions, and batch style requests all land closer to the intended output -- **Stronger consistency** — Character identity, product appearance, style, and lighting all hold up much better across a set of generated images -- **Deeper world knowledge** — Expanded knowledge base makes the model useful for technical illustration, educational diagrams, and knowledge-heavy creative work +- **Web-connected retrieval**: The model pulls from live web data during generation, allowing creations to be tied to real-world trends, current events, and up-to-date knowledge +- **Smarter instruction following**: Handles complex, multi-step instructional prompts far more reliably. Multi-image inputs, local editing instructions, and batch style requests all land closer to the intended output +- **Stronger consistency**: Character identity, product appearance, style, and lighting all hold up much better across a set of generated images +- **Deeper world knowledge**: Expanded knowledge base makes the model useful for technical illustration, educational diagrams, and knowledge-heavy creative work ## Seedream 5.0 lite image edit workflow diff --git a/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/tutorials/partner-nodes/google/gemini-omni-flash.mdx index a63d908e7..51ae47d33 100644 --- a/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -14,7 +14,7 @@ Gemini Omni Flash is Google DeepMind's high-quality, cost-efficient video genera ## What Gemini Omni Flash offers -- **Conversational video editing**: Refine and edit videos using natural language — swap characters, relight scenes, alter angles, add or remove objects while maintaining original audio and video tracks +- **Conversational video editing**: Refine and edit videos using natural language: swap characters, relight scenes, alter angles, add or remove objects while maintaining original audio and video tracks - **Multimodal input**: Combine text, images, and video inputs to guide generation. Natively generates synchronized audio with every video output - **World knowledge and simulation**: Combines physics understanding with Gemini's knowledge of history, science, and cultural context, enabling meaningful storytelling beyond photorealism - **Text and action synchronization**: Render legible text and graphics directly into video, syncing kinetic typography with on-screen movements diff --git a/tutorials/partner-nodes/google/nano-banana-2.mdx b/tutorials/partner-nodes/google/nano-banana-2.mdx index 2ded81bda..d6f0cb7b5 100644 --- a/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -14,28 +14,28 @@ Nano Banana 2 is now available in ComfyUI through Partner Nodes. This release fu ## What's new in Nano Banana 2 -- **Pro-level quality at Flash speed** — Richer textures, sharper detail, stronger lighting, and better prompt adherence without the latency tradeoff -- **Search-grounded world knowledge** — Intelligently triggers Google Image Search grounding to improve factual accuracy for lesser-known landmarks, niche objects, or real-world details -- **Subject consistency** — Maintain character resemblance of up to 5 characters and fidelity of up to 14 objects in a single workflow -- **Configurable thinking levels** — Control how much the model "thinks" before generating (Minimal for fast exploration, High/Dynamic for complex layouts) -- **Precision text rendering and translation** — Generate accurate, legible text for marketing mockups, greeting cards, and localized assets +- **Pro-level quality at Flash speed**: Richer textures, sharper detail, stronger lighting, and better prompt adherence without the latency tradeoff +- **Search-grounded world knowledge**: Intelligently triggers Google Image Search grounding to improve factual accuracy for lesser-known landmarks, niche objects, or real-world details +- **Subject consistency**: Maintain character resemblance of up to 5 characters and fidelity of up to 14 objects in a single workflow +- **Configurable thinking levels**: Control how much the model "thinks" before generating (Minimal for fast exploration, High/Dynamic for complex layouts) +- **Precision text rendering and translation**: Generate accurate, legible text for marketing mockups, greeting cards, and localized assets ## Nano Banana 2 image edit workflow - + + Open in Comfy Cloud - - + Download JSON or search "Nano Banana 2" in Template Library + ![Nano Banana 2 workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_google_nano_banana2_image_edit-1.webp) Get the example input image for this workflow - ### Pro-level quality showcase ![Quality comparison](https://substack-post-media.s3.amazonaws.com/public/images/1f92ae2e-14d8-4a3b-9ed7-e57dacd2584f_1825x1696.png) @@ -80,7 +80,7 @@ Nano Banana 2 is now available in ComfyUI through Partner Nodes. This release fu | Model | Best for | |-------|----------| -| **Nano Banana 2** | Most creators — fast enough to explore, smart enough to refine, polished enough to ship | +| **Nano Banana 2** | Most creators: fast enough to explore, smart enough to refine, polished enough to ship | | **Nano Banana Pro** | Maximum factual precision and high-stakes edge cases | ## Get started @@ -91,5 +91,5 @@ Nano Banana 2 is now available in ComfyUI through Partner Nodes. This release fu 4. Choose your resolution, set your aspect ratio, and decide whether the prompt needs deeper reasoning -The new model version has also been added to the previous "Nano Banana Pro" node. If you have old workflows with this node, you don't need to rebuild them — just reselect the model version. +The new model version has also been added to the previous "Nano Banana Pro" node. If you have old workflows with this node, you don't need to rebuild them: just reselect the model version. diff --git a/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index 98ac05114..d398c0aac 100644 --- a/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -11,10 +11,10 @@ The [Grok Imagine Video 1.5](/built-in-nodes/GrokVideoNode) Partner Node enables The node supports two model variants selected via its `model` parameter: -- **`grok-imagine-video`** — the previous generation model, supports optional image input -- **`grok-imagine-video-1.5`** — the latest model, **always requires** an input image and supports 1080p output +- **`grok-imagine-video`**: the previous generation model, supports optional image input +- **`grok-imagine-video-1.5`**: the latest model, **always requires** an input image and supports 1080p output -Both variants generate **native audio** — sound effects, ambience, and dialogue are synthesized in the same pass, with no separate audio pipeline needed. Video duration ranges from **1 to 15 seconds**. +Both variants generate **native audio**: sound effects, ambience, and dialogue are synthesized in the same pass, with no separate audio pipeline needed. Video duration ranges from **1 to 15 seconds**. @@ -35,23 +35,22 @@ Both variants generate **native audio** — sound effects, ambience, and dialogu Get the example input image for this workflow. - ### Workflow Overview This workflow uses three nodes: -- **LoadImage** — provides the starting image frame -- **GrokVideoNode** — the core node configured with the `grok-imagine-video-1.5` model -- **SaveVideo** — saves the generated video with native audio +- **LoadImage**: provides the starting image frame +- **GrokVideoNode**: the core node configured with the `grok-imagine-video-1.5` model +- **SaveVideo**: saves the generated video with native audio ### Steps to Run -1. **Upload a starting image** — use the **LoadImage** node to load your reference image -2. **Enter your prompt** — describe the motion, atmosphere, and scene dynamics in the **GrokVideoNode** node -3. **Select model** — ensure `grok-imagine-video-1.5` is selected -4. **Set resolution** — choose output resolution (`720p` recommended) -5. **Set duration** — choose the video length in seconds -6. **Set seed** — control reproducibility of results -7. **Click Queue** — press `Ctrl+Enter` to generate +1. **Upload a starting image**: use the **LoadImage** node to load your reference image +2. **Enter your prompt**: describe the motion, atmosphere, and scene dynamics in the **GrokVideoNode** node +3. **Select model**: ensure `grok-imagine-video-1.5` is selected +4. **Set resolution**: choose output resolution (`720p` recommended) +5. **Set duration**: choose the video length in seconds +6. **Set seed**: control reproducibility of results +7. **Click Queue**: press `Ctrl+Enter` to generate ### Output diff --git a/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 0a2649889..f2e595866 100644 --- a/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -39,7 +39,7 @@ Generate cinematic video from a single image with strong aesthetic control and m - + Try the Image-to-Video workflow instantly on Comfy Cloud. @@ -52,7 +52,6 @@ Generate cinematic video from a single image with strong aesthetic control and m Get the example input image for this workflow. - ## HappyHorse 1.0 text-to-video Generate cinematic video from pure text prompts, with multi-shot sequencing and refined visual atmosphere. @@ -64,7 +63,7 @@ Generate cinematic video from pure text prompts, with multi-shot sequencing and > - + Try the Text-to-Video workflow instantly on Comfy Cloud. @@ -79,7 +78,7 @@ Generate cinematic video from pure text prompts, with multi-shot sequencing and Use a reference subject to drive video generation, preserving identity across cinematic multi-shot sequences. - + Try the Reference-to-Video workflow instantly on Comfy Cloud. @@ -89,20 +88,20 @@ Use a reference subject to drive video generation, preserving identity across ci ![HappyHorse 1.0 Reference-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_r2v-1.webp) + Get the example car reference image. - Get the example person reference image. - + ## HappyHorse 1.0 video edit Transform existing footage or replace/insert subjects while preserving motion and composition with V2V and SV2V editing workflows. - + Try the Video Edit workflow instantly on Comfy Cloud. @@ -112,10 +111,11 @@ Transform existing footage or replace/insert subjects while preserving motion an ![HappyHorse 1.0 Video Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_video_edit-1.webp) + Get the example input image for the video edit workflow. - Get the example input video for the video edit workflow. + \ No newline at end of file diff --git a/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 5c95f47d1..aea8038e9 100644 --- a/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -51,7 +51,6 @@ Enter a text description to generate a 3D model in one click. The model accurate Run on Comfy Cloud - ## Image-to-3D workflow Upload one or more images to generate high-quality 3D models. Support for 2–4 multi-view images improves geometry and material fidelity. @@ -61,7 +60,6 @@ Upload one or more images to generate high-quality 3D models. Support for 2–4 Run on Comfy Cloud -
Input materials @@ -83,7 +81,6 @@ Provide multiple view images (front, back, side) to generate 3D models with impr Run on Comfy Cloud -
Input materials @@ -115,38 +112,39 @@ Split a complete 3D model into meaningful structural parts, such as armor pieces Hunyuan 3D parts decomposition workflow preview - + + Run the workflow - Get the JSON file - + ### UV unwrapping Automatically generate UV layouts for supported 3D models, turning raw geometry into assets that are much easier to texture. Instead of manually cutting seams and organizing UV islands, creators can move more quickly into painting, baking, and material work with a cleaner starting point. Hunyuan 3D UV unwrapping workflow preview - + + Run the workflow - Get the JSON file - + ### Smart topology Convert dense geometry into cleaner meshes with more organized edge flow, helping generated models become easier to optimize and reuse in real production pipelines. This is especially useful when preparing assets for game engines, real-time rendering, or any workflow that benefits from lower-density, better-structured geometry. Hunyuan 3D smart topology workflow preview - + + Run the workflow - Get the JSON file + \ No newline at end of file diff --git a/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index d3ff29a5b..5630c754b 100644 --- a/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -51,7 +51,6 @@ Enter a text description to generate a 3D model in one click. The model accurate Run on Comfy Cloud - ## Image-to-3D workflow Upload one or more images to generate high-quality 3D models. Support for 2–4 multi-view images improves geometry and material fidelity. @@ -61,7 +60,6 @@ Upload one or more images to generate high-quality 3D models. Support for 2–4 Run on Comfy Cloud -
Input materials @@ -83,7 +81,6 @@ Provide multiple view images (front, back, side) to generate 3D models with impr Run on Comfy Cloud -
Input materials diff --git a/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 7fe08ea6c..dc7834b0a 100644 --- a/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -16,13 +16,14 @@ Ideogram 4.0 is the latest text-to-image model from Ideogram, offering superior Ideogram 4.0 Text-to-Image workflow preview - + + Open in Comfy Cloud - - + Download JSON or search "Ideogram v4: Text to Image (API)" in Template Library + ![Ideogram 4.0 Example Output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_ideogram_v4_t2i.png) *Example output from the Ideogram 4.0 API* @@ -31,18 +32,18 @@ Ideogram 4.0 is the latest text-to-image model from Ideogram, offering superior The Ideogram V4 node supports two prompt formats: -1. **Natural Language** — Simple and intuitive, best for quick ideas. -2. **Structured JSON** — Offers fine-grained control over style palette, element placement (`bbox`), and literal text rendering in the image. Recommended for poster-like compositions. +1. **Natural Language**: Simple and intuitive, best for quick ideas. +2. **Structured JSON**: Offers fine-grained control over style palette, element placement (`bbox`), and literal text rendering in the image. Recommended for poster-like compositions. The default workflow example uses a structured JSON prompt that includes: -- `high_level_description` — Overall scene description -- `style_description` — Aesthetic, lighting, medium, color palette -- `compositional_deconstruction` — Per-element bounding boxes, descriptions, and color palettes +- `high_level_description`: Overall scene description +- `style_description`: Aesthetic, lighting, medium, color palette +- `compositional_deconstruction`: Per-element bounding boxes, descriptions, and color palettes ### Workflow Steps 1. Download and drag the workflow file into ComfyUI -2. The `Ideogram V4` node already contains a default structured JSON prompt — you can modify it or write plain natural language +2. The `Ideogram V4` node already contains a default structured JSON prompt: you can modify it or write plain natural language 3. Click `Run` or use shortcut `Ctrl(cmd) + Enter` to generate the image 4. After the API returns results, view the generated image in the `Save Image` node. Images are saved to the `ComfyUI/output/` directory @@ -57,8 +58,8 @@ The workflow includes a built-in **Gemini** node (Gemini 3.1 Pro) to help you bu ## Additional Notes -- **Negative prompts are not needed** — The Ideogram 4.0 model uses asymmetric classifier-free guidance, where the unconditional pass drops text tokens, rather than requiring a separate negative prompt string. -- **Model-assisted prompt building** — Consider using the **Ideogram 4 Prompt Builder** (KJNodes) for visual control over prompt construction on ComfyUI. +- **Negative prompts are not needed**: The Ideogram 4.0 model uses asymmetric classifier-free guidance, where the unconditional pass drops text tokens, rather than requiring a separate negative prompt string. +- **Model-assisted prompt building**: Consider using the **Ideogram 4 Prompt Builder** (KJNodes) for visual control over prompt construction on ComfyUI. ## Live Conversation with Ideogram & ComfyOrg diff --git a/tutorials/partner-nodes/kling/kling-motion-control.mdx b/tutorials/partner-nodes/kling/kling-motion-control.mdx index 16de4a9a3..add428521 100644 --- a/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -42,14 +42,14 @@ The `character_orientation` parameter determines how the model interprets spatia Kling 2.6 Motion Control workflow preview + Run the Kling 2.6 Motion Control workflow on Comfy Cloud. - Download the workflow JSON file for local use. - +
Input materials diff --git a/tutorials/partner-nodes/luma/luma-uni-1.mdx b/tutorials/partner-nodes/luma/luma-uni-1.mdx index fa35daacb..92dcc8f31 100644 --- a/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -32,10 +32,10 @@ Uni-1 excels across a wide range of tasks: - **Illustration & stylized art** with strong aesthetic control - **Old photo restoration** and vintage reproduction - **Surreal and conceptual** compositions -- **Text rendering** — readable text inside images, great for infographics and posters +- **Text rendering**: readable text inside images, great for infographics and posters - **Image editing** and multi-turn refinement - **Reference-guided generation** with identity preservation -- **Multi-panel output** — consistent characters/scenes across multiple frames +- **Multi-panel output**: consistent characters/scenes across multiple frames ## The core distinction: Create vs Modify @@ -58,7 +58,7 @@ When in doubt: Luma Uni-1 Image Create workflow preview - + Try the Image Create workflow instantly on Comfy Cloud. @@ -83,7 +83,7 @@ Download this sample input image to try the workflow: Luma Uni-1 Image Edit workflow preview - + Try the Image Edit workflow instantly on Comfy Cloud. diff --git a/tutorials/partner-nodes/meshy/meshy-6.mdx b/tutorials/partner-nodes/meshy/meshy-6.mdx index 36767485d..bc1e47f9b 100644 --- a/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -27,46 +27,45 @@ Meshy 6 is the latest generation of Meshy's 3D model generation technology, feat Generate 3D models directly from text descriptions using Meshy 6. - + + Run the text-to-model workflow instantly on Comfy Cloud. - Download the workflow JSON file for local use. - + ![Meshy 6 Text-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_text_to_model-1.webp) ## Image-to-Model Workflow Convert 2D images into detailed 3D models with Meshy 6's image-to-3D capabilities. - + + Run the image-to-model workflow instantly on Comfy Cloud. - Download the workflow JSON file for local use. - + ![Meshy 6 Image-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_image_to_model-1.webp) Get the example input image for this workflow. - ## Multi-view to Model Workflow Generate 3D models from multiple view images for more accurate geometry and texture reconstruction. - + + Run the multi-view workflow instantly on Comfy Cloud. - Download the workflow JSON file for local use. - + ![Meshy 6 Multi-view to Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_multi_image_to_model-1.webp) diff --git a/tutorials/partner-nodes/openai/gpt-image-2.mdx b/tutorials/partner-nodes/openai/gpt-image-2.mdx index dad603bb0..90336e249 100644 --- a/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -35,7 +35,7 @@ GPT-Image-2 is selected as a `model` option on the **OpenAI GPT Image 1.5** node Generate an image from a text prompt with GPT-Image-2's reasoning-driven composition. - + Try the Text-to-Image workflow instantly on Comfy Cloud. @@ -53,7 +53,7 @@ Generate an image from a text prompt with GPT-Image-2's reasoning-driven composi Edit an input image with high structural fidelity at up to 2K resolution. - + Try the Image Edit workflow instantly on Comfy Cloud. @@ -67,7 +67,6 @@ Edit an input image with high structural fidelity at up to 2K resolution. Get the example input image for this workflow - ![GPT-Image-2 Image-to-Image example](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_i2i_1.jpg) ![GPT-Image-2 Image Edit example 1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_image_edit_1.jpg) @@ -78,7 +77,7 @@ Edit an input image with high structural fidelity at up to 2K resolution. ### Reasoning-driven generation -GPT-Image-2 plans the composition before rendering. This makes it well suited for prompts that have historically broken image models — for example, a poster with a seven-item bulleted list in 11pt Helvetica, centered — and produces clean output for dense text, small UI elements, iconography, infographics, maps, and slides. +GPT-Image-2 plans the composition before rendering. This makes it well suited for prompts that have historically broken image models: for example, a poster with a seven-item bulleted list in 11pt Helvetica, centered: and produces clean output for dense text, small UI elements, iconography, infographics, maps, and slides. ### Image editing that preserves what matters @@ -92,5 +91,5 @@ The model can return up to **eight distinct images from a single prompt** while ## Hybrid Pipelines -GPT-Image-2 slots naturally into hybrid pipelines: use it for the text-heavy hero frame, then hand off to your local models for upscaling, stylization, or video generation — the best model for each step, in one graph. +GPT-Image-2 slots naturally into hybrid pipelines: use it for the text-heavy hero frame, then hand off to your local models for upscaling, stylization, or video generation: the best model for each step, in one graph. diff --git a/tutorials/partner-nodes/recraft/recraft-v4.mdx b/tutorials/partner-nodes/recraft/recraft-v4.mdx index f4d3802b8..47d3758c8 100644 --- a/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -14,23 +14,24 @@ Recraft V4 is a new image generation model built for professional design work. I ## What's new in V4 -- **Raster + vector in one model** — photorealistic images, stylized illustrations, and production-ready SVGs from the same model -- **Better detail** — cleaner textures, more accurate lighting, tighter geometry. Holds up in complex scenes -- **Faces and anatomy that work** — reliable proportions and expressions, even with multiple subjects or action poses -- **Text rendering** — legible text for signage, packaging, infographics. Handles short and mid-length phrases well -- **SVG output** — editable vectors with clean lines and proper color segmentation. Production-quality vector output -- **Prompt accuracy** — follows detailed creative direction closely, including spatial relationships, materials, and reflections -- **V4 Pro** — same model, higher resolution. Outputs at 2048×2048 for print and large-format work +- **Raster + vector in one model**: photorealistic images, stylized illustrations, and production-ready SVGs from the same model +- **Better detail**: cleaner textures, more accurate lighting, tighter geometry. Holds up in complex scenes +- **Faces and anatomy that work**: reliable proportions and expressions, even with multiple subjects or action poses +- **Text rendering**: legible text for signage, packaging, infographics. Handles short and mid-length phrases well +- **SVG output**: editable vectors with clean lines and proper color segmentation. Production-quality vector output +- **Prompt accuracy**: follows detailed creative direction closely, including spatial relationships, materials, and reflections +- **V4 Pro**: same model, higher resolution. Outputs at 2048×2048 for print and large-format work ## Recraft V4 text to image workflow - + + Open in Comfy Cloud - - + Download JSON or search "Recraft V4 Text to Image" in Template Library + ![Recraft V4 Text to Image workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_t2i-1.webp) @@ -51,29 +52,30 @@ Recraft V4 is a new image generation model built for professional design work. I *Full outfit shot: electric blue and coral camo hoodie, matching camo joggers, white G-Shock watch, matte black techwear crossbody harness bag, and New Balance 550s in a cream colorway. Concrete parking garage backdrop, harsh fluorescent overhead lighting casting sharp shadows.* ![Action shot](https://substack-post-media.s3.amazonaws.com/public/images/a8643f80-8cc4-4246-a95a-bed1f9ee10f1_2560x1664.png) -*Fish-eye lens, shot from inside an empty concrete pool — a guy in a shearling bomber and transparent PVC shell is mid-kickflip at the lip of the pool directly above the camera, board separating from his feet, his layered outfit catching air and fanning open revealing the hockey jersey underneath.* +*Fish-eye lens, shot from inside an empty concrete pool: a guy in a shearling bomber and transparent PVC shell is mid-kickflip at the lip of the pool directly above the camera, board separating from his feet, his layered outfit catching air and fanning open revealing the hockey jersey underneath.* ## Recraft V4 text to vector workflow Recraft V4 can generate production-ready SVG vector outputs directly. This is useful for logos, icons, brand assets, or anything that needs to scale. SVG outputs are compatible with Illustrator, Figma, and Sketch. - + + Open in Comfy Cloud - - + Download JSON or search "Recraft V4 Text to Vector" in Template Library + ![Recraft V4 Text to Vector workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_recraft_v4_text_to_vector-1.webp) ### Vector showcase ![Rooftop illustration](https://substack-post-media.s3.amazonaws.com/public/images/571c5214-5c7f-43f9-81c4-80cad1e5578b_1792x1024.png) -*A group of friends lounging on a rooftop at sunset, string lights overhead, one person tending a small grill, another strumming a ukulele, a dog napping on a blanket beside a cooler — warm, relaxed, retro lifestyle illustration* +*A group of friends lounging on a rooftop at sunset, string lights overhead, one person tending a small grill, another strumming a ukulele, a dog napping on a blanket beside a cooler: warm, relaxed, retro lifestyle illustration* ![Boombox illustration](https://substack-post-media.s3.amazonaws.com/public/images/748f04c7-50b9-433d-93cb-c619984c8e89_1792x1024.png) -*A giant neon boombox blasting music, with lightning bolts and sound waves exploding out of the speakers, melting cassette tapes swirling around it, and a pair of oversized headphones draped on top — bold, chaotic, and electrifying* +*A giant neon boombox blasting music, with lightning bolts and sound waves exploding out of the speakers, melting cassette tapes swirling around it, and a pair of oversized headphones draped on top: bold, chaotic, and electrifying* ![Kitten sticker](https://substack-post-media.s3.amazonaws.com/public/images/3ddce7b4-8f37-4256-bd93-2e5796dfacca_1024x1024.png) *Adorable kitten sticker, the kitten is using ComfyUI on a little tiny laptop* diff --git a/tutorials/partner-nodes/rodin/model-generation.mdx b/tutorials/partner-nodes/rodin/model-generation.mdx index 5a5121655..0a886b721 100644 --- a/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/tutorials/partner-nodes/rodin/model-generation.mdx @@ -28,7 +28,7 @@ Generate a 3D model from a single image input with Rodin. Single-view Model Generation (Json Format) - + Open in Comfy Cloud @@ -73,7 +73,7 @@ You can modify the single-view workflow to a multi-view workflow, or directly do Multi-view Model Generation (Json Format) - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/topaz/astra-2.mdx b/tutorials/partner-nodes/topaz/astra-2.mdx index 066e6eba1..5bcaf7766 100644 --- a/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/tutorials/partner-nodes/topaz/astra-2.mdx @@ -19,9 +19,9 @@ It builds on Astra 1 with clearer control over how detail and stylization are pr ## Key capabilities -- **Adds texture and detail** — Not only sharpening. -- **Steerable look** — Creativity, sharpness, optional prompt guidance. -- **Stylized & wides** — Where inferred detail matters. +- **Adds texture and detail**: Not only sharpening. +- **Steerable look**: Creativity, sharpness, optional prompt guidance. +- **Stylized & wides**: Where inferred detail matters. ## In ComfyUI @@ -38,14 +38,14 @@ Optional **Prompt**: max **450** frames (~15 s @ 30 fps). Empty prompt: up t Astra 2 workflow preview - + + Open the workflow instantly on Comfy Cloud. - Download the workflow JSON for local ComfyUI. - +
Input materials @@ -60,4 +60,4 @@ Download this sample input video to try the workflow: ## Related -- [Video upscaling in ComfyUI](/tutorials/utility/video-upscale) — guidance on choosing an upscaling approach. +- [Video upscaling in ComfyUI](/tutorials/utility/video-upscale): guidance on choosing an upscaling approach. diff --git a/tutorials/partner-nodes/tripo/model-generation.mdx b/tutorials/partner-nodes/tripo/model-generation.mdx index c802f0849..377ef512d 100644 --- a/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/tutorials/partner-nodes/tripo/model-generation.mdx @@ -27,7 +27,7 @@ Generate a 3D model from a text prompt with Tripo's text-to-model capabilities. Tripo Text-to-Model workflow preview - + Try the Text-to-Model workflow instantly on Comfy Cloud. @@ -55,7 +55,7 @@ Generate a 3D model from a single image input. Tripo Image-to-Model workflow preview - + Try the Image-to-Model workflow instantly on Comfy Cloud. @@ -97,7 +97,7 @@ Generate a 3D model from multiple view images for enhanced accuracy. Tripo Multi-view Model Generation workflow preview - + Try the Multiview-to-Model workflow instantly on Comfy Cloud. diff --git a/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/tutorials/partner-nodes/tripo/tripo-3-1.mdx index 9dd2c73ea..cb2c0870c 100644 --- a/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -1,5 +1,5 @@ --- -title: "Tripo 3.1 — High-Detail 3D Asset Generation ComfyUI Official Guide" +title: "Tripo 3.1: High-Detail 3D Asset Generation ComfyUI Official Guide" description: "Learn how to generate high-detail 3D assets with Tripo 3.1 in ComfyUI via Partner Nodes, featuring high-density geometry and PBR-ready materials for production use." sidebarTitle: "Tripo 3.1" --- @@ -16,17 +16,17 @@ Tripo 3.1 is the latest model version available through Tripo Partner Nodes in C Compared to previous Tripo model versions, 3.1 provides the following improvements: -- **High-density geometry** — Sharper edges, cleaner silhouettes, and more precise forms for production-level 3D assets. -- **PBR-ready materials** — More reliable lighting response across different render environments. -- **Hero asset quality** — Stable detail retention in close-up shots. -- **Cross-scenario reuse** — A single model can work across game, marketing, and render pipelines without regeneration. +- **High-density geometry**: Sharper edges, cleaner silhouettes, and more precise forms for production-level 3D assets. +- **PBR-ready materials**: More reliable lighting response across different render environments. +- **Hero asset quality**: Stable detail retention in close-up shots. +- **Cross-scenario reuse**: A single model can work across game, marketing, and render pipelines without regeneration. ## Use Cases -- **Game hero assets** — For close-up camera shots and promotional key art. -- **Marketing visuals** — Where geometric fidelity and material quality are critical. -- **Product visualization** — Cleaner shape definition in studio-style renders. -- **3D printing preparation** — Better structural clarity. +- **Game hero assets**: For close-up camera shots and promotional key art. +- **Marketing visuals**: Where geometric fidelity and material quality are critical. +- **Product visualization**: Cleaner shape definition in studio-style renders. +- **3D printing preparation**: Better structural clarity. ## Available Workflows @@ -37,7 +37,7 @@ Generate a high-detail 3D model from a text prompt using Tripo 3.1. Tripo 3.1 Text-to-Model workflow preview - + Try the Text-to-Model workflow instantly on Comfy Cloud. @@ -52,7 +52,7 @@ Generate a high-detail 3D model from an image input using Tripo 3.1. Tripo 3.1 Image-to-Model workflow preview - + Try the Image-to-Model workflow instantly on Comfy Cloud. @@ -67,7 +67,7 @@ Generate a high-detail 3D model from multiple view images using Tripo 3.1. Tripo 3.1 Multiview-to-Model workflow preview - + Try the Multiview-to-Model workflow instantly on Comfy Cloud. diff --git a/tutorials/partner-nodes/wan/wan2-7.mdx b/tutorials/partner-nodes/wan/wan2-7.mdx index 762a63724..5413f3d05 100644 --- a/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/tutorials/partner-nodes/wan/wan2-7.mdx @@ -39,7 +39,7 @@ Generate video from image inputs. Supports first-frame, first+last-frame, and au Get the Wan2.7 Image-to-Video workflow file. - + Try the Image-to-Video workflow instantly on Comfy Cloud. @@ -54,7 +54,7 @@ Generate video from pure text prompts. Optionally include audio input and multi- Get the Wan2.7 Text-to-Video workflow file. - + Try the Text-to-Video workflow instantly on Comfy Cloud. @@ -69,7 +69,7 @@ Use reference images of a subject's visual appearance along with an optional voc Get the Wan2.7 Reference-to-Video workflow file. - + Try the Reference-to-Video workflow instantly on Comfy Cloud. @@ -84,7 +84,7 @@ Edit or replicate existing videos using text prompts, a reference image, or styl Get the Wan2.7 Video Edit workflow file. - + Try the Video Edit workflow instantly on Comfy Cloud. diff --git a/tutorials/utility/moge.mdx b/tutorials/utility/moge.mdx index 73d3b6598..10b42eb8a 100644 --- a/tutorials/utility/moge.mdx +++ b/tutorials/utility/moge.mdx @@ -8,11 +8,11 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' # ComfyUI MoGe Introduction -[MoGe](https://github.com/microsoft/MoGe) (CVPR 2025, from Microsoft Research) is a powerful model for recovering 3D geometry from monocular open-domain images. It estimates metric point maps, metric depth maps, normal maps, and camera FOV — all in a single forward pass. +[MoGe](https://github.com/microsoft/MoGe) (CVPR 2025, from Microsoft Research) is a powerful model for recovering 3D geometry from monocular open-domain images. It estimates metric point maps, metric depth maps, normal maps, and camera FOV: all in a single forward pass. Key capabilities: -- **Accurate 3D geometry estimation**: point maps, depth maps, and normal maps from a single image — one model, one forward pass +- **Accurate 3D geometry estimation**: point maps, depth maps, and normal maps from a single image: one model, one forward pass - **Metric scale** (MoGe-2): point maps and depth maps in real-world scale - **Flexible resolution support**: works with various resolutions and aspect ratios (2:1 to 1:2) - **Fast inference**: ~60ms per image on A100 / RTX 3090 (FP16, ViT-L) @@ -39,8 +39,8 @@ Generated `.glb` format models will be output to the `ComfyUI/output/mesh` folde Download the MoGe checkpoint(s) and save them to the corresponding ComfyUI folder: -- **MoGe-2 (recommended)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_2_vitl_normal_fp16.safetensors) -- **MoGe-1 (baseline)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/resolve/main/moge_1_vitl_fp16.safetensors) +- **MoGe-2 (recommended)**: [moge_2_vitl_normal_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_2_vitl_normal_fp16.safetensors) +- **MoGe-1 (baseline)**: [moge_1_vitl_fp16.safetensors](https://huggingface.co/Comfy-Org/MoGe/blob/main/moge_1_vitl_fp16.safetensors) ``` ComfyUI/ @@ -56,10 +56,11 @@ ComfyUI/ ## 1. Depth Estimation -**What it does:** Takes a single image and estimates its metric depth map, colored depth preview, and mask — outputs the same metric-scale depth that MoGe infers in one forward pass. Useful as a scene depth reference for compositing, depth-based effects, or as preprocessing before mesh generation. +**What it does:** Takes a single image and estimates its metric depth map, colored depth preview, and mask: outputs the same metric-scale depth that MoGe infers in one forward pass. Useful as a scene depth reference for compositing, depth-based effects, or as preprocessing before mesh generation. MoGe also estimates the camera's field of view (FOV) from the image, which can be optionally overridden with a ground-truth value for even more accurate results. + Download JSON or search "MoGe Depth Estimation" in Template Library @@ -67,11 +68,11 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b Open in Comfy Cloud + Get the example input image for this workflow -
depth estimation color preview depth estimation raw preview @@ -88,8 +89,9 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b ## 2. Perspective to Mesh -**What it does:** Converts a single perspective photo into a textured GLB mesh with normal and depth previews. MoGe estimates point maps, depth, and normals from the visible scene, then converts them to a mesh. This is **monocular geometry estimation** — occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. +**What it does:** Converts a single perspective photo into a textured GLB mesh with normal and depth previews. MoGe estimates point maps, depth, and normals from the visible scene, then converts them to a mesh. This is **monocular geometry estimation**: occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. + Download JSON or search "3D MoGe Perspective to Mesh" in Template Library @@ -97,11 +99,11 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b Open in Comfy Cloud + Get the example input image for this workflow - ![perspective to mesh preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_perspective_to_mesh-1.webp) ### 2.1 Steps to Run @@ -115,8 +117,9 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b ## 3. Panorama to Mesh -**What it does:** Converts an equirectangular (360°) panorama into a textured GLB mesh. The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, runs monocular geometry estimation on each view independently, then merges them into a single mesh. Each segment is still single-view estimation, so the result is a rough scene reconstruction — useful for getting a spatial overview of a 360° scene, but occluded areas and geometry behind surfaces will be missing or fragmented. +**What it does:** Converts an equirectangular (360°) panorama into a textured GLB mesh. The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, runs monocular geometry estimation on each view independently, then merges them into a single mesh. Each segment is still single-view estimation, so the result is a rough scene reconstruction: useful for getting a spatial overview of a 360° scene, but occluded areas and geometry behind surfaces will be missing or fragmented. + Download JSON or search "3D MoGe Panorama to Mesh" in Template Library @@ -124,11 +127,11 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b Open in Comfy Cloud + Get the example input image for this workflow - ![panorama to mesh preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_panorama_to_mesh-1.webp) ### 3.1 Steps to Run @@ -141,5 +144,5 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b ## Community Resources -- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe) — Research paper and code -- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe) — Official ComfyUI model weights +- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe): Research paper and code +- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe): Official ComfyUI model weights diff --git a/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 44733df9f..a4e8245d3 100644 --- a/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -70,24 +70,21 @@ If the model download was not successful, you can try to download them manually **Diffusion model** - + cosmos_predict2_2B_video2world_480p_16fps.safetensors - For other weights, please visit [Cosmos_Predict2_repackaged](https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged) to download. **Text encoder** - + oldt5_xxl_fp8_e4m3fn_scaled.safetensors - **VAE** - + wan_2.1_vae.safetensors - File Storage Location ``` 📂 ComfyUI/ diff --git a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 5aa507f36..c4b3dbe52 100644 --- a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -23,13 +23,13 @@ import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; The following models are used in both Text-to-Video and Image-to-Video workflows. Download and save them to the specified directories: - + qwen_2.5_vl_7b_fp8_scaled.safetensors — save to ComfyUI/models/text_encoders/ - + byt5_small_glyphxl_fp16.safetensors — save to ComfyUI/models/text_encoders/ - + hunyuanvideo15_vae_fp16.safetensors — save to ComfyUI/models/vae/ @@ -64,7 +64,7 @@ HunyuanVideo 1.5 Text-to-Video generates 5-10 second videos from natural languag ### Model downloads - + hunyuanvideo1.5_720p_t2v_fp16.safetensors — save to ComfyUI/models/diffusion_models/ @@ -122,10 +122,10 @@ HunyuanVideo 1.5 Image-to-Video transforms static images into smooth, high-quali ### Model downloads - + sigclip_vision_patch14_384.safetensors — save to ComfyUI/models/clip_vision/ - + hunyuanvideo1.5_720p_i2v_fp16.safetensors — save to ComfyUI/models/diffusion_models/ @@ -160,10 +160,9 @@ ComfyUI/ Both workflows include a super-resolution node that can upscale the output video from 720p to 1080p. This optional upscaler uses a distilled model for efficient high-resolution output. - + hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors — save to ComfyUI/models/diffusion_models/ - All shared and workflow-specific models combined: ``` diff --git a/tutorials/video/hunyuan/hunyuan-video.mdx b/tutorials/video/hunyuan/hunyuan-video.mdx index 51d4e77f8..6dc8b76e7 100644 --- a/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/tutorials/video/hunyuan/hunyuan-video.mdx @@ -38,18 +38,17 @@ Alternatively, this guide provides direct model links if automatic downloads fai The following models are used in both Text-to-Video and Image-to-Video workflows. Download and save them to the specified directories: - + Save to ComfyUI/models/text_encoders - + Save to ComfyUI/models/text_encoders - + Save to ComfyUI/models/vae - Storage location: ``` @@ -83,10 +82,9 @@ Download the workflow image below and drag it into ComfyUI to load the workflow: ### 2. Manual models installation - + Save to ComfyUI/models/diffusion_models - Ensure you have all these model files in the correct locations: ``` @@ -137,10 +135,9 @@ Currently, the Hunyuan Image-to-Video model has two versions: ### Shared Model for v1 and v2 Versions - + Save to ComfyUI/models/clip_vision - ### V1 "concat" Image-to-Video Workflow #### 1. Workflow and Asset @@ -163,13 +160,11 @@ Download the image below and use it as the starting frame for the image-to-video Save and use as input image for I2V generation - #### 2. Related models manual installation - + Save to ComfyUI/models/diffusion_models - Ensure you have all these model files in the correct locations: ``` @@ -223,13 +218,11 @@ Download the image below and use it as the starting frame for the image-to-video Save and use as input image for I2V generation - #### 2. Related models manual installation - + Save to ComfyUI/models/diffusion_models - Ensure you have all these model files in the correct locations: ``` diff --git a/tutorials/video/kandinsky/kandinsky-5.mdx b/tutorials/video/kandinsky/kandinsky-5.mdx index 710d04f8b..14032111c 100644 --- a/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/tutorials/video/kandinsky/kandinsky-5.mdx @@ -41,13 +41,13 @@ All models are available in 5-second and 10-second video generation versions. Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 T2V" to load the workflow. -![Kandinsky 5.0 T2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_t2v-1.webp) +![Kandinsky 5.0 T2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_t2v-1.webp) - + Download the T2V workflow to use locally - + Open in Comfy Cloud @@ -57,26 +57,24 @@ Please update your ComfyUI to the latest version, and through the menu `Workflow **Text Encoders** - + Qwen2.5-VL 7B text encoder (FP8). Place in ComfyUI/models/text_encoders/ - + CLIP-L text encoder. Place in ComfyUI/models/text_encoders/ **Diffusion Model** - + Kandinsky 5.0 T2V Lite SFT diffusion model (5s). Place in ComfyUI/models/diffusion_models/ - **VAE** - + HunyuanVideo 3D VAE. Place in ComfyUI/models/vae/ - ``` ComfyUI/ ├── 📂 models/ @@ -95,13 +93,13 @@ ComfyUI/ Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 I2V" to load the workflow. -![Kandinsky 5.0 I2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_i2v-1.webp) +![Kandinsky 5.0 I2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_i2v-1.webp) - + Download the I2V workflow to use locally - + Open in Comfy Cloud @@ -111,32 +109,29 @@ Please update your ComfyUI to the latest version, and through the menu `Workflow Default input image for the I2V workflow. Download and use this image, or replace with your own. - ### 2. Manually download models **Text Encoders** - + Qwen2.5-VL 7B text encoder (FP8). Place in ComfyUI/models/text_encoders/ - + CLIP-L text encoder. Place in ComfyUI/models/text_encoders/ **Diffusion Model** - + Kandinsky 5.0 I2V Lite diffusion model (5s). Place in ComfyUI/models/diffusion_models/ - **VAE** - + HunyuanVideo 3D VAE. Place in ComfyUI/models/vae/ - ``` ComfyUI/ ├── 📂 models/ diff --git a/tutorials/video/ltxv.mdx b/tutorials/video/ltxv.mdx index 7ead602e7..850c4ff0b 100644 --- a/tutorials/video/ltxv.mdx +++ b/tutorials/video/ltxv.mdx @@ -25,7 +25,7 @@ Drag the video directly into ComfyUI to run the workflow. Allows you to control the video with a first frame image. - + Open in Comfy Cloud @@ -50,7 +50,7 @@ Drag the video directly into ComfyUI to run the workflow. ## Text to Video - + Open in Comfy Cloud @@ -68,14 +68,14 @@ Drag the video directly into ComfyUI to run the workflow. Download the following models and place them in the locations specified below: + Download and place in ComfyUI/models/checkpoints/ - Download and place in ComfyUI/models/text_encoders/ - + ``` ├── checkpoints/ │ └── ltx-video-2b-v0.9.5.safetensors diff --git a/tutorials/video/wan/fun-camera.mdx b/tutorials/video/wan/fun-camera.mdx index 3ff0af386..bfdb1fbf6 100644 --- a/tutorials/video/wan/fun-camera.mdx +++ b/tutorials/video/wan/fun-camera.mdx @@ -31,10 +31,10 @@ All of the following models can be found at [Wan_2.1_ComfyUI_repackaged](https:/ Choose either 1.3B or 14B: - + Wan2.1 Fun Camera 1.3B diffusion model - + Wan2.1 Fun Camera 14B diffusion model @@ -46,10 +46,10 @@ If you've used Wan2.1 related models before, you should already have the followi Choose one of the following: - + Full precision text encoder - + FP8 quantized text encoder (recommended for lower VRAM) @@ -57,7 +57,7 @@ Choose one of the following: ### VAE - + Wan2.1 VAE model @@ -65,7 +65,7 @@ Choose one of the following: ### CLIP Vision - + CLIP vision encoder diff --git a/tutorials/video/wan/fun-control.mdx b/tutorials/video/wan/fun-control.mdx index e2966606c..265066de6 100644 --- a/tutorials/video/wan/fun-control.mdx +++ b/tutorials/video/wan/fun-control.mdx @@ -55,7 +55,7 @@ Click the corresponding links to download. If you have used Wan-related workflow **Diffusion models** -- choose 1.3B or 14B. The 14B version has a larger file size (32GB) and higher VRAM requirements: - + Wan2.1 Fun Control 1.3B diffusion model -- lightweight, lower VRAM requirements @@ -66,26 +66,24 @@ Click the corresponding links to download. If you have used Wan-related workflow **Text encoders** -- choose one of the following models (fp16 precision has a larger size and higher performance requirements): - + UMT5 XXL text encoder -- fp16 precision, larger size, higher performance - + UMT5 XXL text encoder -- fp8 precision, smaller size, lower performance requirements **VAE** - + Wan2.1 VAE model - **CLIP Vision** - + CLIP Vision model for processing reference images - File storage location: ``` 📂 ComfyUI/ diff --git a/tutorials/video/wan/fun-inp.mdx b/tutorials/video/wan/fun-inp.mdx index 3f09d0d5e..8d88335a0 100644 --- a/tutorials/video/wan/fun-inp.mdx +++ b/tutorials/video/wan/fun-inp.mdx @@ -89,13 +89,11 @@ All models involved in this guide can be found at [Wan_2.1_ComfyUI_repackaged](h `models/vae/wan_2.1_vae.safetensors` - **CLIP Vision** `models/clip_vision/clip_vision_h.safetensors` - File storage location: ``` 📂 ComfyUI/ diff --git a/tutorials/video/wan/wan-alpha.mdx b/tutorials/video/wan/wan-alpha.mdx index 36a09915b..50317e3cf 100644 --- a/tutorials/video/wan/wan-alpha.mdx +++ b/tutorials/video/wan/wan-alpha.mdx @@ -88,7 +88,7 @@ All models mentioned can be found at [Wan_2.1_ComfyUI_repackaged](https://huggin Alpha generation LoRA. Place in ComfyUI/models/loras/ - + Lightning LoRA for fast inference. Place in ComfyUI/models/loras/ diff --git a/tutorials/video/wan/wan-ati.mdx b/tutorials/video/wan/wan-ati.mdx index 6ff3e114d..4c20d1f2d 100644 --- a/tutorials/video/wan/wan-ati.mdx +++ b/tutorials/video/wan/wan-ati.mdx @@ -91,13 +91,11 @@ All models involved in this guide can be found [here](https://huggingface.co/Com Wan2.1 VAE model. Place in ComfyUI/models/vae/ - **CLIP Vision** CLIP Vision model for processing reference images. Place in ComfyUI/models/clip_vision/ - **File Storage Location** ``` diff --git a/tutorials/video/wan/wan-causal-forcing.mdx b/tutorials/video/wan/wan-causal-forcing.mdx index 94b9dac4d..7b967fdd3 100644 --- a/tutorials/video/wan/wan-causal-forcing.mdx +++ b/tutorials/video/wan/wan-causal-forcing.mdx @@ -45,7 +45,6 @@ This recurrent approach creates **strong temporal consistency**: each frame natu This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. - ### Causal Forcing vs Causal Forcing++ | Mode | Description | diff --git a/tutorials/video/wan/wan-dancer.mdx b/tutorials/video/wan/wan-dancer.mdx index cc40e2d11..6b2f45c7f 100644 --- a/tutorials/video/wan/wan-dancer.mdx +++ b/tutorials/video/wan/wan-dancer.mdx @@ -55,20 +55,20 @@ Update your ComfyUI to the latest version, then download and drag the workflow f ### 3. Manually Download Models **Diffusion Models** -- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) -- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/resolve/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_global_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_global_fp8_scaled.safetensors) +- [wan2.2_dancer_14b_local_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan-Dancer/blob/main/diffusion_models/wan2.2_dancer_14b_local_fp8_scaled.safetensors) **LoRA** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) **Text Encoder** -- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) +- [umt5_xxl_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors) **CLIP Vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **VAE** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) ``` ComfyUI/ diff --git a/tutorials/video/wan/wan-flf.mdx b/tutorials/video/wan/wan-flf.mdx index eaea39267..d91fce0a1 100644 --- a/tutorials/video/wan/wan-flf.mdx +++ b/tutorials/video/wan/wan-flf.mdx @@ -96,13 +96,11 @@ If you have previously tried Wan Video related workflows, you may already have t wan_2.1_vae.safetensors : Wan2.1 VAE for encoding/decoding. Place in ComfyUI/models/vae/ - **CLIP Vision** clip_vision_h.safetensors : CLIP Vision encoder. Place in ComfyUI/models/clip_vision/ - File Storage Location ``` ComfyUI/ diff --git a/tutorials/video/wan/wan-video.mdx b/tutorials/video/wan/wan-video.mdx index 03ad0037a..c95f1c4ed 100644 --- a/tutorials/video/wan/wan-video.mdx +++ b/tutorials/video/wan/wan-video.mdx @@ -48,13 +48,11 @@ Choose one version from **Text encoders** to download: Wan2.1 VAE model. Place in ComfyUI/models/vae/ - **CLIP Vision** CLIP Vision model for image conditioning. Place in ComfyUI/models/clip_vision/ - File storage locations: ``` ComfyUI/ @@ -91,7 +89,6 @@ For diffusion models, we'll use the fp16 precision models in this guide because Diffusion model for Wan2.1 Text-to-Video. Place in ComfyUI/models/diffusion_models/ - > If you need other t2v precision versions, please visit [here](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models) to download them. ### Steps to Run @@ -125,13 +122,11 @@ For diffusion models, we'll use the fp16 precision models in this guide because Download the default input image, or use your own image. - #### Model Downloads Diffusion model for Wan2.1 I2V 480P. Place in ComfyUI/models/diffusion_models/ - #### Steps to Run ![ComfyUI Wan2.1 Workflow Steps](/images/tutorial/video/wan/wan2.1_i2v_14b_480p_flow_diagram.jpg) @@ -161,13 +156,11 @@ For diffusion models, we'll use the fp16 precision models in this guide because Download the default input image, or use your own image. - #### Model Downloads Diffusion model for Wan2.1 I2V 720P. Place in ComfyUI/models/diffusion_models/ - #### Steps to Run ![ComfyUI Wan2.1 Workflow Steps](/images/tutorial/video/wan/wan2.1_i2v_14b_720p_flow_diagram.jpg) diff --git a/tutorials/video/wan/wan2-2-animate.mdx b/tutorials/video/wan/wan2-2-animate.mdx index 7ac1a29d1..56ef45321 100644 --- a/tutorials/video/wan/wan2-2-animate.mdx +++ b/tutorials/video/wan/wan2-2-animate.mdx @@ -80,25 +80,21 @@ All models involved in this guide can be found [here](https://huggingface.co/Com clip_vision_h.safetensors: CLIP Vision encoder - **LoRAs** lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors: 4-step acceleration LoRA - **VAE** wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding - **Text Encoders** umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder - ``` ComfyUI/ ├───📂 models/ diff --git a/tutorials/video/wan/wan2-2-fun-camera.mdx b/tutorials/video/wan/wan2-2-fun-camera.mdx index 4204c50b1..89fb79fbd 100644 --- a/tutorials/video/wan/wan2-2-fun-camera.mdx +++ b/tutorials/video/wan/wan2-2-fun-camera.mdx @@ -96,13 +96,11 @@ The following models can be found in [Wan_2.2_ComfyUI_Repackaged](https://huggin Wan2.1 VAE for encoding/decoding - **Text Encoder** FP8 scaled text encoder - File save location ``` diff --git a/tutorials/video/wan/wan2-2-fun-control.mdx b/tutorials/video/wan/wan2-2-fun-control.mdx index 4d5d31b04..326c5d373 100644 --- a/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/tutorials/video/wan/wan2-2-fun-control.mdx @@ -106,13 +106,11 @@ All models involved in this guide can be found [here](https://huggingface.co/Com wan_2.1_vae.safetensors: Wan2.1 VAE for encoding/decoding - **Text Encoder** umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder - ``` ComfyUI/ ├───📂 models/ diff --git a/tutorials/video/wan/wan2-2-fun-inp.mdx b/tutorials/video/wan/wan2-2-fun-inp.mdx index 2c6fb2aa7..12bdfd33e 100644 --- a/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -109,13 +109,11 @@ All models involved in this guide can be found [here](https://huggingface.co/Com wan_2.1_vae.safetensors: Wan2.1 VAE for encoding and decoding - **Text Encoder** umt5_xxl_fp8_e4m3fn_scaled.safetensors: Scaled FP8 text encoder - ``` ComfyUI/ ├───📂 models/ diff --git a/tutorials/video/wan/wan2-2-s2v.mdx b/tutorials/video/wan/wan2-2-s2v.mdx index 7363fa519..ace2531d4 100644 --- a/tutorials/video/wan/wan2-2-s2v.mdx +++ b/tutorials/video/wan/wan2-2-s2v.mdx @@ -73,20 +73,16 @@ You can find the models in [our repo](https://huggingface.co/Comfy-Org/Wan_2.2_C Audio encoder model. Place in ComfyUI/models/audio_encoders/ - **vae** Wan2.1 VAE model. Place in ComfyUI/models/vae/ - **text_encoders** FP8 scaled text encoder. Place in ComfyUI/models/text_encoders/ - - ``` ComfyUI/ ├───📂 models/ diff --git a/tutorials/video/wan/wan2_2.mdx b/tutorials/video/wan/wan2_2.mdx index 702ac5bbf..d21e119a6 100644 --- a/tutorials/video/wan/wan2_2.mdx +++ b/tutorials/video/wan/wan2_2.mdx @@ -93,7 +93,7 @@ Please update your ComfyUI to the latest version, and through the menu `Workflow Download JSON or search "Wan2.2 5B" in Template Library - + Open in Comfy Cloud @@ -166,7 +166,7 @@ Or update your ComfyUI to the latest version, then download the following video Download JSON or search "Wan2.2 14B T2V" in Template Library - + Open in Comfy Cloud @@ -242,7 +242,7 @@ Or update your ComfyUI to the latest version, then download the following video Download JSON or search "Wan2.2 14B I2V" in Template Library - + Open in Comfy Cloud @@ -321,7 +321,7 @@ Download the video or the JSON workflow below and open it in ComfyUI. Download JSON or search "Wan2.2 14B FLF2V" in Template Library - + Open in Comfy Cloud diff --git a/tutorials/video/zai/scail2.mdx b/tutorials/video/zai/scail2.mdx index a56b86d06..e57801872 100644 --- a/tutorials/video/zai/scail2.mdx +++ b/tutorials/video/zai/scail2.mdx @@ -56,7 +56,6 @@ This workflow uses two subgraph nodes: a **Base** subgraph (first segment) and a This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. - ### Long Videos For longer videos, calculate the number of segments: `ceil(total_frames / 76)`. Each segment except the first uses the Extend subgraph. Duplicate the Extend node for more segments, chain the `previous_frames` output, and increment `segment_index`. @@ -108,23 +107,23 @@ Update ComfyUI to the latest version first for the built-in WanSCAILToVideo and ### Required Models **diffusion_models** -- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) +- [wan2.1_14B_SCAIL_2_fp16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/diffusion_models/wan2.1_14B_SCAIL_2_fp16.safetensors) **text_encoders** (choose one) -- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) +- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) **clip_vision** -- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors) +- [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors) **vae** -- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors) +- [Wan2_1_VAE_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors) **loras** -- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) -- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/resolve/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) +- [lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors) +- [wan2.1_SCAIL_2_DPO_lora_bf16.safetensors](https://huggingface.co/Comfy-Org/SCAIL-2/blob/main/loras/wan2.1_SCAIL_2_DPO_lora_bf16.safetensors) **checkpoints** -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) ### File Storage Locations diff --git a/zh/tutorials/3d/hunyuan3D-2.mdx b/zh/tutorials/3d/hunyuan3D-2.mdx index f4222be41..54416450d 100644 --- a/zh/tutorials/3d/hunyuan3D-2.mdx +++ b/zh/tutorials/3d/hunyuan3D-2.mdx @@ -60,7 +60,7 @@ Hunyuan3D-2mv 工作流中，我们将使用多视角的图片来生成3D模型 - + 在 Comfy Cloud 上立即运行此工作流 @@ -113,7 +113,7 @@ ComfyUI/ Hunyuan3D-2mv-turbo 工作流中，我们将使用 Hunyuan3D-2mv-turbo 模型来生成3D模型，这个模型是 Hunyuan3D-2mv 的分步蒸馏（Step Distillation）版本，可以更快地生成3D模型，在这个版本的工作流中我们设置 `cfg` 为 1.0 并添加 `flux guidance` 节点来控制 `distilled cfg` 的生成。 - + 在 Comfy Cloud 上立即运行此工作流 @@ -158,7 +158,7 @@ ComfyUI/ Hunyuan3D-2 工作流中，我们将使用 Hunyuan3D-2 模型来生成3D模型，这个模型不是一个多视角的模型，在这个工作流中，我们使用`Hunyuan3Dv2Conditioning` 节点替换掉 `Hunyuan3Dv2ConditioningMultiView` 节点。 - + 在 Comfy Cloud 上立即运行此工作流 diff --git a/zh/tutorials/3d/triposplat.mdx b/zh/tutorials/3d/triposplat.mdx index 7aa8dae01..d456036c7 100644 --- a/zh/tutorials/3d/triposplat.mdx +++ b/zh/tutorials/3d/triposplat.mdx @@ -2,7 +2,7 @@ title: "TripoSplat 图片转高斯泼溅 ComfyUI 工作流示例" description: "使用 TripoSplat 从单张 2D 图片生成高质量 3D 高斯泼溅表示，支持可控密度和渲染预算。" sidebarTitle: "TripoSplat" -translationSourceHash: 95c7decf +translationSourceHash: 0cba7aeb translationFrom: tutorials/3d/triposplat.mdx translationBlockHashes: "_intro": 77f8628a @@ -10,11 +10,13 @@ translationBlockHashes: "Workflow node guide": 83cd15a0 "Steps to run": e4ad7fa4 "Output options": 9a1cc408 - "Model downloads": c6ba2321 + "Model downloads": 44894623 --- + + import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **TripoSplat** 是一款开源模型，能够从单张 2D 图片直接生成 **3D 高斯泼溅（Gaussian splat）** 表示。由 VAST-AI 开发，以开源许可证发布。 @@ -25,9 +27,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - + + + 在 Comfy Cloud 上立即运行此工作流 + + 下载 JSON 或在模板库中搜索 "TripoSplat" + ## 工作原理 @@ -41,7 +48,6 @@ TripoSplat 使用 **前馈架构**，接收单张 RGB 图像并直接预测一 本工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 - ## 工作流节点指南 ### LoadImage @@ -107,26 +113,23 @@ TripoSplat 使用 **前馈架构**，接收单张 RGB 图像并直接预测一 下载 TripoSplat 模型及所需文件。放入对应的 `models/` 子目录。 + triposplat_fp16.safetensors — TripoSplat 扩散模型检查点 - triposplat_vae_decoder_fp16.safetensors — VAE 解码器 - flux2-vae.safetensors — Flux.2 VAE，用于潜空间编码 - dino_v3_vit_h.safetensors — CLIP 视觉编码器（DINOv2） - birefnet.safetensors — 用于预处理的背景去除模型 - + ### 模型存放位置 ``` diff --git a/zh/tutorials/flux/flux-1-controlnet.mdx b/zh/tutorials/flux/flux-1-controlnet.mdx index c550fac57..823ba44bf 100644 --- a/zh/tutorials/flux/flux-1-controlnet.mdx +++ b/zh/tutorials/flux/flux-1-controlnet.mdx @@ -49,7 +49,7 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 Download JSON or search "Flux.1 Canny" in Template Library - + 在 Comfy Cloud 中打开 @@ -120,7 +120,7 @@ ComfyUI/ Download JSON or search "Flux.1 Depth LoRA" in Template Library - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/flux/flux-1-fill-dev.mdx b/zh/tutorials/flux/flux-1-fill-dev.mdx index 898f4b946..3fbaf598f 100644 --- a/zh/tutorials/flux/flux-1-fill-dev.mdx +++ b/zh/tutorials/flux/flux-1-fill-dev.mdx @@ -66,7 +66,7 @@ ComfyUI/ Download JSON or search "flux_fill_inpaint" in Template Library - + 在 Comfy Cloud 中打开 @@ -76,7 +76,7 @@ ComfyUI/ Download JSON or search "flux_fill_outpaint" in Template Library - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/flux/flux-1-kontext-dev.mdx b/zh/tutorials/flux/flux-1-kontext-dev.mdx index 35c043a56..73e65d4e2 100644 --- a/zh/tutorials/flux/flux-1-kontext-dev.mdx +++ b/zh/tutorials/flux/flux-1-kontext-dev.mdx @@ -87,7 +87,7 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 Download JSON or search "Flux Kontext Dev" in Template Library - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/flux/flux-1-text-to-image.mdx b/zh/tutorials/flux/flux-1-text-to-image.mdx index 573684b24..d465eddbc 100644 --- a/zh/tutorials/flux/flux-1-text-to-image.mdx +++ b/zh/tutorials/flux/flux-1-text-to-image.mdx @@ -55,7 +55,7 @@ Flux 以其卓越的画面质量和灵活性而闻名，能够生成高质量、 ![Flux Dev 原始版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) - + 在 Comfy Cloud 上运行此工作流 @@ -116,7 +116,7 @@ ComfyUI/ - + 在 Comfy Cloud 上运行此工作流 @@ -126,7 +126,7 @@ ComfyUI/ Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. - + 在 Comfy Cloud 上运行此工作流 diff --git a/zh/tutorials/flux/flux-1-uso.mdx b/zh/tutorials/flux/flux-1-uso.mdx index 16f50ae21..68d40307a 100644 --- a/zh/tutorials/flux/flux-1-uso.mdx +++ b/zh/tutorials/flux/flux-1-uso.mdx @@ -35,7 +35,7 @@ USO 支持三种主要方法：

下载 JSON 工作流

@@ -45,7 +45,7 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', Download the workflow JSON and drag it into ComfyUI - + 在 Comfy Cloud 上运行此工作流 diff --git a/zh/tutorials/flux/flux-2-dev.mdx b/zh/tutorials/flux/flux-2-dev.mdx index 6d500db06..ae5eb5839 100644 --- a/zh/tutorials/flux/flux-2-dev.mdx +++ b/zh/tutorials/flux/flux-2-dev.mdx @@ -41,25 +41,27 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 使用 FLUX.2 Dev 进行文生图的基础工作流。 - + + 直接在 Comfy Cloud 中打开此工作流 - - + 下载 JSON 工作流文件用于本地使用 + ## 多图参考工作流 2 图参考工作流示例，可参考此实现扩展支持更多参考图像。 - + + 直接在 Comfy Cloud 中打开此工作流 - - + 下载 JSON 工作流文件用于本地使用 + ## 模型链接 diff --git a/zh/tutorials/flux/flux-2-klein.mdx b/zh/tutorials/flux/flux-2-klein.mdx index b4de58411..b7f715c38 100644 --- a/zh/tutorials/flux/flux-2-klein.mdx +++ b/zh/tutorials/flux/flux-2-klein.mdx @@ -32,36 +32,33 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 ## Flux.2 Klein 4B 工作流 - + + 下载 Flux.2 Klein 4B 文生图工作流。 - - + 下载 4B Base 模型图像编辑工作流。 - - + 下载 4B 蒸馏版快速图像编辑工作流。 - + ## Flux.2 Klein 4B 模型下载 + 4B 模型文本编码器。 - 扩散模型（4B Base）。 - 扩散模型（4B 蒸馏版）。 - 4B 模型 VAE。 - + **4B 模型存放路径** ``` @@ -78,40 +75,37 @@ FLUX.2 [Klein] 是 Flux 系列中目前(2026年1月15日)最快的模型，将 ## Flux.2 Klein 9B 工作流 - + + 下载 Flux.2 Klein 9B 文生图工作流。 - - + 下载 9B Base 模型图像编辑工作流。 - - + 下载 9B 蒸馏版快速图像编辑工作流。 - + ## Flux.2 Klein 9B 模型下载 扩散模型需要访问 BFL 仓库，接受协议后才能下载。 + 扩散模型（9B Base）。 - 扩散模型（9B 蒸馏版）。 - 9B 模型文本编码器。 - 9B 模型 VAE。 - + **9B 模型存放路径** ``` diff --git a/zh/tutorials/flux/flux1-krea-dev.mdx b/zh/tutorials/flux/flux1-krea-dev.mdx index d85fe90ee..22962640f 100644 --- a/zh/tutorials/flux/flux1-krea-dev.mdx +++ b/zh/tutorials/flux/flux1-krea-dev.mdx @@ -32,7 +32,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ![Flux Krea Dev 工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) - + 在 Comfy Cloud 上运行此工作流 diff --git a/zh/tutorials/image/hidream/hidream-i1.mdx b/zh/tutorials/image/hidream/hidream-i1.mdx index 7e1487336..605c94338 100644 --- a/zh/tutorials/image/hidream/hidream-i1.mdx +++ b/zh/tutorials/image/hidream/hidream-i1.mdx @@ -99,7 +99,7 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 full 版本工作流 - + Run this workflow on Comfy Cloud with zero setup @@ -141,7 +141,7 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 ### HiDream-I1 dev 版本工作流 - + Run this workflow on Comfy Cloud with zero setup @@ -183,7 +183,7 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 - + Run this workflow on Comfy Cloud with zero setup diff --git a/zh/tutorials/image/omnigen/omnigen2.mdx b/zh/tutorials/image/omnigen/omnigen2.mdx index f80b90415..fd70c11b8 100644 --- a/zh/tutorials/image/omnigen/omnigen2.mdx +++ b/zh/tutorials/image/omnigen/omnigen2.mdx @@ -69,7 +69,7 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 ### 1. 工作流文件下载 - + Open and run this workflow directly in Comfy Cloud. @@ -99,7 +99,7 @@ OmniGen2 有丰富的图像编辑能力，并且支持为图像添加文本 ### 1. 工作流文件下载 - + Open and run this workflow directly in Comfy Cloud. diff --git a/zh/tutorials/image/pixeldit/pixeldit.mdx b/zh/tutorials/image/pixeldit/pixeldit.mdx index 18f597d59..cb3834a62 100644 --- a/zh/tutorials/image/pixeldit/pixeldit.mdx +++ b/zh/tutorials/image/pixeldit/pixeldit.mdx @@ -34,7 +34,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 下载 JSON 或在模板库中搜索 "PixelDiT" - 工作流由三个主要节点组成： @@ -46,7 +45,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 本工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 - ### 工作流控制参数 **文生图 (PixelDiT)** 子图节点暴露的控件包括： @@ -63,14 +61,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' PixelDiT 使用两个模型文件：文本编码器和扩散模型。 + gemma_2_2b_it_elm_bf16.safetensors — Gemma-2-2B-IT 文本编码器 - pixeldit_1300m_1024px_bf16.safetensors — PixelDiT 1300M 1024px 扩散模型 - + ### 模型存放位置 ``` diff --git a/zh/tutorials/image/qwen/qwen-image-2512.mdx b/zh/tutorials/image/qwen/qwen-image-2512.mdx index 1e4b05848..927c96793 100644 --- a/zh/tutorials/image/qwen/qwen-image-2512.mdx +++ b/zh/tutorials/image/qwen/qwen-image-2512.mdx @@ -44,7 +44,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - + 在 Comfy Cloud 上运行 @@ -57,7 +57,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **Text to Image (Qwen-Image 2512 4steps)**：使用 Lightning LoRA 的 4 步加速生成 - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx index e5a87c4be..11739c3be 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -35,7 +35,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/image/qwen/qwen-image-edit.mdx b/zh/tutorials/image/qwen/qwen-image-edit.mdx index 70038b415..0955814ab 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit.mdx @@ -52,7 +52,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' Download JSON or search "image_qwen_image_edit" in Template Library - + Run this workflow on Cloud GPUs with zero setup diff --git a/zh/tutorials/image/qwen/qwen-image-layered.mdx b/zh/tutorials/image/qwen/qwen-image-layered.mdx index e524ada1f..ae263a941 100644 --- a/zh/tutorials/image/qwen/qwen-image-layered.mdx +++ b/zh/tutorials/image/qwen/qwen-image-layered.mdx @@ -32,11 +32,11 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Qwen-Image-Layered 工作流 | -| +| | Download the JSON workflow file | | -| +| | Run ComfyUI online with zero setup | | diff --git a/zh/tutorials/image/qwen/qwen-image.mdx b/zh/tutorials/image/qwen/qwen-image.mdx index 4bb0dc365..9e9c600e9 100644 --- a/zh/tutorials/image/qwen/qwen-image.mdx +++ b/zh/tutorials/image/qwen/qwen-image.mdx @@ -63,10 +63,12 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' | - + + + 在本篇文档所附工作流中使用的不同模型有三种 @@ -165,7 +167,7 @@ Qwen_image_distill 这是一个 ControlNet 模型 - + @@ -224,7 +226,7 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches 工作流 - + @@ -294,7 +296,7 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http ## Qwen Image union ControlNet LoRA 工作流 - + diff --git a/zh/tutorials/image/z-image/z-image-turbo.mdx b/zh/tutorials/image/z-image/z-image-turbo.mdx index cc62fbb37..c6cd9c1d5 100644 --- a/zh/tutorials/image/z-image/z-image-turbo.mdx +++ b/zh/tutorials/image/z-image/z-image-turbo.mdx @@ -35,30 +35,29 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Z-Image-Turbo 文生图工作流 - + + 下载 Z-Image-Turbo 文生图工作流 JSON 文件。 - 在 ComfyUI Cloud 上直接运行此工作流。 - + ### Z-Image-Turbo 模型下载 + Z-Image-Turbo 文本编码器。 - Z-Image-Turbo 扩散模型。 - Z-Image-Turbo VAE。 - + **Z-Image-Turbo 模型存储位置** ``` @@ -76,16 +75,14 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 此工作流使用 Z-Image-Turbo Fun Union ControlNet 模型生成带有 ControlNet 引导的图像。它对参考图像应用 Canny 边缘检测，并使用 ControlNet 来引导生成过程。 - + 下载 Z-Image-Turbo Fun Union ControlNet 工作流 JSON 文件。 - ### ControlNet 所需的额外模型 Z-Image-Turbo ControlNet 模型补丁。 - **模型存储位置** ``` diff --git a/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 7d18b4589..ba858cab5 100644 --- a/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -28,14 +28,14 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; Beeble SwitchX 图像编辑工作流 + 打开 Comfy Cloud - 下载 JSON 或在模板库中搜索 "Beeble SwitchX: Image Edit" - + ### 工作原理 该工作流接收两个输入图像： diff --git a/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 064e698a4..45d623c7f 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -105,7 +105,7 @@ ComfyUI 提供两套预置的 Seedance 2.0 真人支持模板。两者都会将 Seedance 2.0 Real Human R2V workflow preview - + Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. @@ -126,7 +126,7 @@ Download this sample input image to try the workflow: Seedance 2.0 Real Human FLF2V workflow preview - + Try the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow instantly on Comfy Cloud. diff --git a/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index 35d08fed0..f06b467b5 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -38,38 +38,38 @@ Seedance 2.0 是字节跳动推出的新一代多模态视频生成模型，现 仅使用文本提示生成视频，由 Seedance 2.0 负责场景、运动与节奏的整体呈现。 - + + 在 Comfy Cloud 上直接体验 Text-to-Video 工作流。 - 下载工作流 JSON。 - + ### 参考生成视频（R2V） 使用参考图片、视频或音频来引导外观、运动与节奏，生成结果在人物/物体与风格上保持一致。 - + + 在 Comfy Cloud 上直接体验 Reference-to-Video 工作流。 - 下载工作流 JSON。 - + ### 首尾帧生成视频（FLF2V） 提供起始帧与结束帧，由 Seedance 2.0 生成中间的运动与过渡。 - + + 在 Comfy Cloud 上直接体验 First-Last-Frame-to-Video 工作流。 - 下载工作流 JSON。 - + ## Seedance 2.0 Mini Seedance 2.0 Mini 是 Seedance 2.0 的轻量快速版本，适用于日常视频生成。它保持了与完整模型相同的音画同步、角色一致性和摄像机控制，而单次运行成本更低。 diff --git a/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index 582c13235..eceaf0e62 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -31,14 +31,14 @@ Seedream 5.0 lite 是 BytePlus 最新的图像生成模型。它是 Seedream 系 ## Seedream 5.0 lite 图像编辑工作流 + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索 "Seedream 5.0 lite" - + ### 图像编辑展示 ![指令跟随示例 1](https://substack-post-media.s3.amazonaws.com/public/images/9a770e2a-c0ed-4253-bd60-14c5beff150d_2048x2048.png) diff --git a/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index 243f92752..a6dd9b74f 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -32,14 +32,14 @@ Seedream 5.0 Pro 是字节跳动推出的专业级图像生成模型，作为 Se 通过文本提示生成高质量图像，Seedream 5.0 Pro 负责处理构图、光照和细节。 + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索“Seedream 5.0 Pro” - + ![Seedream 5.0 Pro 文本转图像预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/Seedream5.0_Pro_T2I.png) ## Seedream 5.0 Pro 图像编辑工作流 diff --git a/zh/tutorials/partner-nodes/google/nano-banana-2.mdx b/zh/tutorials/partner-nodes/google/nano-banana-2.mdx index 90729ca2f..02696c3fd 100644 --- a/zh/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/zh/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -32,14 +32,14 @@ Nano Banana 2 现已通过合作伙伴节点在 ComfyUI 中可用。此版本从 ## Nano Banana 2 图像编辑工作流 + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索 "Nano Banana 2" - + ### Pro 级质量展示 ![质量对比](https://substack-post-media.s3.amazonaws.com/public/images/1f92ae2e-14d8-4a3b-9ed7-e57dacd2584f_1825x1696.png) diff --git a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 321d75da8..e5d9c9455 100644 --- a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -49,14 +49,14 @@ HappyHorse 1.0 聚焦于强美学、多镜头叙事和强大的编辑型工作 > + 获取 HappyHorse 1.0 图生视频工作流文件。 - - + 在 Comfy Cloud 上即刻体验图生视频工作流。 - + ## HappyHorse 1.0 文生视频 仅通过文本提示词生成电影级视频，具备多镜头叙事与精致的视觉氛围。 @@ -68,7 +68,7 @@ HappyHorse 1.0 聚焦于强美学、多镜头叙事和强大的编辑型工作 > - + 在 Comfy Cloud 上立即试用文本转视频工作流 @@ -76,7 +76,7 @@ HappyHorse 1.0 聚焦于强美学、多镜头叙事和强大的编辑型工作 ![HappyHorse 1.0 Text-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_t2v-1.webp) - + 在 Comfy Cloud 上立即试用参考转视频工作流 @@ -84,15 +84,16 @@ HappyHorse 1.0 聚焦于强美学、多镜头叙事和强大的编辑型工作 ![HappyHorse 1.0 Reference-to-Video workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_r2v-1.webp) + Get the example car reference image. - Get the example person reference image. + - + Try the Video Edit workflow instantly on Comfy Cloud. @@ -100,34 +101,35 @@ HappyHorse 1.0 聚焦于强美学、多镜头叙事和强大的编辑型工作 ![HappyHorse 1.0 Video Edit workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_happyhorse1_0_video_edit-1.webp) + Get the example input image for the video edit workflow. - Get the example input video for the video edit workflow. - + ## HappyHorse 1.0 参考生成视频 使用参考主体驱动视频生成，在电影级多镜头序列中保持身份一致性。 + 获取 HappyHorse 1.0 参考生成视频工作流文件。 - - + 在 Comfy Cloud 上即刻体验参考生成视频工作流。 - + ## HappyHorse 1.0 视频编辑 通过 V2V 和 SV2V 编辑工作流，在保留运动与构图的同时改造现有素材或替换/插入主体。 + 获取 HappyHorse 1.0 视频编辑工作流文件。 - - + 在 Comfy Cloud 上即刻体验视频编辑工作流。 + \ No newline at end of file diff --git a/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 3468b13ce..92dcdbb8c 100644 --- a/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -61,7 +61,6 @@ ComfyUI 现已原生集成了相应的 Hunyuan 3D API，让您可以方便地在 在 Comfy Cloud 上运行 - ## 图像到 3D 工作流 上传一张或多张图像以生成高质量的 3D 模型。支持 2-4 张多视图图像可提高几何和材质的保真度。 @@ -69,7 +68,6 @@ ComfyUI 现已原生集成了相应的 Hunyuan 3D API，让您可以方便地在 在 Comfy Cloud 上运行 - ## 多视图到 3D 工作流 提供多个视角的图像（正面、背面、侧面）以生成精度和细节更高的 3D 模型。此功能与图像到 3D 使用相同的工作流——只需上传 2-4 张不同角度的图像即可。 @@ -77,7 +75,6 @@ ComfyUI 现已原生集成了相应的 Hunyuan 3D API，让您可以方便地在 在 Comfy Cloud 上运行 - ## 高级功能 继 [HY 3D 3.0 初始集成](https://blog.comfy.org/p/hunyuan-3d-30-in-comfyui-state-of) 之后，Hunyuan 3D 的高级处理功能现已通过合作伙伴节点提供。这些工作流通过将关键后处理步骤引入 ComfyUI，帮助弥合生成与生产之间的差距。 @@ -86,34 +83,35 @@ ComfyUI 现已原生集成了相应的 Hunyuan 3D API，让您可以方便地在 将完整的 3D 模型拆分为有意义的结构部件，如盔甲、配件、车轮或其他独立组件。这使得编辑资产的特定区域、交换部件以创建变体、以及为模块化工作流、动画或下游组装准备模型变得更加容易。 - + + 运行工作流 - 获取 JSON 文件 - + ### UV 展开 自动为支持的 3D 模型生成 UV 布局，将原始几何体转换为更易于贴图的资产。无需手动切割接缝和组织 UV 岛，创作者可以更快地进入绑定、烘焙和材质工作，并拥有更清晰的起点。 - + + 运行工作流 - 获取 JSON 文件 - + ### 智能拓扑 将密集几何体转换为具有更有组织的边缘流的更干净的网格，帮助生成的模型更容易在实际生产流程中优化和重用。这在为游戏引擎、实时渲染或任何受益于低密度、结构更好的几何体的工作流准备资产时特别有用。 - + + 运行工作流 - 获取 JSON 文件 + \ No newline at end of file diff --git a/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index 1b795429f..de78da338 100644 --- a/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -60,7 +60,6 @@ ComfyUI 现已原生集成了相应的 Hunyuan 3D API，让您可以方便地在 在 Comfy Cloud 上运行 - ## 图像到 3D 工作流 上传一张或多张图像以生成高质量的 3D 模型。支持 2-4 张多视图图像可提高几何和材质的保真度。 @@ -68,11 +67,10 @@ ComfyUI 现已原生集成了相应的 Hunyuan 3D API，让您可以方便地在 在 Comfy Cloud 上运行 - ## 多视图到 3D 工作流 提供多个视角的图像（正面、背面、侧面）以生成精度和细节更高的 3D 模型。此功能与图像到 3D 使用相同的工作流——只需上传 2-4 张不同角度的图像即可。 在 Comfy Cloud 上运行 - + \ No newline at end of file diff --git a/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 1ead03fb9..228395c3d 100644 --- a/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -23,14 +23,14 @@ Ideogram 4.0 是 Ideogram 最新推出的文生图模型，具有出色的照片 ## Ideogram 4.0 文生图工作流 + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索"Ideogram v4: Text to Image (API)" - + ![Ideogram 4.0 生成示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_ideogram_v4_t2i.png) *Ideogram 4.0 API 生成示例* diff --git a/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx b/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx index 24feaed8d..40db73bd1 100644 --- a/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -80,7 +80,7 @@ Uni-1 在广泛的任务中表现优异： Luma Uni-1 Image Create workflow preview - + @@ -100,7 +100,7 @@ Download this sample input image to try the workflow: Luma Uni-1 Image Edit workflow preview - + @@ -120,14 +120,14 @@ Download this sample input image to try the workflow: ### Image Edit 工作流 - + + 在 Comfy Cloud 上即时体验 Image Edit 工作流。 - 下载 JSON 或在模板库中搜索 "Luma UNI-1 Image Edit" - + 工作流很简单：**提示词 → 评估 → 优化**。探索阶段保持种子为空，找到效果不错的结果后再锁定种子并迭代。 ## 核心参数 diff --git a/zh/tutorials/partner-nodes/meshy/meshy-6.mdx b/zh/tutorials/partner-nodes/meshy/meshy-6.mdx index a7c7dc87c..5bdd265b8 100644 --- a/zh/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/zh/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -36,24 +36,24 @@ Meshy 6 是 Meshy 最新一代的 3D 模型生成技术，在几何质量、纹 使用 Meshy 6 直接从文本描述生成 3D 模型。 - + Run the text-to-model workflow instantly on Comfy Cloud. ![Meshy 6 Text-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_text_to_model-1.webp) - + Run the image-to-model workflow instantly on Comfy Cloud. ![Meshy 6 Image-to-Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_image_to_model-1.webp) + 获取此工作流的示例输入图片 - - + Run the multi-view workflow instantly on Comfy Cloud. - + ![Meshy 6 Multi-view to Model workflow preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_meshy_multi_image_to_model-1.webp) @@ -71,15 +71,15 @@ Meshy 6 是 Meshy 最新一代的 3D 模型生成技术，在几何质量、纹 下载工作流 JSON 文件以供本地使用。 - ## 多视角到模型工作流 从多个视角图像生成 3D 模型，以获得更准确的几何和纹理重建。 - + + 在 Comfy Cloud 上即时运行多视角工作流。 - 下载工作流 JSON 文件以供本地使用。 + \ No newline at end of file diff --git a/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx b/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx index 68a0eb502..33a654fb1 100644 --- a/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -45,28 +45,28 @@ GPT-Image-2 在节点库中通过 **OpenAI GPT Image 1.5** 节点的 `model` 选 通过文本提示词生成图像,利用 GPT-Image-2 基于推理的构图能力。 - + + 在 Comfy Cloud 上一键体验文生图工作流。 - 下载工作流 JSON 文件。 - + ![GPT-Image-2 文生图示例](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_t2i_1.jpg) ### 图像编辑 对输入图像进行编辑,在最高 2K 分辨率下保持结构一致性。 - + + 在 Comfy Cloud 上一键体验图像编辑工作流。 - 下载工作流 JSON 文件。 - + ![GPT-Image-2 图生图示例](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_i2i_1.jpg) ![GPT-Image-2 图像编辑示例 1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/openai/gpt_image_2/gpt_image_2_image_edit_1.jpg) diff --git a/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx b/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx index 351c7319f..dfbe0da9a 100644 --- a/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -34,14 +34,14 @@ Recraft V4 是一款专为专业设计工作打造的全新图像生成模型。 ## Recraft V4 文本到图像工作流 + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索 "Recraft V4 Text to Image" - + ### 运行工作流步骤 1. (可选) 修改 `Recraft Style` 节点来控制图像的视觉风格 @@ -65,14 +65,14 @@ Recraft V4 是一款专为专业设计工作打造的全新图像生成模型。 Recraft V4 可以直接生成可用于生产的 SVG 矢量输出。这对于 Logo、图标、品牌资产或任何需要缩放的内容非常有用。SVG 输出与 Illustrator、Figma 和 Sketch 兼容。 + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索 "Recraft V4 Text to Vector" - + ### 矢量展示 ![屋顶插画](https://substack-post-media.s3.amazonaws.com/public/images/571c5214-5c7f-43f9-81c4-80cad1e5578b_1792x1024.png) diff --git a/zh/tutorials/partner-nodes/rodin/model-generation.mdx b/zh/tutorials/partner-nodes/rodin/model-generation.mdx index 7b64890af..fe41bfdcc 100644 --- a/zh/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/zh/tutorials/partner-nodes/rodin/model-generation.mdx @@ -37,7 +37,7 @@ Hyper3D Rodin (hyper3d.ai) 是一个专注于通过人工智能快速生成高 Single-view Model Generation (Json Format) - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/partner-nodes/topaz/astra-2.mdx b/zh/tutorials/partner-nodes/topaz/astra-2.mdx index 7c36435f6..e3c3469d6 100644 --- a/zh/tutorials/partner-nodes/topaz/astra-2.mdx +++ b/zh/tutorials/partner-nodes/topaz/astra-2.mdx @@ -38,14 +38,14 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; **加载视频** → **Topaz Video Enhance** → **保存视频**。打开模板后若默认不是本模型，把放大模型改成 **Astra 2**。 - + + 在 Comfy Cloud 上立即打开该工作流。 - 下载可在本地 ComfyUI 使用的 workflow JSON。 - + ## 相关链接 - [ComfyUI 视频放大](/zh/tutorials/utility/video-upscale) — 视频放大方案选择指南。 diff --git a/zh/tutorials/partner-nodes/tripo/model-generation.mdx b/zh/tutorials/partner-nodes/tripo/model-generation.mdx index be5a0b82f..6be569bf0 100644 --- a/zh/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/zh/tutorials/partner-nodes/tripo/model-generation.mdx @@ -37,7 +37,7 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - + Try the Text-to-Model workflow instantly on Comfy Cloud. @@ -65,7 +65,7 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 下载下面的文件，并拖入 ComfyUI 中加载对应工作流。 - + Try the Image-to-Model workflow instantly on Comfy Cloud. diff --git a/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx index b3d3a21d9..e864acb72 100644 --- a/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -48,7 +48,7 @@ Generate a high-detail 3D model from a text prompt using Tripo 3.1. Tripo 3.1 Text-to-Model workflow preview - + @@ -58,7 +58,7 @@ Generate a high-detail 3D model from an image input using Tripo 3.1. Tripo 3.1 Image-to-Model workflow preview - + @@ -68,7 +68,7 @@ Generate a high-detail 3D model from multiple view images using Tripo 3.1. Tripo 3.1 Multiview-to-Model workflow preview - + @@ -76,24 +76,24 @@ Generate a high-detail 3D model from multiple view images using Tripo 3.1. ### 图片转模型 - + + 在 Comfy Cloud 上立即尝试图片转模型工作流。 - 下载 JSON 或在模板库中搜索 "Tripo 3.1 Image-to-Model"。 - + ### 多视图转模型 - + + 在 Comfy Cloud 上立即尝试多视图转模型工作流。 - 下载 JSON 或在模板库中搜索 "Tripo 3.1 Multiview-to-Model"。 - + ### 版本对比 | 方面 | Tripo 3.1 | 之前版本 | diff --git a/zh/tutorials/partner-nodes/wan/wan2-7.mdx b/zh/tutorials/partner-nodes/wan/wan2-7.mdx index 755408c95..97f48298b 100644 --- a/zh/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/zh/tutorials/partner-nodes/wan/wan2-7.mdx @@ -52,7 +52,7 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 - + Wan2.7 T2V workflow preview @@ -60,7 +60,7 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 - + Wan2.7 R2V workflow preview @@ -68,7 +68,7 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 - + Wan2.7 Video Edit workflow preview @@ -76,7 +76,7 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 - + @@ -84,34 +84,35 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 从纯文本提示词生成视频。可选加入音频输入和多镜头叙事，实现更丰富的故事讲述。 + 获取 Wan2.7 文生视频工作流文件。 - - + 在 Comfy Cloud 上即刻体验文生视频工作流。 - + ## Wan2.7 参考生成视频 使用人物视觉外观的参考图像，并可选配声线参考。支持最多 5 个真人输入，实现多角色交互场景。 + 获取 Wan2.7 参考生成视频工作流文件。 - - + 在 Comfy Cloud 上即刻体验参考生成视频工作流。 - + ## Wan2.7 视频编辑 使用文本提示词、参考图像或风格迁移来编辑或复制现有视频。 + 获取 Wan2.7 视频编辑工作流文件。 - - + 在 Comfy Cloud 上即刻体验视频编辑工作流。 + \ No newline at end of file diff --git a/zh/tutorials/utility/moge.mdx b/zh/tutorials/utility/moge.mdx index f4e05e4cf..b185e0489 100644 --- a/zh/tutorials/utility/moge.mdx +++ b/zh/tutorials/utility/moge.mdx @@ -72,15 +72,14 @@ ComfyUI/ MoGe 还会自动估计图片的相机 FOV，也可以手动输入真实 FOV 以获得更精确的结果。 + 下载 JSON 或在模板库中搜索 "MoGe Depth Estimation" - 获取本工作流所用的示例输入图片 - - + ### 1.1 运行步骤 1. 确保 `LoadMoGeModel` 节点已加载 MoGe 检查点 @@ -94,15 +93,14 @@ MoGe 还会自动估计图片的相机 FOV，也可以手动输入真实 FOV 以 **功能：** 将单张透视照片转换为带纹理的 GLB 网格，同时生成法线和深度预览。MoGe 从可见场景中估计点云、深度和法线，再转换为网格。这是**单目几何估计**——遮挡区域和物体背面会缺失或出现碎片。适合场景快速原型、参考几何体，或将深度/法线可视化展示为网格，不能替代多视角 3D 重建。 + 下载 JSON 或在模板库中搜索 "3D MoGe Perspective to Mesh" - 获取本工作流所用的示例输入图片 - - + ### 2.1 运行步骤 1. 确保 `LoadMoGeModel` 节点已加载 MoGe 检查点 @@ -116,15 +114,14 @@ MoGe 还会自动估计图片的相机 FOV，也可以手动输入真实 FOV 以 **功能：** 将 360° 等距柱状全景图转换为带纹理的 GLB 网格。该工作流使用 `MoGePanoramaInference` 将全景图分割为 12 个透视视角，分别独立进行单目几何估计后合并为单个网格。每个分段仍然是单视图估计，因此结果是粗略的场景重建——适合获得 360° 场景的空间概览，但遮挡区域和表面后的几何结构会缺失或碎片化。 + 下载 JSON 或在模板库中搜索 "3D MoGe Panorama to Mesh" - 获取本工作流所用的示例输入图片 - - + ### 3.1 运行步骤 1. 确保 `LoadMoGeModel` 节点已加载 MoGe 检查点 diff --git a/zh/tutorials/video/bytedance/bernini-r.mdx b/zh/tutorials/video/bytedance/bernini-r.mdx index 6f8bdf07d..68b2f3c28 100644 --- a/zh/tutorials/video/bytedance/bernini-r.mdx +++ b/zh/tutorials/video/bytedance/bernini-r.mdx @@ -108,7 +108,6 @@ ComfyUI/ 此工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 - --- ## 2. 视频编辑 @@ -139,7 +138,6 @@ ComfyUI/ 此工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 - ## 社区资源 - [Bernini GitHub (bytedance/Bernini)](https://github.com/bytedance/Bernini) — 研究论文和任务文档 diff --git a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index aea227b02..7fa04fd6d 100644 --- a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -22,7 +22,6 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- 使用 Cosmos-Predict2 的进行文生图 - {/* ## Cosmos Predict2 Video2World 工作流 diff --git a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 1b9048a36..2daad36be 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -16,7 +16,9 @@ translationBlockHashes: import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; -混元视频 1.5是由腾讯混元团队开发的轻量级 8.3B 参数模型。它可在消费级 GPU（24GB 显存）上提供旗舰级质量的视频生成，大幅降低了使用门槛，同时不影响质量。 +[HunyuanVideo 1.5](https://github.com/Tencent/HunyuanVideo) 是由腾讯混元团队开发的轻量级 8.3B 参数模型。它可在消费级 GPU（24GB 显存）上提供旗舰级质量的视频生成，大幅降低了使用门槛，同时不影响质量。 + + ## 模型亮点 @@ -26,42 +28,165 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; - **电影级质量**：原生 720p 输出（可升级至 1080p），具有专业美学效果。 - **丰富功能**：支持多种风格（写实、动漫、3D）和视频内文字渲染（中文/英文）。 - +## 所有工作流通用模型 + +以下模型在文生视频和图生视频工作流中都会用到。请下载并保存到指定目录： -## 模型链接 + + + qwen_2.5_vl_7b_fp8_scaled.safetensors，保存到 ComfyUI/models/text_encoders/ + + + byt5_small_glyphxl_fp16.safetensors，保存到 ComfyUI/models/text_encoders/ + + + hunyuanvideo15_vae_fp16.safetensors，保存到 ComfyUI/models/vae/ + + -**text_encoders** +#### 存储位置 + +``` +ComfyUI/ +├── 📂 models/ +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors +│ │ └── byt5_small_glyphxl_fp16.safetensors +│ └── 📂 vae/ +│ └── hunyuanvideo15_vae_fp16.safetensors +``` -- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) -- [byt5_small_glyphxl_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/byt5_small_glyphxl_fp16.safetensors) +## 混元视频 1.5 文生视频工作流 -**diffusion_models** +混元视频 1.5 文生视频可根据自然语言描述生成 5-10 秒视频，质量更高且显存需求更低。 -- [hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors) -- [hunyuanvideo1.5_720p_t2v_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/diffusion_models/hunyuanvideo1.5_720p_t2v_fp16.safetensors) +![ComfyUI 工作流 - 混元视频 1.5 T2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_t2v-1.webp) -**vae** + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Hunyuan Video 1.5 T2V" + + -- [hunyuanvideo15_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/vae/hunyuanvideo15_vae_fp16.safetensors) +### 模型下载 + + + hunyuanvideo1.5_720p_t2v_fp16.safetensors，保存到 ComfyUI/models/diffusion_models/ + + -Model Storage Location +#### 模型存储 ``` -:open_file_folder: ComfyUI/ -├── :open_file_folder: models/ -│ ├── :open_file_folder: text_encoders/ -│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors -│ │ └── byt5_small_glyphxl_fp16.safetensors -│ ├── :open_file_folder: diffusion_models/ -│ │ ├── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors -│ │ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors -│ └── :open_file_folder: vae/ -│ └── hunyuanvideo15_vae_fp16.safetensors +ComfyUI/ +├── 📂 models/ +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // 共用模型 +│ │ └── byt5_small_glyphxl_fp16.safetensors // 共用模型 +│ ├── 📂 vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors // 共用模型 +│ └── 📂 diffusion_models/ +│ └── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V 模型 ``` -## 工作流 +### 运行步骤 + +1. 确保 `DualCLIPLoader` 节点已加载以下模型： + - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` + - clip_name2: `byt5_small_glyphxl_fp16.safetensors` +2. 确保 `Load Diffusion Model` 节点已加载 `hunyuanvideo1.5_720p_t2v_fp16.safetensors` +3. 确保 `Load VAE` 节点已加载 `hunyuanvideo15_vae_fp16.safetensors` +4. 点击 `Queue` 按钮或使用快捷键 `Ctrl(Cmd) + Enter` 运行工作流 + + +工作流包含超分辨率放大节点。启用后，会使用 `hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors` 将输出放大到 1080p。 + + +## 混元视频 1.5 图生视频工作流 + +混元视频 1.5 图生视频可将静态图像转换为流畅、高质量的视频，一致性和运动动态均有提升。 + +![ComfyUI 工作流 - 混元视频 1.5 I2V](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_hunyuan_video_1.5_720p_i2v-1.webp) + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Hunyuan Video 1.5 I2V" + + + +#### 输入素材 -[video_hunyuan_video_1.5_720p_i2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_i2v.json) + + + 下载默认输入图片，或使用你自己的图片作为起始帧。 + + -[video_hunyuan_video_1.5_720p_t2v.json](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_hunyuan_video_1.5_720p_t2v.json) +### 模型下载 + + + + sigclip_vision_patch14_384.safetensors，保存到 ComfyUI/models/clip_vision/ + + + hunyuanvideo1.5_720p_i2v_fp16.safetensors，保存到 ComfyUI/models/diffusion_models/ + + + +#### 模型存储 + +``` +ComfyUI/ +├── 📂 models/ +│ ├── 📂 clip_vision/ +│ │ └── sigclip_vision_patch14_384.safetensors // I2V 视觉编码器 +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors // 共用模型 +│ │ └── byt5_small_glyphxl_fp16.safetensors // 共用模型 +│ ├── 📂 vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors // 共用模型 +│ └── 📂 diffusion_models/ +│ └── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V 模型 +``` + +### 运行步骤 + +1. 确保 `DualCLIPLoader` 节点已加载以下模型： + - clip_name1: `qwen_2.5_vl_7b_fp8_scaled.safetensors` + - clip_name2: `byt5_small_glyphxl_fp16.safetensors` +2. 确保 `CLIPVisionLoader` 节点已加载 `sigclip_vision_patch14_384.safetensors` +3. 确保 `Load Diffusion Model` 节点已加载 `hunyuanvideo1.5_720p_i2v_fp16.safetensors` +4. 确保 `Load VAE` 节点已加载 `hunyuanvideo15_vae_fp16.safetensors` +5. 点击 `Queue` 按钮或使用快捷键 `Ctrl(Cmd) + Enter` 运行工作流 + +## 超分辨率放大器 + +两个工作流都包含可选的超分辨率节点，可将 720p 输出视频放大到 1080p。该放大器使用蒸馏模型，可高效输出高分辨率视频。 + + + hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors，保存到 ComfyUI/models/diffusion_models/ + +所有共用模型与工作流专用模型汇总： + +``` +ComfyUI/ +├── 📂 models/ +│ ├── 📂 clip_vision/ +│ │ └── sigclip_vision_patch14_384.safetensors +│ ├── 📂 text_encoders/ +│ │ ├── qwen_2.5_vl_7b_fp8_scaled.safetensors +│ │ └── byt5_small_glyphxl_fp16.safetensors +│ ├── 📂 vae/ +│ │ └── hunyuanvideo15_vae_fp16.safetensors +│ └── 📂 diffusion_models/ +│ ├── hunyuanvideo1.5_720p_t2v_fp16.safetensors // T2V 模型 +│ ├── hunyuanvideo1.5_720p_i2v_fp16.safetensors // I2V 模型 +│ └── hunyuanvideo1.5_1080p_sr_distilled_fp16.safetensors // 放大模型 +``` diff --git a/zh/tutorials/video/kandinsky/kandinsky-5.mdx b/zh/tutorials/video/kandinsky/kandinsky-5.mdx index 3520fa9c7..19b2fa0e9 100644 --- a/zh/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/zh/tutorials/video/kandinsky/kandinsky-5.mdx @@ -53,10 +53,10 @@ Kandinsky 5.0 使用带有 Flow Matching 的潜在扩散管道，具有以下特 请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `视频` 找到 "Kandinsky 5.0 T2V" 以加载工作流。 - + Download the T2V workflow to use locally - + 在 Comfy Cloud 中打开 @@ -92,13 +92,13 @@ ComfyUI/ 请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `视频` 找到 "Kandinsky 5.0 I2V" 以加载工作流。 - ![Kandinsky 5.0 I2V Workflow Preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/video_kandinsky5_i2v-1.webp) + ![Kandinsky 5.0 I2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_i2v-1.webp) - + Download the I2V workflow to use locally - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/video/ltx/ltx-2-3.mdx b/zh/tutorials/video/ltx/ltx-2-3.mdx index c60e8e284..d02f9d8fc 100644 --- a/zh/tutorials/video/ltx/ltx-2-3.mdx +++ b/zh/tutorials/video/ltx/ltx-2-3.mdx @@ -300,7 +300,6 @@ ComfyUI/ 此工作流使用子图节点实现模块化处理。查阅子图文档，了解如何自定义和扩展工作流。 - #### 输入素材 diff --git a/zh/tutorials/video/ltx/ltx-2.mdx b/zh/tutorials/video/ltx/ltx-2.mdx index 00e360f8b..cdb7bd7f9 100644 --- a/zh/tutorials/video/ltx/ltx-2.mdx +++ b/zh/tutorials/video/ltx/ltx-2.mdx @@ -57,67 +57,65 @@ LTX-2 已原生支持 ComfyUI。开始使用： 从文本提示词生成视频。 - + + 下载工作流 - 在云端打开 - + **蒸馏版本**（更快，8 步）： - + 下载工作流 - ### 图生视频 从输入图像生成视频。 - + + 下载工作流 - 在云端打开 - + **蒸馏版本**（更快，8 步）： - + 下载工作流 - ### 控制生成视频 使用 IC-LoRAs 进行结构控制生成视频。 **深度控制：** + 下载工作流 - 在云端打开 - + **Canny 边缘控制：** - + + 下载工作流 - 在云端打开 - + **姿态控制：** - + + 下载工作流 - 在云端打开 - + ## 提示词技巧 编写 LTX-2 提示词时，请专注于详细、按时间顺序描述动作和场景。在一个连贯的段落中包含具体的动作、外观、镜头角度和环境细节。直接从动作开始，保持描述的字面性和精确性。 diff --git a/zh/tutorials/video/ltxv.mdx b/zh/tutorials/video/ltxv.mdx index 455c2785b..845a5a82b 100644 --- a/zh/tutorials/video/ltxv.mdx +++ b/zh/tutorials/video/ltxv.mdx @@ -34,7 +34,7 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 通过首帧图像控制视频生成：[示例首帧](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)。 - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/video/wan/vace.mdx b/zh/tutorials/video/wan/vace.mdx index e650ee9c1..5f657fce9 100644 --- a/zh/tutorials/video/wan/vace.mdx +++ b/zh/tutorials/video/wan/vace.mdx @@ -235,10 +235,11 @@ VACE 还支持在一张图像中输入多个参考图像，来生成对应的视 请查阅下面的文档了解相关的节点 + WanVaceToVideo 节点文档 - ComfyUI TrimVideoLatent 节点文档 + \ No newline at end of file diff --git a/zh/tutorials/video/wan/wan-alpha.mdx b/zh/tutorials/video/wan/wan-alpha.mdx index deb0828d1..d28ef3e6f 100644 --- a/zh/tutorials/video/wan/wan-alpha.mdx +++ b/zh/tutorials/video/wan/wan-alpha.mdx @@ -11,11 +11,11 @@ translationBlockHashes: --- - import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; -Wan-Alpha 是一个专门的文本生成视频模型,可以生成带有 Alpha 通道透明度的高质量视频。它基于 Wan2.1-14B-T2V 基础模型构建,能够创建具有透明背景和半透明物体的视频,非常适合合成工作流程。 -该模型擅长生成透明背景、半透明物体(气泡、玻璃、水)、发光效果以及具有适当 Alpha 通道的精细细节(头发、烟雾、粒子)。 +Wan-Alpha 是一个专门的文本生成视频模型，可以生成带有 Alpha 通道透明度的高质量视频。它基于 Wan2.1-14B-T2V 基础模型构建，能够创建具有透明背景和半透明物体的视频，非常适合合成工作流程。 + +该模型擅长生成透明背景、半透明物体（气泡、玻璃、水）、发光效果以及具有适当 Alpha 通道的精细细节（头发、烟雾、粒子）。 # 视频教程 @@ -29,11 +29,103 @@ Wan-Alpha 是一个专门的文本生成视频模型,可以生成带有 Alpha allowfullscreen > -[下载工作流](https://github.com/Comfy-Org/workflows/blob/main/tutorial_workflows/Get_Comfy_With_Comfy_Wan_Alpha.json) - ## 资源 - [Wan-Alpha GitHub](https://github.com/WeChatCV/Wan-Alpha) - [Hugging Face 模型](https://huggingface.co/htdong/Wan-Alpha) - [ComfyUI 版本](https://huggingface.co/htdong/Wan-Alpha_ComfyUI) - [研究论文](https://arxiv.org/pdf/2509.24979) + + + +## Wan-Alpha 文生视频工作流 (14B) + +### 1. 下载工作流 + +请将 ComfyUI 更新到最新版本，然后下载工作流文件并拖入 ComfyUI，或在模板库中通过 `Workflow` → `Browse Templates` → `Video` 搜索 "Wan 2.1 Alpha T2V 14B"。 + +Wan-Alpha T2V 工作流 + + + + 在 Comfy Cloud 中打开 + + + 下载 JSON 或在模板库中搜索 "Wan 2.1 Alpha T2V 14B" + + + +### 2. 安装模型 + +所有相关模型可在 [Wan_2.1_ComfyUI_repackaged](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged) 找到。 + +**扩散模型**：选择一个版本： + + + + FP8 量化扩散模型。保存到 ComfyUI/models/diffusion_models/ + + + BF16 扩散模型。保存到 ComfyUI/models/diffusion_models/ + + + +**文本编码器** + + + + FP8 文本编码器。保存到 ComfyUI/models/text_encoders/ + + + +**VAE 模型** + + + + RGB 通道 VAE。保存到 ComfyUI/models/vae/ + + + Alpha 通道 VAE。保存到 ComfyUI/models/vae/ + + + +**LoRA 模型** + + + + Alpha 生成 LoRA。保存到 ComfyUI/models/loras/ + + + 快速推理 Lightning LoRA。保存到 ComfyUI/models/loras/ + + + +文件保存位置： + +``` +ComfyUI/ +├── models/ +│ ├── diffusion_models/ +│ │ ├── wan2.1_t2v_14B_fp8_scaled.safetensors +│ │ └── wan2.1_t2v_14B_bf16.safetensors +│ ├── text_encoders/ +│ │ └── umt5_xxl_fp8_e4m3fn_scaled.safetensors +│ ├── vae/ +│ │ ├── wan_alpha_2.1_vae_rgb_channel.safetensors +│ │ └── wan_alpha_2.1_vae_alpha_channel.safetensors +│ └── loras/ +│ ├── wan_alpha_2.1_rgba_lora.safetensors +│ └── lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors +``` + +### 3. 按步骤完成工作流 + +1. 确保 `Load Diffusion Model` 节点已加载正确的扩散模型 +2. 确保 `Load CLIP` 节点已加载 `umt5_xxl_fp8_e4m3fn_scaled.safetensors` +3. 确保 RGB 的 `Load VAE` 节点已加载 `wan_alpha_2.1_vae_rgb_channel.safetensors` +4. 确保 Alpha 的第二个 `Load VAE` 节点已加载 `wan_alpha_2.1_vae_alpha_channel.safetensors` +5. 确保 `LoRA Loader` 节点已加载 `wan_alpha_2.1_rgba_lora.safetensors` +6. （可选）在 Lightning LoRA 节点中加载 `lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16.safetensors` 以加快生成 +7. 在 `CLIP Text Encode (Positive Prompt)` 节点中设置正向提示词 +8. （可选）在 `EmptyHunyuanLatentVideo` 节点中修改视频尺寸 +9. 点击 `Run` 按钮或使用 `Ctrl(Cmd) + Enter` 执行生成 diff --git a/zh/tutorials/video/wan/wan-causal-forcing.mdx b/zh/tutorials/video/wan/wan-causal-forcing.mdx index f45014dbb..9842a427f 100644 --- a/zh/tutorials/video/wan/wan-causal-forcing.mdx +++ b/zh/tutorials/video/wan/wan-causal-forcing.mdx @@ -43,7 +43,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 本工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 - ### Causal Forcing 与 Causal Forcing++ | 模式 | 说明 | diff --git a/zh/tutorials/video/wan/wan2_2.mdx b/zh/tutorials/video/wan/wan2_2.mdx index 44406d6cb..ea705fa99 100644 --- a/zh/tutorials/video/wan/wan2_2.mdx +++ b/zh/tutorials/video/wan/wan2_2.mdx @@ -106,7 +106,7 @@ Wan2.2 5B 版本配合 ComfyUI 原生 offloading功能，能很好地适配 8GB Download JSON or search "Wan2.2 5B" in Template Library - + 在 Comfy Cloud 中打开 @@ -231,7 +231,7 @@ ComfyUI/ Download JSON or search "Wan2.2 14B T2V" in Template Library - + 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/video/zai/scail2.mdx b/zh/tutorials/video/zai/scail2.mdx index 415b478b3..c2a423081 100644 --- a/zh/tutorials/video/zai/scail2.mdx +++ b/zh/tutorials/video/zai/scail2.mdx @@ -54,7 +54,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 本工作流使用 Subgraph 节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 - ### 长视频 对于较长的视频，计算段落数量：`ceil(total_frames / 76)`。除第一段外均使用 Extend 子图。复制 Extend 节点以增加更多段落，链接 `previous_frames` 输出，并递增 `segment_index`。 From 2a3e669187c0b64fc885738ef258e6a3eb1d0ebb Mon Sep 17 00:00:00 2001 From: ComfyUI Wiki Date: Fri, 24 Jul 2026 20:53:02 +0800 Subject: [PATCH 12/19] docs: finish tutorial workflow standardization and Partner Nodes nav grouping Standardize workflow pages across tutorials (CardGroup, blob URLs, UTM, preview/output assets) in EN/zh/ja/ko, group Partner Nodes sidebar by modality, reorder Image/LTX sections, and add tutorial-workflow-standard skill scripts. --- .../skills/tutorial-workflow-standard/SPEC.md | 142 ++ .../fix_webp_misuse.py | 140 ++ .../sync_i18n_workflows.py | 161 ++ docs.json | 1528 ++++++++++------- ja/tutorials/3d/hunyuan3D-2.mdx | 200 ++- ja/tutorials/3d/triposplat.mdx | 139 +- ja/tutorials/audio/ace-step/ace-step-v1-5.mdx | 93 +- ja/tutorials/audio/ace-step/ace-step-v1.mdx | 1 - .../audio/stable-audio/stable-audio-1.mdx | 55 +- .../audio/stable-audio/stable-audio-3.mdx | 101 +- ja/tutorials/basic/image-to-image.mdx | 1 - ja/tutorials/basic/inpaint.mdx | 1 - ja/tutorials/basic/lora.mdx | 1 - ja/tutorials/basic/multiple-loras.mdx | 1 - ja/tutorials/basic/outpaint.mdx | 1 - ja/tutorials/basic/text-to-image.mdx | 1 - ja/tutorials/basic/upscale.mdx | 2 - ja/tutorials/controlnet/controlnet.mdx | 1 - ja/tutorials/controlnet/depth-controlnet.mdx | 1 - ja/tutorials/controlnet/depth-t2i-adapter.mdx | 1 - .../controlnet/mixing-controlnets.mdx | 1 - .../controlnet/pose-controlnet-2-pass.mdx | 1 - ja/tutorials/flux/flux-1-controlnet.mdx | 56 +- ja/tutorials/flux/flux-1-fill-dev.mdx | 41 +- ja/tutorials/flux/flux-1-kontext-dev.mdx | 37 +- ja/tutorials/flux/flux-1-text-to-image.mdx | 118 +- ja/tutorials/flux/flux-1-uso.mdx | 43 +- ja/tutorials/flux/flux-2-dev.mdx | 1 - ja/tutorials/flux/flux-2-klein.mdx | 2 - ja/tutorials/flux/flux1-krea-dev.mdx | 79 +- ja/tutorials/image/anima/anima.mdx | 59 +- ja/tutorials/image/boogu/boogu-image-0.1.mdx | 1 - .../image/cosmos/cosmos-predict2-t2i.mdx | 2 - .../image/ernie-image/ernie-image.mdx | 74 +- ja/tutorials/image/hidream/hidream-e1.mdx | 69 +- ja/tutorials/image/hidream/hidream-i1.mdx | 161 +- ja/tutorials/image/hidream/hidream-o1.mdx | 98 +- ja/tutorials/image/ideogram/ideogram-v4.mdx | 35 +- ja/tutorials/image/krea/krea-2.mdx | 64 +- ja/tutorials/image/lens/lens.mdx | 55 +- .../newbie-image/newbie-image-exp-0-1.mdx | 27 +- ja/tutorials/image/omnigen/omnigen2.mdx | 50 +- ja/tutorials/image/ovis/ovis-image.mdx | 17 +- ja/tutorials/image/pixeldit/pixeldit.mdx | 2 - ja/tutorials/image/qwen/qwen-image-2512.mdx | 4 - ja/tutorials/image/qwen/qwen-image-edit.mdx | 40 +- .../image/qwen/qwen-image-layered.mdx | 2 - ja/tutorials/image/qwen/qwen-image.mdx | 156 +- ja/tutorials/image/z-image/z-image-turbo.mdx | 63 +- ja/tutorials/llm/gemma4/gemma4.mdx | 86 +- ja/tutorials/llm/qwen/qwen3.mdx | 69 +- ja/tutorials/llm/qwen/qwen3_5.mdx | 79 +- .../partner-nodes/beeble/beeble-switchx.mdx | 2 - .../flux-1-1-pro-ultra-image.mdx | 1 - .../black-forest-labs/flux-1-kontext.mdx | 1 - .../partner-nodes/bria/background-removal.mdx | 2 - .../bytedance/seed-audio-1-0.mdx | 1 - .../bytedance/seedance-2-0-real-human.mdx | 2 - .../partner-nodes/bytedance/seedance-2-0.mdx | 2 - .../bytedance/seedream-5-lite.mdx | 2 - .../google/gemini-omni-flash.mdx | 2 - .../google/nano-banana-2-lite.mdx | 1 - .../partner-nodes/google/nano-banana-2.mdx | 2 - .../partner-nodes/google/nano-banana-pro.mdx | 1 - .../happyhorse/happyhorse1-0.mdx | 2 - .../happyhorse/happyhorse1-1.mdx | 2 - .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 2 - .../hunyuan3d/model-generation.mdx | 2 - .../partner-nodes/ideogram/ideogram-v4.mdx | 2 - .../kling/kling-motion-control.mdx | 2 - .../partner-nodes/krea2/krea2-t2i.mdx | 1 - .../partner-nodes/luma/luma-uni-1.mdx | 2 - ja/tutorials/partner-nodes/meshy/meshy-6.mdx | 1 - .../moonvalley-video-generation.mdx | 6 - .../partner-nodes/openai/dall-e-2.mdx | 1 - .../partner-nodes/openai/dall-e-3.mdx | 2 - .../partner-nodes/openai/gpt-image-1.mdx | 1 - .../partner-nodes/openai/gpt-image-2.mdx | 2 - ja/tutorials/partner-nodes/openrouter/llm.mdx | 1 - ja/tutorials/partner-nodes/overview.mdx | 6 - ja/tutorials/partner-nodes/pricing.mdx | 1 - .../partner-nodes/recraft/recraft-v4.mdx | 2 - .../partner-nodes/reve/reve-image.mdx | 1 - .../partner-nodes/rodin/model-generation.mdx | 2 - .../partner-nodes/runway/image-generation.mdx | 1 - .../partner-nodes/runway/video-generation.mdx | 1 - .../partner-nodes/sonilo/video-to-music.mdx | 1 - .../partner-nodes/tripo/model-generation.mdx | 2 - .../partner-nodes/tripo/tripo-3-1.mdx | 3 - ja/tutorials/partner-nodes/wan/wan2-7.mdx | 2 - ja/tutorials/utility/depth-anything-3.mdx | 110 +- .../utility/face-detection/mediapipe.mdx | 120 +- ja/tutorials/utility/image-upscale.mdx | 1 - ja/tutorials/utility/moge.mdx | 127 +- .../utility/pose-detection-sdpose.mdx | 217 ++- ja/tutorials/utility/preprocessors.mdx | 206 ++- .../utility/remove-background-birefnet.mdx | 64 +- ja/tutorials/utility/seedvr2.mdx | 144 +- ja/tutorials/utility/video-segment-sam3.mdx | 115 +- ja/tutorials/utility/video-upscale.mdx | 1 - .../utility/void-video-inpainting.mdx | 100 +- ja/tutorials/video/bytedance/bernini-r.mdx | 43 +- .../video/hunyuan/hunyuan-video-1-5.mdx | 1 - ja/tutorials/video/hunyuan/hunyuan-video.mdx | 3 - ja/tutorials/video/kandinsky/kandinsky-5.mdx | 6 +- ja/tutorials/video/ltx/ltx-2-3.mdx | 11 +- ja/tutorials/video/ltx/ltx-2.mdx | 2 - ja/tutorials/video/ltxv.mdx | 18 +- ja/tutorials/video/wan/fun-camera.mdx | 2 - ja/tutorials/video/wan/fun-control.mdx | 2 - ja/tutorials/video/wan/fun-inp.mdx | 2 - ja/tutorials/video/wan/vace.mdx | 2 - ja/tutorials/video/wan/wan-alpha.mdx | 1 - ja/tutorials/video/wan/wan-ati.mdx | 3 - ja/tutorials/video/wan/wan-causal-forcing.mdx | 1 - ja/tutorials/video/wan/wan-dancer.mdx | 2 + ja/tutorials/video/wan/wan-flf.mdx | 2 - ja/tutorials/video/wan/wan-move.mdx | 4 +- ja/tutorials/video/wan/wan-video.mdx | 3 - ja/tutorials/video/wan/wan2-2-animate.mdx | 4 +- ja/tutorials/video/wan/wan2-2-fun-camera.mdx | 2 + ja/tutorials/video/wan/wan2-2-fun-control.mdx | 6 +- ja/tutorials/video/wan/wan2-2-fun-inp.mdx | 4 +- ja/tutorials/video/wan/wan2-2-s2v.mdx | 8 +- ja/tutorials/video/wan/wan2_2.mdx | 12 +- ja/tutorials/video/zai/scail2.mdx | 4 +- ko/tutorials/3d/hunyuan3D-2.mdx | 230 ++- ko/tutorials/3d/triposplat.mdx | 136 +- ko/tutorials/audio/ace-step/ace-step-v1-5.mdx | 93 +- ko/tutorials/audio/ace-step/ace-step-v1.mdx | 1 - .../audio/stable-audio/stable-audio-1.mdx | 55 +- .../audio/stable-audio/stable-audio-3.mdx | 101 +- ko/tutorials/basic/image-to-image.mdx | 1 - ko/tutorials/basic/inpaint.mdx | 1 - ko/tutorials/basic/lora.mdx | 1 - ko/tutorials/basic/multiple-loras.mdx | 1 - ko/tutorials/basic/outpaint.mdx | 1 - ko/tutorials/basic/text-to-image.mdx | 1 - ko/tutorials/basic/upscale.mdx | 2 - ko/tutorials/controlnet/controlnet.mdx | 1 - ko/tutorials/controlnet/depth-controlnet.mdx | 1 - ko/tutorials/controlnet/depth-t2i-adapter.mdx | 1 - .../controlnet/mixing-controlnets.mdx | 1 - .../controlnet/pose-controlnet-2-pass.mdx | 1 - ko/tutorials/flux/flux-1-controlnet.mdx | 60 +- ko/tutorials/flux/flux-1-fill-dev.mdx | 41 +- ko/tutorials/flux/flux-1-kontext-dev.mdx | 39 +- ko/tutorials/flux/flux-1-text-to-image.mdx | 122 +- ko/tutorials/flux/flux-1-uso.mdx | 34 +- ko/tutorials/flux/flux-2-dev.mdx | 1 - ko/tutorials/flux/flux-2-klein.mdx | 2 - ko/tutorials/flux/flux1-krea-dev.mdx | 68 +- ko/tutorials/image/anima/anima.mdx | 59 +- ko/tutorials/image/boogu/boogu-image-0.1.mdx | 1 - .../image/cosmos/cosmos-predict2-t2i.mdx | 3 - .../image/ernie-image/ernie-image.mdx | 74 +- ko/tutorials/image/hidream/hidream-e1.mdx | 75 +- ko/tutorials/image/hidream/hidream-i1.mdx | 192 ++- ko/tutorials/image/hidream/hidream-o1.mdx | 96 +- ko/tutorials/image/ideogram/ideogram-v4.mdx | 37 +- ko/tutorials/image/krea/krea-2.mdx | 63 +- ko/tutorials/image/lens/lens.mdx | 63 +- .../newbie-image/newbie-image-exp-0-1.mdx | 19 +- ko/tutorials/image/omnigen/omnigen2.mdx | 54 +- ko/tutorials/image/ovis/ovis-image.mdx | 15 +- ko/tutorials/image/pixeldit/pixeldit.mdx | 2 - ko/tutorials/image/qwen/qwen-image-2512.mdx | 2 - .../image/qwen/qwen-image-edit-2511.mdx | 1 - ko/tutorials/image/qwen/qwen-image-edit.mdx | 36 +- .../image/qwen/qwen-image-layered.mdx | 3 - ko/tutorials/image/qwen/qwen-image.mdx | 122 +- ko/tutorials/image/z-image/z-image-turbo.mdx | 65 +- ko/tutorials/llm/gemma4/gemma4.mdx | 87 +- ko/tutorials/llm/qwen/qwen3.mdx | 70 +- ko/tutorials/llm/qwen/qwen3_5.mdx | 80 +- .../partner-nodes/anthropic/claude.mdx | 1 - .../partner-nodes/beeble/beeble-switchx.mdx | 2 - .../flux-1-1-pro-ultra-image.mdx | 1 - .../black-forest-labs/flux-1-kontext.mdx | 1 - .../partner-nodes/bria/background-removal.mdx | 2 - .../bytedance/seed-audio-1-0.mdx | 1 - .../bytedance/seedance-2-0-real-human.mdx | 2 - .../partner-nodes/bytedance/seedance-2-0.mdx | 2 - .../bytedance/seedream-5-lite.mdx | 2 - .../google/gemini-omni-flash.mdx | 2 - .../google/nano-banana-2-lite.mdx | 1 - .../partner-nodes/google/nano-banana-2.mdx | 2 - .../partner-nodes/google/nano-banana-pro.mdx | 1 - .../happyhorse/happyhorse1-0.mdx | 2 - .../happyhorse/happyhorse1-1.mdx | 2 - .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 2 - .../hunyuan3d/model-generation.mdx | 2 - .../partner-nodes/ideogram/ideogram-v4.mdx | 2 - .../kling/kling-motion-control.mdx | 2 - .../partner-nodes/krea2/krea2-t2i.mdx | 1 - .../partner-nodes/luma/luma-uni-1.mdx | 2 - ko/tutorials/partner-nodes/meshy/meshy-6.mdx | 1 - .../moonvalley-video-generation.mdx | 6 - .../partner-nodes/openai/dall-e-2.mdx | 1 - .../partner-nodes/openai/dall-e-3.mdx | 2 - .../partner-nodes/openai/gpt-image-1.mdx | 1 - .../partner-nodes/openai/gpt-image-2.mdx | 2 - ko/tutorials/partner-nodes/openrouter/llm.mdx | 1 - ko/tutorials/partner-nodes/overview.mdx | 5 - ko/tutorials/partner-nodes/pricing.mdx | 1 - .../partner-nodes/recraft/recraft-v4.mdx | 2 - .../partner-nodes/reve/reve-image.mdx | 1 - .../partner-nodes/rodin/model-generation.mdx | 2 - .../partner-nodes/runway/image-generation.mdx | 1 - .../partner-nodes/runway/video-generation.mdx | 2 - .../partner-nodes/sonilo/video-to-music.mdx | 1 - .../partner-nodes/tripo/model-generation.mdx | 5 - .../partner-nodes/tripo/tripo-3-1.mdx | 2 - ko/tutorials/partner-nodes/wan/wan2-7.mdx | 2 - ko/tutorials/utility/depth-anything-3.mdx | 114 +- .../utility/face-detection/mediapipe.mdx | 118 +- ko/tutorials/utility/image-upscale.mdx | 1 - ko/tutorials/utility/moge.mdx | 130 +- .../utility/pose-detection-sdpose.mdx | 217 ++- ko/tutorials/utility/preprocessors.mdx | 206 ++- .../utility/remove-background-birefnet.mdx | 64 +- ko/tutorials/utility/seedvr2.mdx | 144 +- ko/tutorials/utility/video-segment-sam3.mdx | 115 +- ko/tutorials/utility/video-upscale.mdx | 1 - .../utility/void-video-inpainting.mdx | 98 +- ko/tutorials/video/bytedance/bernini-r.mdx | 43 +- .../video/hunyuan/hunyuan-video-1-5.mdx | 1 - ko/tutorials/video/hunyuan/hunyuan-video.mdx | 3 - ko/tutorials/video/kandinsky/kandinsky-5.mdx | 6 +- ko/tutorials/video/ltx/ltx-2-3.mdx | 11 +- ko/tutorials/video/ltx/ltx-2.mdx | 81 +- ko/tutorials/video/ltxv.mdx | 18 +- ko/tutorials/video/wan/fun-camera.mdx | 6 +- ko/tutorials/video/wan/fun-control.mdx | 2 - ko/tutorials/video/wan/fun-inp.mdx | 2 - ko/tutorials/video/wan/wan-alpha.mdx | 1 - ko/tutorials/video/wan/wan-ati.mdx | 4 - ko/tutorials/video/wan/wan-causal-forcing.mdx | 1 - ko/tutorials/video/wan/wan-dancer.mdx | 2 + ko/tutorials/video/wan/wan-flf.mdx | 2 - ko/tutorials/video/wan/wan-move.mdx | 2 - ko/tutorials/video/wan/wan-video.mdx | 3 - ko/tutorials/video/wan/wan2-2-animate.mdx | 5 +- ko/tutorials/video/wan/wan2-2-fun-control.mdx | 6 +- ko/tutorials/video/wan/wan2-2-fun-inp.mdx | 3 +- ko/tutorials/video/wan/wan2-2-s2v.mdx | 7 +- ko/tutorials/video/wan/wan2_2.mdx | 14 +- ko/tutorials/video/zai/scail2.mdx | 4 +- tutorials/3d/hunyuan3D-2.mdx | 76 +- tutorials/3d/triposplat.mdx | 22 +- tutorials/audio/ace-step/ace-step-v1-5.mdx | 42 +- tutorials/audio/ace-step/ace-step-v1.mdx | 2 - .../audio/stable-audio/stable-audio-1.mdx | 21 +- .../audio/stable-audio/stable-audio-3.mdx | 46 +- tutorials/basic/multiple-loras.mdx | 1 - tutorials/basic/upscale.mdx | 1 - tutorials/flux/flux-1-controlnet.mdx | 58 +- tutorials/flux/flux-1-fill-dev.mdx | 67 +- tutorials/flux/flux-1-kontext-dev.mdx | 37 +- tutorials/flux/flux-1-text-to-image.mdx | 51 +- tutorials/flux/flux-1-uso.mdx | 37 +- tutorials/flux/flux-2-dev.mdx | 70 +- tutorials/flux/flux-2-klein.mdx | 122 ++ tutorials/flux/flux1-krea-dev.mdx | 15 +- tutorials/image/anima/anima.mdx | 52 +- tutorials/image/boogu/boogu-image-0.1.mdx | 61 +- .../image/cosmos/cosmos-predict2-t2i.mdx | 3 - tutorials/image/ernie-image/ernie-image.mdx | 73 +- tutorials/image/hidream/hidream-e1.mdx | 62 +- tutorials/image/hidream/hidream-i1.mdx | 32 +- tutorials/image/hidream/hidream-o1.mdx | 83 +- tutorials/image/ideogram/ideogram-v4.mdx | 37 +- tutorials/image/krea/krea-2.mdx | 71 +- tutorials/image/lens/lens.mdx | 38 +- .../newbie-image/newbie-image-exp-0-1.mdx | 18 +- tutorials/image/omnigen/omnigen2.mdx | 58 +- tutorials/image/ovis/ovis-image.mdx | 10 +- tutorials/image/qwen/qwen-image-2512.mdx | 13 +- tutorials/image/qwen/qwen-image-edit-2511.mdx | 36 +- tutorials/image/qwen/qwen-image-edit.mdx | 37 +- tutorials/image/qwen/qwen-image-layered.mdx | 39 +- tutorials/image/qwen/qwen-image.mdx | 114 +- tutorials/image/z-image/z-image-turbo.mdx | 49 +- tutorials/image/z-image/z-image.mdx | 33 +- tutorials/llm/gemma4/gemma4.mdx | 37 +- tutorials/llm/qwen/qwen3.mdx | 26 +- tutorials/llm/qwen/qwen3_5.mdx | 38 +- .../partner-nodes/beeble/beeble-switchx.mdx | 4 +- .../partner-nodes/bria/background-removal.mdx | 124 +- tutorials/partner-nodes/bria/fibo.mdx | 60 +- .../bytedance/seed-audio-1-0.mdx | 54 +- .../bytedance/seedream-5-lite.mdx | 4 +- .../bytedance/seedream-5-pro.mdx | 4 +- .../google/gemini-omni-flash.mdx | 6 +- tutorials/partner-nodes/google/gemini.mdx | 30 +- .../google/nano-banana-2-lite.mdx | 43 +- .../partner-nodes/google/nano-banana-2.mdx | 2 +- .../partner-nodes/google/nano-banana-pro.mdx | 27 +- .../grok/grok-imagine-video-1-5.mdx | 2 +- .../happyhorse/happyhorse1-1.mdx | 6 +- .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 118 +- .../hunyuan3d/model-generation.mdx | 74 +- tutorials/partner-nodes/kling/kling-3-0.mdx | 81 +- .../kling/kling-motion-control.mdx | 33 +- tutorials/partner-nodes/krea2/krea2-t2i.mdx | 4 +- .../moonvalley-video-generation.mdx | 5 - tutorials/partner-nodes/openai/chat.mdx | 28 +- tutorials/partner-nodes/openai/dall-e-3.mdx | 1 - tutorials/partner-nodes/openrouter/llm.mdx | 27 +- tutorials/partner-nodes/overview.mdx | 4 - .../partner-nodes/recraft/recraft-v4.mdx | 4 +- tutorials/partner-nodes/reve/reve-image.mdx | 86 +- .../partner-nodes/runway/video-generation.mdx | 1 - .../partner-nodes/sonilo/video-to-music.mdx | 44 +- .../partner-nodes/tripo/model-generation.mdx | 1 - tutorials/partner-nodes/tripo/tripo-3-1.mdx | 1 - tutorials/utility/depth-anything-3.mdx | 65 +- .../utility/face-detection/mediapipe.mdx | 37 +- tutorials/utility/frame-interpolation.mdx | 1 - tutorials/utility/moge.mdx | 107 +- tutorials/utility/pose-detection-sdpose.mdx | 130 +- tutorials/utility/preprocessors.mdx | 139 +- .../utility/remove-background-birefnet.mdx | 27 +- tutorials/utility/seedvr2.mdx | 116 +- tutorials/utility/video-segment-sam3.mdx | 64 +- tutorials/utility/void-video-inpainting.mdx | 43 +- tutorials/video/bytedance/bernini-r.mdx | 27 +- .../cosmos/cosmos-predict2-video2world.mdx | 6 +- tutorials/video/hunyuan/hunyuan-video-1-5.mdx | 2 +- tutorials/video/hunyuan/hunyuan-video.mdx | 13 +- tutorials/video/kandinsky/kandinsky-5.mdx | 41 +- tutorials/video/ltx/ltx-2-3.mdx | 44 +- tutorials/video/ltx/ltx-2.mdx | 24 +- tutorials/video/ltxv.mdx | 30 +- tutorials/video/wan/fun-camera.mdx | 44 +- tutorials/video/wan/fun-control.mdx | 8 +- tutorials/video/wan/fun-inp.mdx | 2 +- tutorials/video/wan/vace.mdx | 92 +- tutorials/video/wan/wan-alpha.mdx | 2 +- tutorials/video/wan/wan-ati.mdx | 2 +- tutorials/video/wan/wan-causal-forcing.mdx | 4 +- tutorials/video/wan/wan-dancer.mdx | 24 +- tutorials/video/wan/wan-flf.mdx | 2 +- tutorials/video/wan/wan-move.mdx | 16 +- tutorials/video/wan/wan-video.mdx | 34 +- tutorials/video/wan/wan2-2-animate.mdx | 32 +- tutorials/video/wan/wan2-2-fun-camera.mdx | 4 +- tutorials/video/wan/wan2-2-fun-control.mdx | 5 +- tutorials/video/wan/wan2-2-fun-inp.mdx | 8 +- tutorials/video/wan/wan2-2-s2v.mdx | 30 +- tutorials/video/wan/wan2_2.mdx | 104 +- tutorials/video/zai/scail2.mdx | 14 +- zh/tutorials/3d/hunyuan3D-2.mdx | 204 ++- zh/tutorials/3d/triposplat.mdx | 139 +- zh/tutorials/audio/ace-step/ace-step-v1-5.mdx | 97 +- zh/tutorials/audio/ace-step/ace-step-v1.mdx | 8 - .../audio/stable-audio/stable-audio-1.mdx | 55 +- .../audio/stable-audio/stable-audio-3.mdx | 101 +- zh/tutorials/basic/image-to-image.mdx | 9 - zh/tutorials/basic/inpaint.mdx | 1 - zh/tutorials/basic/lora.mdx | 1 - zh/tutorials/basic/multiple-loras.mdx | 1 - zh/tutorials/basic/outpaint.mdx | 2 - zh/tutorials/basic/text-to-image.mdx | 2 - zh/tutorials/basic/upscale.mdx | 1 - zh/tutorials/controlnet/controlnet.mdx | 1 - zh/tutorials/controlnet/depth-controlnet.mdx | 2 - zh/tutorials/controlnet/depth-t2i-adapter.mdx | 1 - .../controlnet/mixing-controlnets.mdx | 1 - .../controlnet/pose-controlnet-2-pass.mdx | 1 - zh/tutorials/flux/flux-1-controlnet.mdx | 56 +- zh/tutorials/flux/flux-1-fill-dev.mdx | 82 +- zh/tutorials/flux/flux-1-kontext-dev.mdx | 38 +- zh/tutorials/flux/flux-1-text-to-image.mdx | 76 +- zh/tutorials/flux/flux-1-uso.mdx | 43 +- zh/tutorials/flux/flux-2-dev.mdx | 1 - zh/tutorials/flux/flux-2-klein.mdx | 2 - zh/tutorials/flux/flux1-krea-dev.mdx | 59 +- zh/tutorials/image/anima/anima.mdx | 59 +- zh/tutorials/image/boogu/boogu-image-0.1.mdx | 1 - .../image/cosmos/cosmos-predict2-t2i.mdx | 1 - .../image/ernie-image/ernie-image.mdx | 74 +- zh/tutorials/image/hidream/hidream-e1.mdx | 71 +- zh/tutorials/image/hidream/hidream-i1.mdx | 169 +- zh/tutorials/image/hidream/hidream-o1.mdx | 98 +- zh/tutorials/image/ideogram/ideogram-v4.mdx | 37 +- zh/tutorials/image/krea/krea-2.mdx | 67 +- zh/tutorials/image/lens/lens.mdx | 59 +- .../newbie-image/newbie-image-exp-0-1.mdx | 21 +- zh/tutorials/image/omnigen/omnigen2.mdx | 55 +- zh/tutorials/image/ovis/ovis-image.mdx | 13 +- zh/tutorials/image/pixeldit/pixeldit.mdx | 2 - zh/tutorials/image/qwen/qwen-image-2512.mdx | 2 - .../image/qwen/qwen-image-edit-2511.mdx | 2 - zh/tutorials/image/qwen/qwen-image-edit.mdx | 36 +- .../image/qwen/qwen-image-layered.mdx | 3 - zh/tutorials/image/qwen/qwen-image.mdx | 128 +- zh/tutorials/image/z-image/z-image-turbo.mdx | 66 +- zh/tutorials/llm/gemma4/gemma4.mdx | 86 +- zh/tutorials/llm/qwen/qwen3.mdx | 69 +- zh/tutorials/llm/qwen/qwen3_5.mdx | 79 +- .../partner-nodes/anthropic/claude.mdx | 2 - .../partner-nodes/beeble/beeble-switchx.mdx | 2 - .../flux-1-1-pro-ultra-image.mdx | 1 - .../black-forest-labs/flux-1-kontext.mdx | 2 - .../partner-nodes/bria/background-removal.mdx | 2 - zh/tutorials/partner-nodes/bria/fibo.mdx | 1 - .../bytedance/seed-audio-1-0.mdx | 1 - .../bytedance/seedance-2-0-real-human.mdx | 10 +- .../partner-nodes/bytedance/seedance-2-0.mdx | 2 - .../bytedance/seedream-5-lite.mdx | 4 - .../bytedance/seedream-5-pro.mdx | 1 - .../google/gemini-omni-flash.mdx | 2 - .../google/nano-banana-2-lite.mdx | 1 - .../partner-nodes/google/nano-banana-2.mdx | 2 - .../partner-nodes/google/nano-banana-pro.mdx | 1 - .../happyhorse/happyhorse1-0.mdx | 2 - .../happyhorse/happyhorse1-1.mdx | 2 - .../partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx | 2 - .../hunyuan3d/model-generation.mdx | 2 - .../partner-nodes/ideogram/ideogram-v4.mdx | 2 - .../kling/kling-motion-control.mdx | 35 +- .../partner-nodes/krea2/krea2-t2i.mdx | 1 - .../partner-nodes/luma/luma-uni-1.mdx | 4 - zh/tutorials/partner-nodes/meshy/meshy-6.mdx | 1 - .../moonvalley-video-generation.mdx | 6 - .../partner-nodes/openai/dall-e-2.mdx | 2 - .../partner-nodes/openai/dall-e-3.mdx | 3 - .../partner-nodes/openai/gpt-image-1.mdx | 3 - .../partner-nodes/openai/gpt-image-2.mdx | 2 - zh/tutorials/partner-nodes/openrouter/llm.mdx | 1 - zh/tutorials/partner-nodes/overview.mdx | 2 - zh/tutorials/partner-nodes/pricing.mdx | 1 - .../partner-nodes/recraft/recraft-v4.mdx | 2 - .../partner-nodes/reve/reve-image.mdx | 1 - .../partner-nodes/rodin/model-generation.mdx | 2 - .../partner-nodes/runway/image-generation.mdx | 1 - .../partner-nodes/runway/video-generation.mdx | 2 - .../partner-nodes/sonilo/video-to-music.mdx | 1 - .../partner-nodes/tripo/model-generation.mdx | 6 - .../partner-nodes/tripo/tripo-3-1.mdx | 3 - zh/tutorials/partner-nodes/wan/wan2-7.mdx | 64 +- zh/tutorials/utility/depth-anything-3.mdx | 110 +- .../utility/face-detection/mediapipe.mdx | 120 +- zh/tutorials/utility/frame-interpolation.mdx | 1 - zh/tutorials/utility/image-upscale.mdx | 1 - zh/tutorials/utility/moge.mdx | 127 +- .../utility/pose-detection-sdpose.mdx | 217 ++- zh/tutorials/utility/preprocessors.mdx | 204 ++- .../utility/remove-background-birefnet.mdx | 64 +- zh/tutorials/utility/seedvr2.mdx | 144 +- zh/tutorials/utility/video-segment-sam3.mdx | 117 +- zh/tutorials/utility/video-upscale.mdx | 1 - .../utility/void-video-inpainting.mdx | 100 +- zh/tutorials/video/bytedance/bernini-r.mdx | 43 +- .../cosmos/cosmos-predict2-video2world.mdx | 2 - .../video/hunyuan/hunyuan-video-1-5.mdx | 1 - zh/tutorials/video/hunyuan/hunyuan-video.mdx | 3 - zh/tutorials/video/kandinsky/kandinsky-5.mdx | 7 +- zh/tutorials/video/ltx/ltx-2-3.mdx | 11 +- zh/tutorials/video/ltx/ltx-2.mdx | 43 +- zh/tutorials/video/ltxv.mdx | 18 +- zh/tutorials/video/wan/fun-camera.mdx | 3 - zh/tutorials/video/wan/fun-control.mdx | 5 - zh/tutorials/video/wan/fun-inp.mdx | 3 - zh/tutorials/video/wan/vace.mdx | 2 - zh/tutorials/video/wan/wan-alpha.mdx | 1 - zh/tutorials/video/wan/wan-ati.mdx | 5 - zh/tutorials/video/wan/wan-causal-forcing.mdx | 1 - zh/tutorials/video/wan/wan-dancer.mdx | 2 + zh/tutorials/video/wan/wan-move.mdx | 2 - zh/tutorials/video/wan/wan-video.mdx | 3 - zh/tutorials/video/wan/wan2-2-animate.mdx | 3 +- zh/tutorials/video/wan/wan2-2-fun-camera.mdx | 6 +- zh/tutorials/video/wan/wan2-2-fun-control.mdx | 8 +- zh/tutorials/video/wan/wan2-2-fun-inp.mdx | 13 +- zh/tutorials/video/wan/wan2-2-s2v.mdx | 6 +- zh/tutorials/video/wan/wan2_2.mdx | 9 +- zh/tutorials/video/zai/scail2.mdx | 4 +- 479 files changed, 10794 insertions(+), 6203 deletions(-) create mode 100644 .cursor/skills/tutorial-workflow-standard/SPEC.md create mode 100644 .cursor/skills/tutorial-workflow-standard/fix_webp_misuse.py create mode 100644 .cursor/skills/tutorial-workflow-standard/sync_i18n_workflows.py diff --git a/.cursor/skills/tutorial-workflow-standard/SPEC.md b/.cursor/skills/tutorial-workflow-standard/SPEC.md new file mode 100644 index 000000000..759905a1c --- /dev/null +++ b/.cursor/skills/tutorial-workflow-standard/SPEC.md @@ -0,0 +1,142 @@ +# Tutorial workflow page standardization spec + +Apply to any MDX tutorial page that links to `cloud.comfy.org/?template=` or `workflow_templates`. + +## Source of truth for template metadata + +- Index: `/Users/linmoumou/Documents/comfy/workflow_templates/templates/index.json` +- Local templates JSON: `/Users/linmoumou/Documents/comfy/workflow_templates/templates/{name}.json` +- Manifest (webp filenames): `/Users/linmoumou/Documents/comfy/workflow_templates/packages/core/src/comfyui_workflow_templates_core/manifest.json` + +For each `template` name referenced on a page, look up `name`, `title`, `description`, `io.inputs`, `io.outputs`, `thumbnail`, `thumbnailVariant` in `index.json`. + +## Per-workflow section structure + +Each distinct template on a page gets its own `###` subsection: + +```mdx +### {title from index} (`{template_name}`) + +{description from index — one short paragraph, can trim if already covered above} + +{title} workflow preview + +### Workflow preview images (`templates/*-N.webp`) + +- Files like `{template_name}-1.webp`, `-2.webp`, etc. under `templates/` are **workflow preview thumbnails only**. +- Use them **only** in the preview `` above (immediately after the `###` heading, before ``). +- **Never** use any `templates/*.webp` in **Input materials** or **Example output** sections. Those sections must use `input/{file}` and `output/{file}` from `index.json` only. + + + + Open in Comfy Cloud + + + Download JSON or search "{title}" in Template Library + + +``` + +- **`page-stem`**: MDX filename without extension (e.g. `z-image-turbo` for `z-image-turbo.mdx`). +- **UTM**: always `utm_source=docs&utm_medium=referral&utm_campaign={page-stem}`. Replace `inhouse_social`, `boogu_image_launch`, or other legacy campaign values. +- **CardGroup**: always `cols={2}` for Cloud + Download pairs. Never use `cols={1}` for these. +- **CardGroup lines must NOT start with `|`** — that renders as broken markdown tables. + +## Input materials (when `io.inputs` exists) + +```mdx +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node {nodeId} · `{filename}` + + + +
+ {filename} +
+``` + +Use `repeat(2, ...)` even for a single image. For multiple inputs, one card and one `` per input in the grid. + +## Example output + +**Text-to-image** (only `io.outputs`, no inputs): single output image from `output/{file}`. + +```mdx +**Example output** + +![{alt}](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/{file}) +``` + +**Edit / compare** (`thumbnailVariant: compareSlider` or `thumbnail` array with input + output): + +```mdx +**Example output** + +
+ Input image + {title} example output +
+``` + +**URL verification**: run `curl -s -o /dev/null -w "%{http_code}" -L {url}` before embedding. If `output/{file}` 404s, **omit the Example output section** entirely. Do not substitute `templates/*-N.webp` preview thumbnails. + +## Model download cards + +- Wrap in `` when there are multiple model cards. +- Hugging Face links: use `https://huggingface.co/{org}/{repo}/blob/main/...` — **not** `/resolve/main`. +- Do not change model filenames or storage paths unless fixing an obvious broken link. + +## Reference pages (already standardized — copy patterns from these) + +| Page | Notes | +|------|-------| +| `tutorials/flux/flux-2-dev.mdx` | Multiple workflows, inputs + compare output | +| `tutorials/flux/flux-2-klein.mdx` | Six workflow subsections | +| `tutorials/image/qwen/qwen-image-edit-2511.mdx` | Two inputs, compare output | +| `tutorials/image/qwen/qwen-image-layered.mdx` | Single input | +| `tutorials/image/z-image/z-image.mdx` | T2I output, CardGroup model downloads | +| `tutorials/image/boogu/boogu-image-0.1.mdx` | Turbo + Edit, UTM `boogu-image-0-1` | + +## i18n (zh / ja / ko) + +For every English file updated, apply the same structural changes to: + +- `zh/{same-path}` +- `ja/{same-path}` +- `ko/{same-path}` + +Translate user-facing labels only (`Run on Comfy Cloud`, `Input materials`, `Example output`, card titles, alt text). Keep URLs, template names, node IDs, and filenames identical. + +Preserve existing frontmatter (`translationSourceHash`, `translationBlockHashes`, etc.). Do not remove `translationFrom`. + +Label translations: +| EN | zh | ja | ko | +|----|----|----|-----| +| Run on Comfy Cloud | 在 Comfy Cloud 上运行 | Comfy Cloud で実行 | Comfy Cloud에서 실행 | +| Download Workflow | 下载工作流 | ワークフローをダウンロード | 워크플로 다운로드 | +| Input materials | 输入素材 | 入力素材 | 입력 자료 | +| Example output | 输出示例 | 出力例 | 출력 예시 | + +## Prose style + +- Avoid em dashes (—). Use periods, commas, or colons. +- Match technical reference tone of surrounding pages. + +## Scope rules + +- **Only edit workflow-related sections** — do not rewrite intros, model storage trees, or unrelated content unless fixing HF `resolve` → `blob` links in model download cards on the same page. +- **Skip pages already passing audit** (have `-1.webp` preview, `CardGroup cols={2}`, correct UTM, no `resolve/main` in HF links on page). +- **Do not commit** unless explicitly asked. + +## Already completed (skip unless still failing audit) + +- `tutorials/flux/flux-2-dev.mdx` (+ zh/ja/ko partial — sync i18n if EN-only) +- `tutorials/flux/flux-2-klein.mdx` +- `tutorials/image/qwen/qwen-image-2512.mdx`, `qwen-image-edit-2511.mdx`, `qwen-image-layered.mdx` (+ zh/ja/ko for layered) +- `tutorials/image/z-image/z-image.mdx` (+ zh/ja/ko) +- `tutorials/image/boogu/boogu-image-0.1.mdx` (+ zh/ja/ko) diff --git a/.cursor/skills/tutorial-workflow-standard/fix_webp_misuse.py b/.cursor/skills/tutorial-workflow-standard/fix_webp_misuse.py new file mode 100644 index 000000000..add2633d2 --- /dev/null +++ b/.cursor/skills/tutorial-workflow-standard/fix_webp_misuse.py @@ -0,0 +1,140 @@ +#!/usr/bin/env python3 +"""Remove templates/*.webp from Example output sections; fix SPEC violations.""" + +import json +import re +from pathlib import Path + +ROOT = Path(__file__).resolve().parents[3] +INDEX_PATH = Path("/Users/linmoumou/Documents/comfy/workflow_templates/templates/index.json") + +OUTPUT_MARKERS = [ + "**Example output**", + "**输出示例**", + "**出力例**", + "**출력 예시**", +] + +WEBP_IN_TEMPLATES = re.compile( + r"https://raw\.githubusercontent\.com/Comfy-Org/workflow_templates/main/templates/[^\s\"')]+-\d+\.webp" +) + + +def load_templates() -> dict: + data = json.loads(INDEX_PATH.read_text()) + out = {} + for mod in data: + for t in mod.get("templates", []): + out[t["name"]] = t + return out + + +def output_url_for_template(templates: dict, template_name: str) -> str | None: + t = templates.get(template_name) + if not t: + return None + thumb = t.get("thumbnail") or [] + for item in thumb: + if item.startswith("output/"): + return f"https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/{item}" + for o in t.get("io", {}).get("outputs", []): + f = o.get("file") + if f: + return f"https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/{f}" + return None + + +def find_example_output_blocks(text: str) -> list[tuple[int, int]]: + blocks = [] + for marker in OUTPUT_MARKERS: + start = 0 + while True: + idx = text.find(marker, start) + if idx == -1: + break + block_start = idx + rest = text[idx + len(marker) :] + # end at next ## or ### heading at line start, or **Model / next major section + end_match = re.search(r"\n(?:## |### |\*\*[A-Z])", rest) + block_end = idx + len(marker) + (end_match.start() if end_match else len(rest)) + blocks.append((block_start, block_end)) + start = idx + len(marker) + return sorted(set(blocks)) + + +def fix_file(path: Path, templates: dict) -> bool: + text = path.read_text(encoding="utf-8") + original = text + + blocks = find_example_output_blocks(text) + # process from end to preserve offsets + for start, end in reversed(blocks): + block = text[start:end] + if not WEBP_IN_TEMPLATES.search(block): + continue + # try replace webp output side with real output URL + webp_match = WEBP_IN_TEMPLATES.search(block) + if webp_match: + webp_url = webp_match.group(0) + m = re.search(r"templates/([^/\"')]+)-\d+\.webp", webp_url) + template_name = m.group(1) if m else None + real_url = output_url_for_template(templates, template_name) if template_name else None + if real_url: + block = block.replace(webp_url, real_url) + text = text[:start] + block + text[end:] + else: + # remove entire example output section + text = text[:start] + text[end:] + # trim extra blank lines + text = re.sub(r"\n{3,}", "\n\n", text) + + # safety: remove any remaining templates/*-2.webp anywhere (never valid in docs) + text = WEBP_IN_TEMPLATES.sub( + lambda m: "" if "-2.webp" in m.group(0) else m.group(0), + text, + ) + # remove broken empty output imgs left after stripping webp + text = re.sub(r'\n\s*]*/>\s*', "\n", text) + # clean up empty grids left behind + text = re.sub( + r"\*\*Example output\*\*\s*\n\s*
]*\}\}>\s*
\s*\n", + "", + text, + ) + text = re.sub( + r"\*\*输出示例\*\*\s*\n\s*
]*\}\}>\s*
\s*\n", + "", + text, + ) + text = re.sub( + r"\*\*出力例\*\*\s*\n\s*
]*\}\}>\s*
\s*\n", + "", + text, + ) + text = re.sub( + r"\*\*출력 예시\*\*\s*\n\s*
]*\}\}>\s*
\s*\n", + "", + text, + ) + + if text != original: + path.write_text(text, encoding="utf-8") + return True + return False + + +def main() -> None: + templates = load_templates() + changed = [] + for path in ROOT.rglob("*.mdx"): + if "node_modules" in path.parts: + continue + if fix_file(path, templates): + changed.append(str(path.relative_to(ROOT))) + print(f"Fixed {len(changed)} files") + for p in sorted(changed): + print(f" {p}") + + +if __name__ == "__main__": + main() diff --git a/.cursor/skills/tutorial-workflow-standard/sync_i18n_workflows.py b/.cursor/skills/tutorial-workflow-standard/sync_i18n_workflows.py new file mode 100644 index 000000000..6953bbd33 --- /dev/null +++ b/.cursor/skills/tutorial-workflow-standard/sync_i18n_workflows.py @@ -0,0 +1,161 @@ +#!/usr/bin/env python3 +"""Sync workflow-standard blocks from EN MDX to zh/ja/ko.""" + +import re +from pathlib import Path + +ROOT = Path("/Users/linmoumou/Documents/comfy/docs") + +LOCALES = { + "zh": { + "Run on Comfy Cloud": "在 Comfy Cloud 上运行", + "Download Workflow": "下载工作流", + "Input materials": "输入素材", + "Example output": "输出示例", + "Open in Comfy Cloud": "在 Comfy Cloud 上打开", + }, + "ja": { + "Run on Comfy Cloud": "Comfy Cloud で実行", + "Download Workflow": "ワークフローをダウンロード", + "Input materials": "入力素材", + "Example output": "出力例", + "Open in Comfy Cloud": "Comfy Cloud で開く", + }, + "ko": { + "Run on Comfy Cloud": "Comfy Cloud에서 실행", + "Download Workflow": "워크플로 다운로드", + "Input materials": "입력 자료", + "Example output": "출력 예시", + "Open in Comfy Cloud": "Comfy Cloud에서 열기", + }, +} + +FILES = { + "tutorials/utility/depth-anything-3.mdx": "## Example Workflows", + "tutorials/utility/face-detection/mediapipe.mdx": "## MediaPipe Face Detection Workflow", + "tutorials/utility/pose-detection-sdpose.mdx": "## SDPose Workflows", + "tutorials/utility/preprocessors.mdx": "## Depth estimation", + "tutorials/utility/remove-background-birefnet.mdx": "## BiRefNet Background Removal Workflow", + "tutorials/utility/video-segment-sam3.mdx": "## SAM 3.1 Segment Workflows", + "tutorials/utility/void-video-inpainting.mdx": "## VOID Video Inpainting Workflow", + "tutorials/utility/moge.mdx": "## Example Workflows", + "tutorials/utility/seedvr2.mdx": "## Example Workflows", + "tutorials/audio/ace-step/ace-step-v1-5.mdx": "## Option 1:", + "tutorials/audio/stable-audio/stable-audio-1.mdx": "## Workflow", + "tutorials/audio/stable-audio/stable-audio-3.mdx": "## Available workflows", + "tutorials/llm/gemma4/gemma4.mdx": "## Available workflow", + "tutorials/llm/qwen/qwen3.mdx": "## Available workflow", + "tutorials/llm/qwen/qwen3_5.mdx": "## Available workflow", + "tutorials/3d/hunyuan3D-2.mdx": "## ComfyUI Hunyuan3D-2mv Workflow Example", + "tutorials/3d/triposplat.mdx": "### TripoSplat:", +} + +LOC_START = { + "tutorials/utility/depth-anything-3.mdx": [ + "## 示例工作流", "## Example Workflows", + "## サンプルワークフロー", "## 예제 워크플로우", + ], + "tutorials/utility/face-detection/mediapipe.mdx": [ + "## MediaPipe 人脸检测工作流", "## MediaPipe Face Detection Workflow", "## MediaPipe", + ], + "tutorials/utility/pose-detection-sdpose.mdx": [ + "## SDPose 工作流", "## SDPose Workflows", "## SDPose ワークフロー", "## SDPose 워크플로", + ], + "tutorials/utility/preprocessors.mdx": [ + "## 深度估计", "## Depth estimation", "## 深度推定", "## 깊이 추정", + ], + "tutorials/utility/remove-background-birefnet.mdx": [ + "## BiRefNet 背景移除工作流", "## BiRefNet Background Removal Workflow", + "## BiRefNet 背景除去ワークフロー", "## BiRefNet 배경 제거 워크플로우", + ], + "tutorials/utility/video-segment-sam3.mdx": [ + "## SAM 3.1 分割工作流", "## SAM 3.1 Segment Workflows", + "## SAM 3.1 セグメンテーションワークフロー", "## SAM 3.1 분할 워크플로우", + ], + "tutorials/utility/void-video-inpainting.mdx": [ + "## VOID 视频修复工作流", "## VOID Video Inpainting Workflow", + "## VOID ビデオ修復ワークフロー", "## VOID 비디오 인페인팅 워크플로우", + ], + "tutorials/utility/moge.mdx": [ + "## 示例工作流", "## Example Workflows", "## 工作流示例", "## ワークフロー例", "## 예제 워크플로우", + ], + "tutorials/utility/seedvr2.mdx": [ + "## 示例工作流", "## Example Workflows", + "## 1. Image Upscale", "## 1. 图像缩放", "## 1. 画像のアップスケール", "## 1. 이미지 업스케일", + ], + "tutorials/audio/ace-step/ace-step-v1-5.mdx": [ + "## 选项 1", "## Option 1", "## オプション1", "## 옵션 1", + ], + "tutorials/audio/stable-audio/stable-audio-1.mdx": [ + "## 工作流", "## Workflow", "## ワークフロー", "## 워크플로우", + ], + "tutorials/audio/stable-audio/stable-audio-3.mdx": [ + "## 可用工作流", "## Available workflows", "## 利用可能なワークフロー", + "## 이용 가능한 워크플로우", "## 사용 가능한 워크플로", + ], + "tutorials/llm/gemma4/gemma4.mdx": [ + "## 可用工作流", "## Available workflow", "## 利用可能なワークフロー", + "## 이용 가능한 워크플로", "## 사용 가능한 워크플로", + ], + "tutorials/llm/qwen/qwen3.mdx": [ + "## 可用工作流", "## Available workflow", "## 利用可能なワークフロー", + "## 이용 가능한 워크플로", "## 사용 가능한 워크플로", + ], + "tutorials/llm/qwen/qwen3_5.mdx": [ + "## 可用工作流", "## Available workflow", "## 利用可能なワークフロー", + "## 이용 가능한 워크플로", "## 사용 가능한 워크플로", + ], + "tutorials/3d/hunyuan3D-2.mdx": [ + "## ComfyUI Hunyuan3D-2mv 工作流示例", "## ComfyUI Hunyuan3D-2mv Workflow Example", + "## ComfyUI Hunyuan3D-2mv ワークフロー例", "## ComfyUI Hunyuan3D-2mv 워크플로우 예시", + ], + "tutorials/3d/triposplat.mdx": ["### TripoSplat", ""], +} + + +def translate_labels(text: str, locale: str) -> str: + for en, loc in LOCALES[locale].items(): + text = text.replace(f'title="{en}"', f'title="{loc}"') + text = text.replace(f"**{en}**", f"**{loc}**") + return text + + +def main() -> None: + for rel, en_marker in FILES.items(): + en_path = ROOT / rel + en_text = en_path.read_text() + start = en_text.find(en_marker) + if start == -1: + print(f"NO EN MARKER: {rel}") + continue + workflow_block = en_text[start:] + + for locale in LOCALES: + loc_path = ROOT / locale / rel + if not loc_path.exists(): + print(f"SKIP missing {locale}/{rel}") + continue + loc_text = loc_path.read_text() + fm_match = re.match(r"(---\n.*?\n---\n)", loc_text, re.S) + if not fm_match: + print(f"NO FM {locale}/{rel}") + continue + fm = fm_match.group(1) + body = loc_text[len(fm) :] + loc_start = None + for marker in LOC_START.get(rel, [en_marker]): + idx = body.find(marker) + if idx != -1: + loc_start = idx + break + if loc_start is None: + print(f"NO LOC START {locale}/{rel}") + continue + prefix = body[:loc_start] + translated = translate_labels(workflow_block, locale) + loc_path.write_text(fm + prefix + translated) + print(f"UPDATED {locale}/{rel}") + + +if __name__ == "__main__": + main() diff --git a/docs.json b/docs.json index 0264535b9..4f49085e9 100644 --- a/docs.json +++ b/docs.json @@ -206,6 +206,18 @@ "tutorials/image/z-image/z-image-turbo" ] }, + { + "group": "Krea 2", + "pages": [ + "tutorials/image/krea/krea-2" + ] + }, + { + "group": "Ideogram", + "pages": [ + "tutorials/image/ideogram/ideogram-v4" + ] + }, { "group": "Boogu", "pages": [ @@ -256,18 +268,6 @@ "tutorials/image/pixeldit/pixeldit" ] }, - { - "group": "Krea 2", - "pages": [ - "tutorials/image/krea/krea-2" - ] - }, - { - "group": "Ideogram", - "pages": [ - "tutorials/image/ideogram/ideogram-v4" - ] - }, "tutorials/image/cosmos/cosmos-predict2-t2i", "tutorials/image/omnigen/omnigen2" ] @@ -293,9 +293,9 @@ { "group": "LTX", "pages": [ - "tutorials/video/ltxv", + "tutorials/video/ltx/ltx-2-3", "tutorials/video/ltx/ltx-2", - "tutorials/video/ltx/ltx-2-3" + "tutorials/video/ltxv" ] }, { @@ -416,170 +416,230 @@ "tutorials/partner-nodes/pricing", "tutorials/partner-nodes/concurrency-limits", { - "group": "Black Forest Labs", - "pages": [ - "tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", - "tutorials/partner-nodes/black-forest-labs/flux-1-kontext" - ] - }, - { - "group": "Beeble", - "pages": [ - "tutorials/partner-nodes/beeble/beeble-switchx" - ] - }, - { - "group": "ByteDance", - "pages": [ - "tutorials/partner-nodes/bytedance/seedance-2-0", - "tutorials/partner-nodes/bytedance/seedance-2-0-real-human", - "tutorials/partner-nodes/bytedance/seedream-5-pro", - "tutorials/partner-nodes/bytedance/seedream-5-lite", - "tutorials/partner-nodes/bytedance/seed-audio-1-0" - ] - }, - { - "group": "Google", - "pages": [ - "tutorials/partner-nodes/google/gemini", - "tutorials/partner-nodes/google/gemini-omni-flash", - "tutorials/partner-nodes/google/nano-banana-pro", - "tutorials/partner-nodes/google/nano-banana-2", - "tutorials/partner-nodes/google/nano-banana-2-lite" - ] - }, - { - "group": "Anthropic", - "pages": [ - "tutorials/partner-nodes/anthropic/claude" - ] - }, - { - "group": "Ideogram", - "pages": [ - "tutorials/partner-nodes/ideogram/ideogram-v4" - ] - }, - { - "group": "Luma", - "pages": [ - "tutorials/partner-nodes/luma/luma-uni-1", - "tutorials/partner-nodes/luma/luma-text-to-image", - "tutorials/partner-nodes/luma/luma-image-to-image", - "tutorials/partner-nodes/luma/luma-text-to-video", - "tutorials/partner-nodes/luma/luma-image-to-video" - ] - }, - { - "group": "Moonvalley", - "pages": [ - "tutorials/partner-nodes/moonvalley/moonvalley-video-generation" - ] - }, - { - "group": "OpenAI", - "pages": [ - "tutorials/partner-nodes/openai/gpt-image-2", - "tutorials/partner-nodes/openai/gpt-image-1", - "tutorials/partner-nodes/openai/dall-e-2", - "tutorials/partner-nodes/openai/dall-e-3", - "tutorials/partner-nodes/openai/chat" - ] - }, - { - "group": "OpenRouter", - "pages": [ - "tutorials/partner-nodes/openrouter/llm" - ] - }, - { - "group": "Recraft", - "pages": [ - "tutorials/partner-nodes/recraft/recraft-v4", - "tutorials/partner-nodes/recraft/recraft-text-to-image" - ] - }, - { - "group": "Krea 2", - "pages": [ - "tutorials/partner-nodes/krea2/krea2-t2i" - ] - }, - { - "group": "Kling", - "pages": [ - "tutorials/partner-nodes/kling/kling-3-0", - "tutorials/partner-nodes/kling/kling-motion-control" - ] - }, - { - "group": "Runway", - "pages": [ - "tutorials/partner-nodes/runway/image-generation", - "tutorials/partner-nodes/runway/video-generation" - ] - }, - { - "group": "Rodin", - "pages": [ - "tutorials/partner-nodes/rodin/model-generation" - ] - }, - { - "group": "Tripo", - "pages": [ - "tutorials/partner-nodes/tripo/model-generation", - "tutorials/partner-nodes/tripo/tripo-3-1" - ] - }, - { - "group": "Hunyuan 3D", - "pages": [ - "tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" - ] - }, - { - "group": "Meshy", - "pages": [ - "tutorials/partner-nodes/meshy/meshy-6" - ] - }, - { - "group": "Bria", - "pages": [ - "tutorials/partner-nodes/bria/fibo", - "tutorials/partner-nodes/bria/background-removal" - ] - }, - { - "group": "Reve", + "group": "Image", "pages": [ - "tutorials/partner-nodes/reve/reve-image" + { + "group": "ByteDance", + "pages": [ + "tutorials/partner-nodes/bytedance/seedream-5-pro", + "tutorials/partner-nodes/bytedance/seedream-5-lite" + ] + }, + { + "group": "Google", + "pages": [ + "tutorials/partner-nodes/google/nano-banana-pro", + "tutorials/partner-nodes/google/nano-banana-2", + "tutorials/partner-nodes/google/nano-banana-2-lite" + ] + }, + { + "group": "Black Forest Labs", + "pages": [ + "tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", + "tutorials/partner-nodes/black-forest-labs/flux-1-kontext" + ] + }, + { + "group": "Ideogram", + "pages": [ + "tutorials/partner-nodes/ideogram/ideogram-v4" + ] + }, + { + "group": "Krea 2", + "pages": [ + "tutorials/partner-nodes/krea2/krea2-t2i" + ] + }, + { + "group": "Recraft", + "pages": [ + "tutorials/partner-nodes/recraft/recraft-v4", + "tutorials/partner-nodes/recraft/recraft-text-to-image" + ] + }, + { + "group": "Bria", + "pages": [ + "tutorials/partner-nodes/bria/fibo", + "tutorials/partner-nodes/bria/background-removal" + ] + }, + { + "group": "Reve", + "pages": [ + "tutorials/partner-nodes/reve/reve-image" + ] + }, + { + "group": "Luma", + "pages": [ + "tutorials/partner-nodes/luma/luma-uni-1", + "tutorials/partner-nodes/luma/luma-text-to-image", + "tutorials/partner-nodes/luma/luma-image-to-image" + ] + }, + { + "group": "OpenAI", + "pages": [ + "tutorials/partner-nodes/openai/gpt-image-2", + "tutorials/partner-nodes/openai/gpt-image-1", + "tutorials/partner-nodes/openai/dall-e-2", + "tutorials/partner-nodes/openai/dall-e-3" + ] + }, + { + "group": "Runway", + "pages": [ + "tutorials/partner-nodes/runway/image-generation" + ] + }, + { + "group": "Beeble", + "pages": [ + "tutorials/partner-nodes/beeble/beeble-switchx" + ] + } ] }, { - "group": "Wan", + "group": "Video", "pages": [ - "tutorials/partner-nodes/wan/wan2-7" + { + "group": "ByteDance", + "pages": [ + "tutorials/partner-nodes/bytedance/seedance-2-0", + "tutorials/partner-nodes/bytedance/seedance-2-0-real-human" + ] + }, + { + "group": "Google", + "pages": [ + "tutorials/partner-nodes/google/gemini-omni-flash" + ] + }, + { + "group": "Kling", + "pages": [ + "tutorials/partner-nodes/kling/kling-3-0", + "tutorials/partner-nodes/kling/kling-motion-control" + ] + }, + { + "group": "Luma", + "pages": [ + "tutorials/partner-nodes/luma/luma-text-to-video", + "tutorials/partner-nodes/luma/luma-image-to-video" + ] + }, + { + "group": "Moonvalley", + "pages": [ + "tutorials/partner-nodes/moonvalley/moonvalley-video-generation" + ] + }, + { + "group": "Runway", + "pages": [ + "tutorials/partner-nodes/runway/video-generation" + ] + }, + { + "group": "Wan", + "pages": [ + "tutorials/partner-nodes/wan/wan2-7" + ] + }, + { + "group": "HappyHorse", + "pages": [ + "tutorials/partner-nodes/happyhorse/happyhorse1-1", + "tutorials/partner-nodes/happyhorse/happyhorse1-0" + ] + }, + { + "group": "Topaz", + "pages": [ + "tutorials/partner-nodes/topaz/astra-2" + ] + } ] }, { - "group": "HappyHorse", + "group": "3D", "pages": [ - "tutorials/partner-nodes/happyhorse/happyhorse1-1", - "tutorials/partner-nodes/happyhorse/happyhorse1-0" + { + "group": "Rodin", + "pages": [ + "tutorials/partner-nodes/rodin/model-generation" + ] + }, + { + "group": "Tripo", + "pages": [ + "tutorials/partner-nodes/tripo/model-generation", + "tutorials/partner-nodes/tripo/tripo-3-1" + ] + }, + { + "group": "Hunyuan 3D", + "pages": [ + "tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" + ] + }, + { + "group": "Meshy", + "pages": [ + "tutorials/partner-nodes/meshy/meshy-6" + ] + } ] }, { - "group": "Sonilo", + "group": "Audio", "pages": [ - "tutorials/partner-nodes/sonilo/video-to-music" + { + "group": "ByteDance", + "pages": [ + "tutorials/partner-nodes/bytedance/seed-audio-1-0" + ] + }, + { + "group": "Sonilo", + "pages": [ + "tutorials/partner-nodes/sonilo/video-to-music" + ] + } ] }, { - "group": "Topaz", + "group": "LLM", "pages": [ - "tutorials/partner-nodes/topaz/astra-2" + { + "group": "Google", + "pages": [ + "tutorials/partner-nodes/google/gemini" + ] + }, + { + "group": "Anthropic", + "pages": [ + "tutorials/partner-nodes/anthropic/claude" + ] + }, + { + "group": "OpenAI", + "pages": [ + "tutorials/partner-nodes/openai/chat" + ] + }, + { + "group": "OpenRouter", + "pages": [ + "tutorials/partner-nodes/openrouter/llm" + ] + } ] } ] @@ -2972,6 +3032,18 @@ "zh/tutorials/image/z-image/z-image-turbo" ] }, + { + "group": "Krea 2", + "pages": [ + "zh/tutorials/image/krea/krea-2" + ] + }, + { + "group": "Ideogram", + "pages": [ + "zh/tutorials/image/ideogram/ideogram-v4" + ] + }, { "group": "Boogu", "pages": [ @@ -3022,18 +3094,6 @@ "zh/tutorials/image/pixeldit/pixeldit" ] }, - { - "group": "Krea 2", - "pages": [ - "zh/tutorials/image/krea/krea-2" - ] - }, - { - "group": "Ideogram", - "pages": [ - "zh/tutorials/image/ideogram/ideogram-v4" - ] - }, "zh/tutorials/image/cosmos/cosmos-predict2-t2i", "zh/tutorials/image/omnigen/omnigen2" ] @@ -3059,9 +3119,9 @@ { "group": "LTX", "pages": [ - "zh/tutorials/video/ltxv", + "zh/tutorials/video/ltx/ltx-2-3", "zh/tutorials/video/ltx/ltx-2", - "zh/tutorials/video/ltx/ltx-2-3" + "zh/tutorials/video/ltxv" ] }, { @@ -3182,170 +3242,230 @@ "zh/tutorials/partner-nodes/pricing", "zh/tutorials/partner-nodes/concurrency-limits", { - "group": "Black Forest Labs", - "pages": [ - "zh/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", - "zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext" - ] - }, - { - "group": "Beeble", - "pages": [ - "zh/tutorials/partner-nodes/beeble/beeble-switchx" - ] - }, - { - "group": "ByteDance", - "pages": [ - "zh/tutorials/partner-nodes/bytedance/seedance-2-0", - "zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human", - "zh/tutorials/partner-nodes/bytedance/seedream-5-pro", - "zh/tutorials/partner-nodes/bytedance/seedream-5-lite", - "zh/tutorials/partner-nodes/bytedance/seed-audio-1-0" - ] - }, - { - "group": "Google", - "pages": [ - "zh/tutorials/partner-nodes/google/gemini", - "zh/tutorials/partner-nodes/google/gemini-omni-flash", - "zh/tutorials/partner-nodes/google/nano-banana-pro", - "zh/tutorials/partner-nodes/google/nano-banana-2", - "zh/tutorials/partner-nodes/google/nano-banana-2-lite" - ] - }, - { - "group": "Anthropic", - "pages": [ - "zh/tutorials/partner-nodes/anthropic/claude" - ] - }, - { - "group": "Ideogram", - "pages": [ - "zh/tutorials/partner-nodes/ideogram/ideogram-v4" - ] - }, - { - "group": "Luma", - "pages": [ - "zh/tutorials/partner-nodes/luma/luma-uni-1", - "zh/tutorials/partner-nodes/luma/luma-text-to-image", - "zh/tutorials/partner-nodes/luma/luma-image-to-image", - "zh/tutorials/partner-nodes/luma/luma-text-to-video", - "zh/tutorials/partner-nodes/luma/luma-image-to-video" - ] - }, - { - "group": "Moonvalley", - "pages": [ - "zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation" - ] - }, - { - "group": "OpenAI", - "pages": [ - "zh/tutorials/partner-nodes/openai/gpt-image-2", - "zh/tutorials/partner-nodes/openai/gpt-image-1", - "zh/tutorials/partner-nodes/openai/dall-e-2", - "zh/tutorials/partner-nodes/openai/dall-e-3", - "zh/tutorials/partner-nodes/openai/chat" - ] - }, - { - "group": "OpenRouter", - "pages": [ - "zh/tutorials/partner-nodes/openrouter/llm" - ] - }, - { - "group": "Recraft", - "pages": [ - "zh/tutorials/partner-nodes/recraft/recraft-v4", - "zh/tutorials/partner-nodes/recraft/recraft-text-to-image" - ] - }, - { - "group": "Krea 2", - "pages": [ - "zh/tutorials/partner-nodes/krea2/krea2-t2i" - ] - }, - { - "group": "Kling", - "pages": [ - "zh/tutorials/partner-nodes/kling/kling-3-0", - "zh/tutorials/partner-nodes/kling/kling-motion-control" - ] - }, - { - "group": "Runway", - "pages": [ - "zh/tutorials/partner-nodes/runway/image-generation", - "zh/tutorials/partner-nodes/runway/video-generation" - ] - }, - { - "group": "Rodin", - "pages": [ - "zh/tutorials/partner-nodes/rodin/model-generation" - ] - }, - { - "group": "Tripo", - "pages": [ - "zh/tutorials/partner-nodes/tripo/model-generation", - "zh/tutorials/partner-nodes/tripo/tripo-3-1" - ] - }, - { - "group": "Hunyuan 3D", - "pages": [ - "zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" - ] - }, - { - "group": "Meshy", - "pages": [ - "zh/tutorials/partner-nodes/meshy/meshy-6" - ] - }, - { - "group": "Bria", - "pages": [ - "zh/tutorials/partner-nodes/bria/fibo", - "zh/tutorials/partner-nodes/bria/background-removal" - ] - }, - { - "group": "Reve", + "group": "图像", "pages": [ - "zh/tutorials/partner-nodes/reve/reve-image" + { + "group": "ByteDance", + "pages": [ + "zh/tutorials/partner-nodes/bytedance/seedream-5-pro", + "zh/tutorials/partner-nodes/bytedance/seedream-5-lite" + ] + }, + { + "group": "Google", + "pages": [ + "zh/tutorials/partner-nodes/google/nano-banana-pro", + "zh/tutorials/partner-nodes/google/nano-banana-2", + "zh/tutorials/partner-nodes/google/nano-banana-2-lite" + ] + }, + { + "group": "Black Forest Labs", + "pages": [ + "zh/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", + "zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext" + ] + }, + { + "group": "Ideogram", + "pages": [ + "zh/tutorials/partner-nodes/ideogram/ideogram-v4" + ] + }, + { + "group": "Krea 2", + "pages": [ + "zh/tutorials/partner-nodes/krea2/krea2-t2i" + ] + }, + { + "group": "Recraft", + "pages": [ + "zh/tutorials/partner-nodes/recraft/recraft-v4", + "zh/tutorials/partner-nodes/recraft/recraft-text-to-image" + ] + }, + { + "group": "Bria", + "pages": [ + "zh/tutorials/partner-nodes/bria/fibo", + "zh/tutorials/partner-nodes/bria/background-removal" + ] + }, + { + "group": "Reve", + "pages": [ + "zh/tutorials/partner-nodes/reve/reve-image" + ] + }, + { + "group": "Luma", + "pages": [ + "zh/tutorials/partner-nodes/luma/luma-uni-1", + "zh/tutorials/partner-nodes/luma/luma-text-to-image", + "zh/tutorials/partner-nodes/luma/luma-image-to-image" + ] + }, + { + "group": "OpenAI", + "pages": [ + "zh/tutorials/partner-nodes/openai/gpt-image-2", + "zh/tutorials/partner-nodes/openai/gpt-image-1", + "zh/tutorials/partner-nodes/openai/dall-e-2", + "zh/tutorials/partner-nodes/openai/dall-e-3" + ] + }, + { + "group": "Runway", + "pages": [ + "zh/tutorials/partner-nodes/runway/image-generation" + ] + }, + { + "group": "Beeble", + "pages": [ + "zh/tutorials/partner-nodes/beeble/beeble-switchx" + ] + } ] }, { - "group": "Wan", + "group": "视频", "pages": [ - "zh/tutorials/partner-nodes/wan/wan2-7" + { + "group": "ByteDance", + "pages": [ + "zh/tutorials/partner-nodes/bytedance/seedance-2-0", + "zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human" + ] + }, + { + "group": "Google", + "pages": [ + "zh/tutorials/partner-nodes/google/gemini-omni-flash" + ] + }, + { + "group": "Kling", + "pages": [ + "zh/tutorials/partner-nodes/kling/kling-3-0", + "zh/tutorials/partner-nodes/kling/kling-motion-control" + ] + }, + { + "group": "Luma", + "pages": [ + "zh/tutorials/partner-nodes/luma/luma-text-to-video", + "zh/tutorials/partner-nodes/luma/luma-image-to-video" + ] + }, + { + "group": "Moonvalley", + "pages": [ + "zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation" + ] + }, + { + "group": "Runway", + "pages": [ + "zh/tutorials/partner-nodes/runway/video-generation" + ] + }, + { + "group": "Wan", + "pages": [ + "zh/tutorials/partner-nodes/wan/wan2-7" + ] + }, + { + "group": "HappyHorse", + "pages": [ + "zh/tutorials/partner-nodes/happyhorse/happyhorse1-1", + "zh/tutorials/partner-nodes/happyhorse/happyhorse1-0" + ] + }, + { + "group": "Topaz", + "pages": [ + "zh/tutorials/partner-nodes/topaz/astra-2" + ] + } ] }, { - "group": "HappyHorse", + "group": "3D", "pages": [ - "zh/tutorials/partner-nodes/happyhorse/happyhorse1-1", - "zh/tutorials/partner-nodes/happyhorse/happyhorse1-0" + { + "group": "Rodin", + "pages": [ + "zh/tutorials/partner-nodes/rodin/model-generation" + ] + }, + { + "group": "Tripo", + "pages": [ + "zh/tutorials/partner-nodes/tripo/model-generation", + "zh/tutorials/partner-nodes/tripo/tripo-3-1" + ] + }, + { + "group": "Hunyuan 3D", + "pages": [ + "zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" + ] + }, + { + "group": "Meshy", + "pages": [ + "zh/tutorials/partner-nodes/meshy/meshy-6" + ] + } ] }, { - "group": "Sonilo", + "group": "音频", "pages": [ - "zh/tutorials/partner-nodes/sonilo/video-to-music" + { + "group": "ByteDance", + "pages": [ + "zh/tutorials/partner-nodes/bytedance/seed-audio-1-0" + ] + }, + { + "group": "Sonilo", + "pages": [ + "zh/tutorials/partner-nodes/sonilo/video-to-music" + ] + } ] }, { - "group": "Topaz", + "group": "LLM", "pages": [ - "zh/tutorials/partner-nodes/topaz/astra-2" + { + "group": "Google", + "pages": [ + "zh/tutorials/partner-nodes/google/gemini" + ] + }, + { + "group": "Anthropic", + "pages": [ + "zh/tutorials/partner-nodes/anthropic/claude" + ] + }, + { + "group": "OpenAI", + "pages": [ + "zh/tutorials/partner-nodes/openai/chat" + ] + }, + { + "group": "OpenRouter", + "pages": [ + "zh/tutorials/partner-nodes/openrouter/llm" + ] + } ] } ] @@ -5712,6 +5832,18 @@ "ja/tutorials/image/z-image/z-image-turbo" ] }, + { + "group": "Krea 2", + "pages": [ + "ja/tutorials/image/krea/krea-2" + ] + }, + { + "group": "Ideogram", + "pages": [ + "ja/tutorials/image/ideogram/ideogram-v4" + ] + }, { "group": "Boogu", "pages": [ @@ -5762,18 +5894,6 @@ "ja/tutorials/image/pixeldit/pixeldit" ] }, - { - "group": "Krea 2", - "pages": [ - "ja/tutorials/image/krea/krea-2" - ] - }, - { - "group": "Ideogram", - "pages": [ - "ja/tutorials/image/ideogram/ideogram-v4" - ] - }, "ja/tutorials/image/cosmos/cosmos-predict2-t2i", "ja/tutorials/image/omnigen/omnigen2" ] @@ -5799,9 +5919,9 @@ { "group": "LTX", "pages": [ - "ja/tutorials/video/ltxv", + "ja/tutorials/video/ltx/ltx-2-3", "ja/tutorials/video/ltx/ltx-2", - "ja/tutorials/video/ltx/ltx-2-3" + "ja/tutorials/video/ltxv" ] }, { @@ -5923,176 +6043,230 @@ "ja/tutorials/partner-nodes/pricing", "ja/tutorials/partner-nodes/concurrency-limits", { - "group": "Black Forest Labs", - "pages": [ - "ja/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", - "ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext" - ] - }, - { - "group": "Beeble", - "pages": [ - "ja/tutorials/partner-nodes/beeble/beeble-switchx" - ] - }, - { - "group": "ByteDance", - "pages": [ - "ja/tutorials/partner-nodes/bytedance/seedance-2-0", - "ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human", - "ja/tutorials/partner-nodes/bytedance/seedream-5-pro", - "ja/tutorials/partner-nodes/bytedance/seedream-5-lite", - "ja/tutorials/partner-nodes/bytedance/seed-audio-1-0" - ] - }, - { - "group": "Google", - "pages": [ - "ja/tutorials/partner-nodes/google/gemini", - "ja/tutorials/partner-nodes/google/gemini-omni-flash", - "ja/tutorials/partner-nodes/google/nano-banana-pro", - "ja/tutorials/partner-nodes/google/nano-banana-2", - "ja/tutorials/partner-nodes/google/nano-banana-2-lite" - ] - }, - { - "group": "Anthropic", - "pages": [ - "ja/tutorials/partner-nodes/anthropic/claude" - ] - }, - { - "group": "Ideogram", - "pages": [ - "ja/tutorials/partner-nodes/ideogram/ideogram-v4" - ] - }, - { - "group": "Luma", - "pages": [ - "ja/tutorials/partner-nodes/luma/luma-uni-1", - "ja/tutorials/partner-nodes/luma/luma-text-to-image", - "ja/tutorials/partner-nodes/luma/luma-image-to-image", - "ja/tutorials/partner-nodes/luma/luma-text-to-video", - "ja/tutorials/partner-nodes/luma/luma-image-to-video" - ] - }, - { - "group": "Moonvalley", - "pages": [ - "ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation" - ] - }, - { - "group": "OpenAI", - "pages": [ - "ja/tutorials/partner-nodes/openai/gpt-image-2", - "ja/tutorials/partner-nodes/openai/gpt-image-1", - "ja/tutorials/partner-nodes/openai/dall-e-2", - "ja/tutorials/partner-nodes/openai/dall-e-3", - "ja/tutorials/partner-nodes/openai/chat" - ] - }, - { - "group": "OpenRouter", - "pages": [ - "ja/tutorials/partner-nodes/openrouter/llm" - ] - }, - { - "group": "Recraft", - "pages": [ - "ja/tutorials/partner-nodes/recraft/recraft-v4", - "ja/tutorials/partner-nodes/recraft/recraft-text-to-image" - ] - }, - { - "group": "Krea 2", - "pages": [ - "ja/tutorials/partner-nodes/krea2/krea2-t2i" - ] - }, - { - "group": "Kling", - "pages": [ - "ja/tutorials/partner-nodes/kling/kling-3-0", - "ja/tutorials/partner-nodes/kling/kling-motion-control" - ] - }, - { - "group": "Runway", - "pages": [ - "ja/tutorials/partner-nodes/runway/image-generation", - "ja/tutorials/partner-nodes/runway/video-generation" - ] - }, - { - "group": "Rodin", - "pages": [ - "ja/tutorials/partner-nodes/rodin/model-generation" - ] - }, - { - "group": "Tripo", - "pages": [ - "ja/tutorials/partner-nodes/tripo/model-generation", - "ja/tutorials/partner-nodes/tripo/tripo-3-1" - ] - }, - { - "group": "Hunyuan 3D", - "pages": [ - "ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" - ] - }, - { - "group": "Meshy", - "pages": [ - "ja/tutorials/partner-nodes/meshy/meshy-6" - ] - }, - { - "group": "Bria", - "pages": [ - "ja/tutorials/partner-nodes/bria/fibo", - "ja/tutorials/partner-nodes/bria/background-removal" - ] - }, - { - "group": "Reve", - "pages": [ - "ja/tutorials/partner-nodes/reve/reve-image" - ] - }, - { - "group": "Wan", + "group": "画像", "pages": [ - "ja/tutorials/partner-nodes/wan/wan2-7" + { + "group": "ByteDance", + "pages": [ + "ja/tutorials/partner-nodes/bytedance/seedream-5-pro", + "ja/tutorials/partner-nodes/bytedance/seedream-5-lite" + ] + }, + { + "group": "Google", + "pages": [ + "ja/tutorials/partner-nodes/google/nano-banana-pro", + "ja/tutorials/partner-nodes/google/nano-banana-2", + "ja/tutorials/partner-nodes/google/nano-banana-2-lite" + ] + }, + { + "group": "Black Forest Labs", + "pages": [ + "ja/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", + "ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext" + ] + }, + { + "group": "Ideogram", + "pages": [ + "ja/tutorials/partner-nodes/ideogram/ideogram-v4" + ] + }, + { + "group": "Krea 2", + "pages": [ + "ja/tutorials/partner-nodes/krea2/krea2-t2i" + ] + }, + { + "group": "Recraft", + "pages": [ + "ja/tutorials/partner-nodes/recraft/recraft-v4", + "ja/tutorials/partner-nodes/recraft/recraft-text-to-image" + ] + }, + { + "group": "Bria", + "pages": [ + "ja/tutorials/partner-nodes/bria/fibo", + "ja/tutorials/partner-nodes/bria/background-removal" + ] + }, + { + "group": "Reve", + "pages": [ + "ja/tutorials/partner-nodes/reve/reve-image" + ] + }, + { + "group": "Luma", + "pages": [ + "ja/tutorials/partner-nodes/luma/luma-uni-1", + "ja/tutorials/partner-nodes/luma/luma-text-to-image", + "ja/tutorials/partner-nodes/luma/luma-image-to-image" + ] + }, + { + "group": "OpenAI", + "pages": [ + "ja/tutorials/partner-nodes/openai/gpt-image-2", + "ja/tutorials/partner-nodes/openai/gpt-image-1", + "ja/tutorials/partner-nodes/openai/dall-e-2", + "ja/tutorials/partner-nodes/openai/dall-e-3" + ] + }, + { + "group": "Runway", + "pages": [ + "ja/tutorials/partner-nodes/runway/image-generation" + ] + }, + { + "group": "Beeble", + "pages": [ + "ja/tutorials/partner-nodes/beeble/beeble-switchx" + ] + } ] }, { - "group": "HappyHorse", + "group": "ビデオ", "pages": [ - "ja/tutorials/partner-nodes/happyhorse/happyhorse1-1", - "ja/tutorials/partner-nodes/happyhorse/happyhorse1-0" + { + "group": "ByteDance", + "pages": [ + "ja/tutorials/partner-nodes/bytedance/seedance-2-0", + "ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human" + ] + }, + { + "group": "Google", + "pages": [ + "ja/tutorials/partner-nodes/google/gemini-omni-flash" + ] + }, + { + "group": "Kling", + "pages": [ + "ja/tutorials/partner-nodes/kling/kling-3-0", + "ja/tutorials/partner-nodes/kling/kling-motion-control" + ] + }, + { + "group": "Luma", + "pages": [ + "ja/tutorials/partner-nodes/luma/luma-text-to-video", + "ja/tutorials/partner-nodes/luma/luma-image-to-video" + ] + }, + { + "group": "Moonvalley", + "pages": [ + "ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation" + ] + }, + { + "group": "Runway", + "pages": [ + "ja/tutorials/partner-nodes/runway/video-generation" + ] + }, + { + "group": "Wan", + "pages": [ + "ja/tutorials/partner-nodes/wan/wan2-7" + ] + }, + { + "group": "HappyHorse", + "pages": [ + "ja/tutorials/partner-nodes/happyhorse/happyhorse1-1", + "ja/tutorials/partner-nodes/happyhorse/happyhorse1-0" + ] + }, + { + "group": "Topaz", + "pages": [ + "ja/tutorials/partner-nodes/topaz/astra-2" + ] + } ] }, { - "group": "Sonilo", + "group": "3D", "pages": [ - "ja/tutorials/partner-nodes/sonilo/video-to-music" + { + "group": "Rodin", + "pages": [ + "ja/tutorials/partner-nodes/rodin/model-generation" + ] + }, + { + "group": "Tripo", + "pages": [ + "ja/tutorials/partner-nodes/tripo/model-generation", + "ja/tutorials/partner-nodes/tripo/tripo-3-1" + ] + }, + { + "group": "Hunyuan 3D", + "pages": [ + "ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" + ] + }, + { + "group": "Meshy", + "pages": [ + "ja/tutorials/partner-nodes/meshy/meshy-6" + ] + } ] }, { - "group": "Topaz", + "group": "オーディオ", "pages": [ - "ja/tutorials/partner-nodes/topaz/astra-2" + { + "group": "ByteDance", + "pages": [ + "ja/tutorials/partner-nodes/bytedance/seed-audio-1-0" + ] + }, + { + "group": "Sonilo", + "pages": [ + "ja/tutorials/partner-nodes/sonilo/video-to-music" + ] + } ] }, { - "group": "Krea 2", + "group": "LLM", "pages": [ - "ko/tutorials/partner-nodes/krea2/krea2-t2i" + { + "group": "Google", + "pages": [ + "ja/tutorials/partner-nodes/google/gemini" + ] + }, + { + "group": "Anthropic", + "pages": [ + "ja/tutorials/partner-nodes/anthropic/claude" + ] + }, + { + "group": "OpenAI", + "pages": [ + "ja/tutorials/partner-nodes/openai/chat" + ] + }, + { + "group": "OpenRouter", + "pages": [ + "ja/tutorials/partner-nodes/openrouter/llm" + ] + } ] } ] @@ -8537,6 +8711,18 @@ "ko/tutorials/image/z-image/z-image-turbo" ] }, + { + "group": "Krea 2", + "pages": [ + "ko/tutorials/image/krea/krea-2" + ] + }, + { + "group": "Ideogram", + "pages": [ + "ko/tutorials/image/ideogram/ideogram-v4" + ] + }, { "group": "Boogu", "pages": [ @@ -8587,18 +8773,6 @@ "ko/tutorials/image/pixeldit/pixeldit" ] }, - { - "group": "Krea 2", - "pages": [ - "ko/tutorials/image/krea/krea-2" - ] - }, - { - "group": "Ideogram", - "pages": [ - "ko/tutorials/image/ideogram/ideogram-v4" - ] - }, "ko/tutorials/image/cosmos/cosmos-predict2-t2i", "ko/tutorials/image/omnigen/omnigen2" ] @@ -8624,9 +8798,9 @@ { "group": "LTX", "pages": [ - "ko/tutorials/video/ltxv", + "ko/tutorials/video/ltx/ltx-2-3", "ko/tutorials/video/ltx/ltx-2", - "ko/tutorials/video/ltx/ltx-2-3" + "ko/tutorials/video/ltxv" ] }, { @@ -8746,164 +8920,230 @@ "ko/tutorials/partner-nodes/pricing", "ko/tutorials/partner-nodes/concurrency-limits", { - "group": "Black Forest Labs", - "pages": [ - "ko/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", - "ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext" - ] - }, - { - "group": "Beeble", - "pages": [ - "ko/tutorials/partner-nodes/beeble/beeble-switchx" - ] - }, - { - "group": "ByteDance", - "pages": [ - "ko/tutorials/partner-nodes/bytedance/seedance-2-0", - "ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human", - "ko/tutorials/partner-nodes/bytedance/seedream-5-pro", - "ko/tutorials/partner-nodes/bytedance/seedream-5-lite", - "ko/tutorials/partner-nodes/bytedance/seed-audio-1-0" - ] - }, - { - "group": "Google", - "pages": [ - "ko/tutorials/partner-nodes/google/gemini", - "ko/tutorials/partner-nodes/google/gemini-omni-flash", - "ko/tutorials/partner-nodes/google/nano-banana-pro", - "ko/tutorials/partner-nodes/google/nano-banana-2", - "ko/tutorials/partner-nodes/google/nano-banana-2-lite" - ] - }, - { - "group": "Anthropic", - "pages": [ - "ko/tutorials/partner-nodes/anthropic/claude" - ] - }, - { - "group": "Ideogram", - "pages": [ - "ko/tutorials/partner-nodes/ideogram/ideogram-v4" - ] - }, - { - "group": "Luma", - "pages": [ - "ko/tutorials/partner-nodes/luma/luma-uni-1", - "ko/tutorials/partner-nodes/luma/luma-text-to-image", - "ko/tutorials/partner-nodes/luma/luma-image-to-image", - "ko/tutorials/partner-nodes/luma/luma-text-to-video", - "ko/tutorials/partner-nodes/luma/luma-image-to-video" - ] - }, - { - "group": "Moonvalley", - "pages": [ - "ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation" - ] - }, - { - "group": "OpenAI", - "pages": [ - "ko/tutorials/partner-nodes/openai/gpt-image-2", - "ko/tutorials/partner-nodes/openai/gpt-image-1", - "ko/tutorials/partner-nodes/openai/dall-e-2", - "ko/tutorials/partner-nodes/openai/dall-e-3", - "ko/tutorials/partner-nodes/openai/chat" - ] - }, - { - "group": "OpenRouter", - "pages": [ - "ko/tutorials/partner-nodes/openrouter/llm" - ] - }, - { - "group": "Recraft", - "pages": [ - "ko/tutorials/partner-nodes/recraft/recraft-v4", - "ko/tutorials/partner-nodes/recraft/recraft-text-to-image" - ] - }, - { - "group": "Kling", - "pages": [ - "ko/tutorials/partner-nodes/kling/kling-3-0", - "ko/tutorials/partner-nodes/kling/kling-motion-control" - ] - }, - { - "group": "Runway", - "pages": [ - "ko/tutorials/partner-nodes/runway/image-generation", - "ko/tutorials/partner-nodes/runway/video-generation" - ] - }, - { - "group": "Rodin", - "pages": [ - "ko/tutorials/partner-nodes/rodin/model-generation" - ] - }, - { - "group": "Tripo", - "pages": [ - "ko/tutorials/partner-nodes/tripo/model-generation", - "ko/tutorials/partner-nodes/tripo/tripo-3-1" - ] - }, - { - "group": "Hunyuan 3D", - "pages": [ - "ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" - ] - }, - { - "group": "Meshy", - "pages": [ - "ko/tutorials/partner-nodes/meshy/meshy-6" - ] - }, - { - "group": "Bria", - "pages": [ - "ko/tutorials/partner-nodes/bria/fibo", - "ko/tutorials/partner-nodes/bria/background-removal" - ] - }, - { - "group": "Reve", + "group": "이미지", "pages": [ - "ko/tutorials/partner-nodes/reve/reve-image" + { + "group": "ByteDance", + "pages": [ + "ko/tutorials/partner-nodes/bytedance/seedream-5-pro", + "ko/tutorials/partner-nodes/bytedance/seedream-5-lite" + ] + }, + { + "group": "Google", + "pages": [ + "ko/tutorials/partner-nodes/google/nano-banana-pro", + "ko/tutorials/partner-nodes/google/nano-banana-2", + "ko/tutorials/partner-nodes/google/nano-banana-2-lite" + ] + }, + { + "group": "Black Forest Labs", + "pages": [ + "ko/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image", + "ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext" + ] + }, + { + "group": "Ideogram", + "pages": [ + "ko/tutorials/partner-nodes/ideogram/ideogram-v4" + ] + }, + { + "group": "Krea 2", + "pages": [ + "ko/tutorials/partner-nodes/krea2/krea2-t2i" + ] + }, + { + "group": "Recraft", + "pages": [ + "ko/tutorials/partner-nodes/recraft/recraft-v4", + "ko/tutorials/partner-nodes/recraft/recraft-text-to-image" + ] + }, + { + "group": "Bria", + "pages": [ + "ko/tutorials/partner-nodes/bria/fibo", + "ko/tutorials/partner-nodes/bria/background-removal" + ] + }, + { + "group": "Reve", + "pages": [ + "ko/tutorials/partner-nodes/reve/reve-image" + ] + }, + { + "group": "Luma", + "pages": [ + "ko/tutorials/partner-nodes/luma/luma-uni-1", + "ko/tutorials/partner-nodes/luma/luma-text-to-image", + "ko/tutorials/partner-nodes/luma/luma-image-to-image" + ] + }, + { + "group": "OpenAI", + "pages": [ + "ko/tutorials/partner-nodes/openai/gpt-image-2", + "ko/tutorials/partner-nodes/openai/gpt-image-1", + "ko/tutorials/partner-nodes/openai/dall-e-2", + "ko/tutorials/partner-nodes/openai/dall-e-3" + ] + }, + { + "group": "Runway", + "pages": [ + "ko/tutorials/partner-nodes/runway/image-generation" + ] + }, + { + "group": "Beeble", + "pages": [ + "ko/tutorials/partner-nodes/beeble/beeble-switchx" + ] + } ] }, { - "group": "Wan", + "group": "비디오", "pages": [ - "ko/tutorials/partner-nodes/wan/wan2-7" + { + "group": "ByteDance", + "pages": [ + "ko/tutorials/partner-nodes/bytedance/seedance-2-0", + "ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human" + ] + }, + { + "group": "Google", + "pages": [ + "ko/tutorials/partner-nodes/google/gemini-omni-flash" + ] + }, + { + "group": "Kling", + "pages": [ + "ko/tutorials/partner-nodes/kling/kling-3-0", + "ko/tutorials/partner-nodes/kling/kling-motion-control" + ] + }, + { + "group": "Luma", + "pages": [ + "ko/tutorials/partner-nodes/luma/luma-text-to-video", + "ko/tutorials/partner-nodes/luma/luma-image-to-video" + ] + }, + { + "group": "Moonvalley", + "pages": [ + "ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation" + ] + }, + { + "group": "Runway", + "pages": [ + "ko/tutorials/partner-nodes/runway/video-generation" + ] + }, + { + "group": "Wan", + "pages": [ + "ko/tutorials/partner-nodes/wan/wan2-7" + ] + }, + { + "group": "HappyHorse", + "pages": [ + "ko/tutorials/partner-nodes/happyhorse/happyhorse1-1", + "ko/tutorials/partner-nodes/happyhorse/happyhorse1-0" + ] + }, + { + "group": "Topaz", + "pages": [ + "ko/tutorials/partner-nodes/topaz/astra-2" + ] + } ] }, { - "group": "HappyHorse", + "group": "3D", "pages": [ - "ko/tutorials/partner-nodes/happyhorse/happyhorse1-1", - "ko/tutorials/partner-nodes/happyhorse/happyhorse1-0" + { + "group": "Rodin", + "pages": [ + "ko/tutorials/partner-nodes/rodin/model-generation" + ] + }, + { + "group": "Tripo", + "pages": [ + "ko/tutorials/partner-nodes/tripo/model-generation", + "ko/tutorials/partner-nodes/tripo/tripo-3-1" + ] + }, + { + "group": "Hunyuan 3D", + "pages": [ + "ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0" + ] + }, + { + "group": "Meshy", + "pages": [ + "ko/tutorials/partner-nodes/meshy/meshy-6" + ] + } ] }, { - "group": "Sonilo", + "group": "오디오", "pages": [ - "ko/tutorials/partner-nodes/sonilo/video-to-music" + { + "group": "ByteDance", + "pages": [ + "ko/tutorials/partner-nodes/bytedance/seed-audio-1-0" + ] + }, + { + "group": "Sonilo", + "pages": [ + "ko/tutorials/partner-nodes/sonilo/video-to-music" + ] + } ] }, { - "group": "Topaz", + "group": "LLM", "pages": [ - "ko/tutorials/partner-nodes/topaz/astra-2" + { + "group": "Google", + "pages": [ + "ko/tutorials/partner-nodes/google/gemini" + ] + }, + { + "group": "Anthropic", + "pages": [ + "ko/tutorials/partner-nodes/anthropic/claude" + ] + }, + { + "group": "OpenAI", + "pages": [ + "ko/tutorials/partner-nodes/openai/chat" + ] + }, + { + "group": "OpenRouter", + "pages": [ + "ko/tutorials/partner-nodes/openrouter/llm" + ] + } ] } ] diff --git a/ja/tutorials/3d/hunyuan3D-2.mdx b/ja/tutorials/3d/hunyuan3D-2.mdx index bb5a410d6..3779ec1dc 100644 --- a/ja/tutorials/3d/hunyuan3D-2.mdx +++ b/ja/tutorials/3d/hunyuan3D-2.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Hunyuan3D 2.0 Open-Source Model Series": 363ff037 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # Hunyuan3D 2.0 はじめに @@ -28,7 +26,6 @@ Hunyuan3D 2.0 は 2 段階の生成アプローチを採用しており、まず 1. **幾何形状生成モデル（Hunyuan3D-DiT）**：フロー拡散 Transformer アーキテクチャに基づき、入力条件に正確に一致するテクスチャのない幾何形状モデルを生成します。 2. **テクスチャ生成モデル（Hunyuan3D-Paint）**：幾何条件とマルチビュー拡散技術を組み合わせて、モデルに高解像度のテクスチャを追加し、PBR マテリアルをサポートします。 - **主な利点** - **高精度生成**：鋭い幾何構造、豊かなテクスチャカラー、PBR マテリアル生成のサポートにより、ほぼ現実的な照明効果を実現。 @@ -52,179 +49,242 @@ ComfyUI は現在 Hunyuan3D-2mv をネイティブサポートしていますが これにより、対応するワークフローが読み込まれ、必要なモデルのダウンロードを促されます。生成された `.glb` 形式のモデルは `ComfyUI/output/mesh` フォルダに出力されます。 -## ComfyUI Hunyuan3D-2mv ワークフロー例 +## ComfyUI Hunyuan3D-2mv Workflow Example -Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して 3D モデルを生成します。このワークフローでは複数のビュー画像は必須ではありません - `front` ビュー画像のみを使用して 3D モデルを生成できます。 +In the Hunyuan3D-2mv workflow, we'll use multi-view images to generate a 3D model. Note that multiple view images are not mandatory in this workflow - you can use only the `front` view image to generate a 3D model. +### HY 3D 2.0 MV (`3d_hunyuan3d_multiview_to_model`) + +Generate 3D models from multiple views using Hunyuan3D 2.0 MV. + +HY 3D 2.0 MV workflow preview + - Comfy Cloud でこのワークフローを即座に実行 + Run this workflow instantly on Comfy Cloud - ワークフロー JSON ファイルをダウンロード + Download JSON or search "HY 3D 2.0 MV" in Template Library + + + +**入力素材** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 56 · front view + + + `LoadImage` node 78 · left view + + + `LoadImage` node 80 · back view + + + `LoadImage` node 87 · right view -### 1. ワークフロー +### 1. Workflow -以下の画像をダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください。 +Please download the images below and drag into ComfyUI to load the workflow. ![Hunyuan3D-2mv workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/hunyuan-3d-multiview-elf.webp) -以下の画像をダウンロードしてください。これらを入力画像として使用します。 +Download the images below we will use them as input images. ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/front.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/left.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/back.png) - -この例では、入力画像はすでに余分な背景を除去するように前処理されています。実際の使用では、[ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) のようなカスタムノードを使用して、余分な背景を自動的に除去できます。 +In this example, the input images have already been preprocessed to remove excess background. In actual use, you can use custom nodes like [ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) to automatically remove excess background. -### 2. 手動モデルインストール +### 2. Manual Model Installation -以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv.safetensors // リネーム後のファイル +│ │ └── hunyuan3d-dit-v2-mv.safetensors // renamed file ``` -### 3. ワークフローの実行手順 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2mv](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv.jpg) -1. Image Only Checkpoint Loader(img2vid model) が、ダウンロードしてリネームした `hunyuan3d-dit-v2-mv.safetensors` モデルを読み込んでいることを確認します -2. 各 `Load Image` ノードに対応するビュー画像を読み込みます -3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します +1. Ensure that the Image Only Checkpoint Loader(img2vid model) has loaded our downloaded and renamed `hunyuan3d-dit-v2-mv.safetensors` model +2. Load the corresponding view images in each of the `Load Image` nodes +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow + +If you need to add more views, make sure to load other view images in the `Hunyuan3Dv2ConditioningMultiView` node, and ensure that you load the corresponding view images in the `Load Image` nodes. + +## Hunyuan3D-2mv-turbo Workflow -より多くのビューを追加する必要がある場合は、`Hunyuan3Dv2ConditioningMultiView` ノードに他のビュー画像を読み込み、`Load Image` ノードに対応するビュー画像を読み込むことを確認してください。 +In the Hunyuan3D-2mv-turbo workflow, we'll use the Hunyuan3D-2mv-turbo model to generate 3D models. This model is a step distillation version of Hunyuan3D-2mv, allowing for faster 3D model generation. In this version of the workflow, we set `cfg` to 1.0 and add a `flux guidance` node to control the `distilled cfg` generation. -## Hunyuan3D-2mv-turbo ワークフロー +### HY 3D 2.0 MV Turbo (`3d_hunyuan3d_multiview_to_model_turbo`) -Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを使用して 3D モデルを生成します。このモデルは Hunyuan3D-2mv のステップ蒸留バージョンで、より高速な 3D モデル生成を可能にします。このバージョンのワークフローでは、`cfg` を 1.0 に設定し、`flux guidance` ノードを追加して `distilled cfg` 生成を制御します。 +Generate 3D models from multiple views using Hunyuan3D 2.0 MV Turbo. + +HY 3D 2.0 MV Turbo workflow preview - Comfy Cloud でこのワークフローを即座に実行 + Run this workflow instantly on Comfy Cloud - ワークフロー JSON ファイルをダウンロード + Download JSON or search "HY 3D 2.0 MV Turbo" in Template Library -### 1. ワークフロー +**入力素材** -以下の画像をダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください。 +Upload these files to the matching `LoadImage` nodes: -![Hunyuan3D-2mv-turbo workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) + + + `LoadImage` node 56 · front view + + + `LoadImage` node 82 · back view + + + `LoadImage` node 85 · left view + + + `LoadImage` node 87 · right view + + + +### 1. Workflow -以下の画像をダウンロードしてください。これらを入力画像として使用します。 +Please download the images below and drag into ComfyUI to load the workflow. +![Hunyuan3D-2mv-turbo workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) + +Download the images below we will use them as input images. ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) +### 2. Manual Model Installation -### 2. 手動モデルインストール - -以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv-turbo.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // リネーム後のファイル +│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // renamed file ``` -### 3. ワークフローの実行手順 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2mv_turbo](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv_turbo.jpg) -1. `Image Only Checkpoint Loader(img2vid model)` ノードが、リネームした `hunyuan3d-dit-v2-mv-turbo.safetensors` モデルを読み込んでいることを確認します -2. 各 `Load Image` ノードに対応するビュー画像を読み込みます -3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します +1. Ensure that the `Image Only Checkpoint Loader(img2vid model)` node has loaded our renamed `hunyuan3d-dit-v2-mv-turbo.safetensors` model +2. Load the corresponding view images in each of the `Load Image` nodes +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -## Hunyuan3D-2 単一ビューワークフロー +## Hunyuan3D-2 Single View Workflow -Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D モデルを生成します。このモデルはマルチビューモデルではありません。このワークフローでは、`Hunyuan3Dv2ConditioningMultiView` ノードの代わりに `Hunyuan3Dv2Conditioning` ノードを使用します。 +In the Hunyuan3D-2 workflow, we'll use the Hunyuan3D-2 model to generate 3D models. This model is not a multi-view model. In this workflow, we use the `Hunyuan3Dv2Conditioning` node instead of the `Hunyuan3Dv2ConditioningMultiView` node. + +### HY 3D 2.0 (`3d_hunyuan3d_image_to_model`) + +Generate 3D models from single images using Hunyuan3D 2.0. + +HY 3D 2.0 workflow preview - Comfy Cloud でこのワークフローを即座に実行 + Run this workflow instantly on Comfy Cloud - ワークフロー JSON ファイルをダウンロード + Download JSON or search "HY 3D 2.0" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 56 · `3d_hunyuan3d_image_to_model_input_image.png` -### 1. ワークフロー +### 1. Workflow -以下の画像をダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください。 +Please download the image below and drag it into ComfyUI to load the workflow. ![Hunyuan3D-2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) -以下の画像をダウンロードしてください。これを入力画像として使用します。 +Download the image below we will use it as input image. ![ComfyUI Hunyuan 3D 2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan_3d_v2_non_multiview_train.png) -### 2. 手動モデルインストール +### 2. Manual Model Installation -以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2.safetensors // リネーム後のファイル +│ │ └── hunyuan3d-dit-v2.safetensors // renamed file ``` -### 3. ワークフローの実行手順 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2_non_multiview.jpg) -1. `Image Only Checkpoint Loader(img2vid model)` ノードが、リネームした `hunyuan3d-dit-v2.safetensors` モデルを読み込んでいることを確認します -2. `Load Image` ノードに画像を読み込みます -3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します +1. Ensure that the `Image Only Checkpoint Loader(img2vid model)` node has loaded our renamed `hunyuan3d-dit-v2.safetensors` model +2. Load the image in the `Load Image` node +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -## コミュニティリソース +## Community Resources -以下は Hunyuan3D-2 に関連する ComfyUI コミュニティリソースです +Below are ComfyUI community resources related to Hunyuan3D-2 - [ComfyUI-Hunyuan3DWrapper](https://github.com/kijai/ComfyUI-Hunyuan3DWrapper) - [Kijai/Hunyuan3D-2_safetensors](https://huggingface.co/Kijai/Hunyuan3D-2_safetensors/tree/main) - [ComfyUI-3D-Pack](https://github.com/MrForExample/ComfyUI-3D-Pack) -## Hunyuan3D 2.0 オープンソースモデルシリーズ +## Hunyuan3D 2.0 Open-Source Model Series -現在、Hunyuan3D 2.0 は完全な 3D 生成プロセスをカバーする複数のモデルをオープンソース化しています。詳細については [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2) をご覧ください。 +Currently, Hunyuan3D 2.0 has open-sourced multiple models covering the complete 3D generation process. You can visit [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2) for more information. -**Hunyuan3D-2mini シリーズ** +**Hunyuan3D-2mini Series** -| モデル | 説明 | 日付 | パラメータ | Huggingface | +| Model | Description | Date | Parameters | Huggingface | |-----------------------|---------------------------|------------|------------|--------------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-mini | Mini 画像から形状モデル | 2025-03-18 | 0.6B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2mini/tree/main/hunyuan3d-dit-v2-mini) | +| Hunyuan3D-DiT-v2-mini | Mini Image to Shape Model | 2025-03-18 | 0.6B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mini/tree/main/hunyuan3d-dit-v2-mini) | -**Hunyuan3D-2mv シリーズ** +**Hunyuan3D-2mv Series** -| モデル | 説明 | 日付 | パラメータ | Huggingface | +| Model | Description | Date | Parameters | Huggingface | |--------------------------|-------------------------------------------------------------------------------------------------------------|------------|------------|------------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-mv-Fast | ガイダンス蒸留バージョン、DIT 推論時間を半分に短縮可能 | 2025-03-18 | 1.1B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv-fast) | -| Hunyuan3D-DiT-v2-mv | マルチビュー画像から形状モデル、シーンを理解するために複数の角度が必要な 3D 制作に適する | 2025-03-18 | 1.1B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv) | +| Hunyuan3D-DiT-v2-mv-Fast | Guidance Distillation Version, can halve DIT inference time | 2025-03-18 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv-fast) | +| Hunyuan3D-DiT-v2-mv | Multi-view Image to Shape Model, suitable for 3D creation requiring multiple angles to understand the scene | 2025-03-18 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv) | -**Hunyuan3D-2 シリーズ** +**Hunyuan3D-2 Series** -| モデル | 説明 | 日付 | パラメータ | Huggingface | +| Model | Description | Date | Parameters | Huggingface | |-------------------------|-----------------------------|------------|------------|---------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-0-Fast | ガイダンス蒸留モデル | 2025-02-03 | 1.1B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast) | -| Hunyuan3D-DiT-v2-0 | 画像から形状モデル | 2025-01-21 | 1.1B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0) | -| Hunyuan3D-Paint-v2-0 | テクスチャ生成モデル | 2025-01-21 | 1.3B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-paint-v2-0) | -| Hunyuan3D-Delight-v2-0 | 画像デライトモデル | 2025-01-21 | 1.3B | [訪問](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-delight-v2-0) | +| Hunyuan3D-DiT-v2-0-Fast | Guidance Distillation Model | 2025-02-03 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast) | +| Hunyuan3D-DiT-v2-0 | Image to Shape Model | 2025-01-21 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0) | +| Hunyuan3D-Paint-v2-0 | Texture Generation Model | 2025-01-21 | 1.3B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-paint-v2-0) | +| Hunyuan3D-Delight-v2-0 | Image Delight Model | 2025-01-21 | 1.3B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-delight-v2-0) | diff --git a/ja/tutorials/3d/triposplat.mdx b/ja/tutorials/3d/triposplat.mdx index 1153991fe..07f8f4cb9 100644 --- a/ja/tutorials/3d/triposplat.mdx +++ b/ja/tutorials/3d/triposplat.mdx @@ -13,10 +13,6 @@ translationBlockHashes: "Model downloads": 44894623 --- - - - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **TripoSplat** は、単一の 2D 画像から直接 **3D ガウシアンスプラット（Gaussian splat）** 表現を生成するオープンソースモデルです。VAST-AI によって開発され、オープンソースライセンスで公開されています。 @@ -54,83 +50,132 @@ TripoSplat は **フィードフォワードアーキテクチャ** を使用し - 入力画像を読み込みます（PNG/JPG） - サンプル画像：`white-hotel-on-rocky-island.png`（テンプレートライブラリから入手可能） -### TripoSplat（サブグラフ） +### TripoSplat: Image to Gaussian Splat (`3d_triposplat_image_to_gaussian_splat`) + +Upload a single 2D image. Generate a high-quality 3D Gaussian splat representation with controllable density and budget for rendering. + +TripoSplat workflow preview + + + + + + Run this workflow instantly on Comfy Cloud + + + Download JSON or search "TripoSplat: Image to Gaussian Splat" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 99 · `white-hotel-on-rocky-island.png` + + + +
+ Input image +
+ +## How it works + +TripoSplat uses a **feed-forward architecture** that takes a single RGB image and directly predicts a set of 3D Gaussian primitives. The pipeline involves: + +1. **Image encoding**: the input image is processed by a vision encoder (DINOv2) +2. **Triplane generation**: features are decoded into a triplane representation +3. **Gaussian prediction**: the triplane is sampled to produce Gaussian parameters (position, scale, rotation, opacity, color) +4. **Rendering**: Gaussians are rendered from arbitrary viewpoints using differentiable splatting + + + This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. + +## Workflow node guide + +### LoadImage +- Loads your input image (PNG/JPG) +- Sample image: `white-hotel-on-rocky-island.png` (available in Template Library) + +### TripoSplat (subgraph) -画像を処理して 3D ガウシアンスプラットを生成するメインのサブグラフノード。公開パラメータ： +The main subgraph node processes the image and generates the 3D Gaussian splat. Exposed parameters: -| パラメータ | デフォルト | 説明 | +| Parameter | Default | Description | |---|---|---| -| `switch` | — | サブグラフの有効/無効 | -| `num_gaussians` | — | 生成するガウシアンプリミティブの数（品質/パフォーマンス制御） | -| `seed` | — | 再現性のためのシード値 | -| `unet_name` | — | TripoSplat 拡散モデルチェックポイント | -| `clip_name` | — | CLIP ビジョンエンコーダーモデル | -| `vae_name` | — | エンコード/デコード用の VAE（2つ：メイン VAE とエンコーダー） | -| `bg_removal_name` | — | 背景除去モデル | +| `switch` | — | Enable/disable the subgraph | +| `num_gaussians` | — | Number of Gaussian primitives to generate (controls quality/performance) | +| `seed` | — | Random seed for reproducibility | +| `unet_name` | — | TripoSplat diffusion model checkpoint | +| `clip_name` | — | CLIP vision encoder model | +| `vae_name` | — | VAE for encoding/decoding (2 entries: one for the main VAE, one for the encoder) | +| `bg_removal_name` | — | Background removal model | ### CreateCameraInfo -- 結果をレンダリングするカメラ軌道を定義 -- パラメータ：軌道タイプ、角度、距離、視野など -- デフォルト：仰角 35°、距離 30、ズーム 2.5 +- Defines the camera orbit for rendering the result +- Parameters: orbit type, angle, distance, field of view, etc. +- Default: orbit at 35° elevation, 30 distance, 2.5 zoom ### RenderSplat -- 定義されたカメラ角度からガウシアンスプラットを 2D 画像にレンダリング -- パラメータ：出力解像度（デフォルト 1024×1024）、画質設定 +- Renders the Gaussian splat into a 2D image from the defined camera angle +- Parameters: output resolution (default 1024×1024), image quality settings ### SplatToMesh -- ガウシアンスプラットをメッシュに変換（オプション） -- パラメータ：メッシュ密度、スムージング、簡略化 +- Converts the Gaussian splat to a mesh (optional) +- Parameters: mesh density, smoothing, simplification ### SaveGLB -- 結果を GLB 3D ファイルとして保存 +- Saves the result as a GLB 3D file ### SaveVideo -- レンダリングされた 3D シーンの動画を保存 +- Saves a video of the rendered 3D scene ### SplatToFile3D -- ガウシアンスプラットを SPZ 形式でエクスポート +- Exports the Gaussian splat in SPZ format ### CreateVideo -- レンダリングフレームから動画を作成 +- Creates a video from rendered frames -## 実行手順 +## Steps to run -1. **画像を読み込む** — **LoadImage** ノードで単一の 2D 画像を読み込みます -2. **TripoSplat サブグラフを実行** — モデルがガウシアンスプラット表現を生成します -3. **出力形式を選択** — GLB、SPZ、動画、またはメッシュに変換 -4. **結果を表示** — 生成された 3D ファイルまたはレンダリングプレビューを確認 +1. **Load an image**: use the **LoadImage** node to load a single 2D image +2. **Run the TripoSplat subgraph**: the model will generate a Gaussian splat representation +3. **Choose output format**: export as GLB, SPZ, video, or render to mesh +4. **View results**: use the created 3D file or rendered preview -## 出力オプション +## Output options -| ノード | 形式 | 用途 | +| Node | Format | Use case | |---|---|---| -| **SaveGLB** | `.glb` | 標準的な 3D ファイル形式。3D ソフトウェアにインポート可能 | -| **SplatToFile3D** | `.spz` | 圧縮されたガウシアンスプラット形式。効率的な保存 | -| **RenderSplat** | 2D 画像 | 任意の角度から結果を素早くプレビュー | -| **SplatToMesh** | メッシュ | 従来のメッシュに変換してさらに編集 | +| **SaveGLB** | `.glb` | Standard 3D file format, importable into 3D software | +| **SplatToFile3D** | `.spz` | Compressed Gaussian splat format for efficient storage | +| **RenderSplat** | 2D image | Quick preview of the result from any angle | +| **SplatToMesh** | Mesh | Convert to traditional mesh for further editing | -## モデルダウンロード +## Model downloads -TripoSplat モデルと必要なファイルをダウンロードします。対応する `models/` サブディレクトリに配置してください。 +Download the TripoSplat model and required files. Place them in the corresponding `models/` subdirectories. - - triposplat_fp16.safetensors — TripoSplat 拡散モデルチェックポイント + + triposplat_fp16.safetensors: TripoSplat diffusion model checkpoint - - triposplat_vae_decoder_fp16.safetensors — VAE デコーダー + + triposplat_vae_decoder_fp16.safetensors: VAE decoder - flux2-vae.safetensors — Flux.2 VAE、潜在表現エンコード用 + flux2-vae.safetensors: Flux.2 VAE for latent encoding - dino_v3_vit_h.safetensors — CLIP ビジョンエンコーダー（DINOv2） + dino_v3_vit_h.safetensors: CLIP vision encoder (DINOv2) - - birefnet.safetensors — 前処理用の背景除去モデル + + birefnet.safetensors: Background removal model for preprocessing -### モデル保存場所 +### Model storage location ``` 📂 ComfyUI/ diff --git a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx index 97dd4f880..9abced842 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "ACE-Step 1.5 ComfyUI Related Resources": 518b9b68 --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## ComfyUIにおけるACE-Step 1.5について @@ -29,27 +28,30 @@ ACE-Step 1.5は、オープンソースの音楽生成モデルに対する重 -## オプション1：オールインワンチェックポイント（推奨） +## Option 1: All-in-One Checkpoint (Recommended) -AIO（All-in-One）版は、すべてのモデルを単一のチェックポイントファイルにパッケージ化しており、ダウンロードおよび管理が容易です。 +The AIO version packages all models into a single checkpoint file, making it easier to download and manage. -### AIOワークフロー +### ACE-Step 1.5 Music Generation AIO (`audio_ace_step_1_5_checkpoint`) - - AIOワークフローをComfy Cloud上で直接実行します。 - +Input style tags and lyrics to generate a full song. The workflow uses the ACE-Step 1.5 model to produce commercial-grade music in under 10 seconds on consumer hardware. - - ローカル環境で使用するためのオールインワンチェックポイントワークフローをダウンロードします。 + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation AIO" in Template Library + -### AIOモデルのダウンロード +### AIO Model Download - - オールインワンチェックポイントファイル（大多数のユーザーに推奨）。 + + All-in-one checkpoint file (recommended for most users). -**AIOモデルの保存場所** +**AIO Model Storage Location** ``` 📂 ComfyUI/ @@ -58,39 +60,44 @@ AIO（All-in-One）版は、すべてのモデルを単一のチェックポイ │ └── ace_step_1.5_turbo_aio.safetensors ``` -## オプション2：分割モデルファイル +## Option 2: Split Model Files -分割版では、個別のモデルコンポーネントをそれぞれ別々にダウンロードできます。 +The split version allows you to download individual model components separately. -### 分割モデルワークフロー +### ACE-Step 1.5 Music Generation Workflow (`audio_ace_step_1_5_split`) - - 分割モデルワークフローをComfy Cloud上で直接実行します。 - +Input a text prompt describing the music style and optional lyrics. Generate a full, high-quality audio song in under 10 seconds on consumer hardware. - - ローカル環境で使用するための分割モデルワークフローをダウンロードします。 + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation Workflow" in Template Library + -### 分割モデルのダウンロード +### Split Model Downloads - - 拡散モデル（Diffusion Model）。 + + + Diffusion model. - - テキストエンコーダー（0.6B）。 + + Text encoder (0.6B). - - テキストエンコーダー（1.7B）。 + + Text encoder (1.7B). - - VAEモデル。 + + VAE model. + -**分割モデルの保存場所** +**Split Models Storage Location** ``` 📂 ComfyUI/ @@ -104,26 +111,26 @@ AIO（All-in-One）版は、すべてのモデルを単一のチェックポイ │ └── ace_1.5_vae.safetensors ``` -## ComfyUIにおけるACE-Step 1.5の主要機能 +## ACE-Step 1.5 Key Features in ComfyUI -### 思考連鎖（Chain-of-Thought）によるプランニング +### Chain-of-Thought Planning -ACE-Step 1.5モデルは、思考連鎖（Chain-of-Thought）推論を用いてメタデータ、歌詞、キャプションを統合し、拡散プロセスを効果的に制御します。その結果、より整合性の高い長尺楽曲の生成が可能になります。 +The ACE-Step 1.5 model synthesizes metadata, lyrics, and captions via Chain-of-Thought reasoning to guide the diffusion process, resulting in more coherent long-form compositions. -### ハイブリッドLM＋DiTアーキテクチャ +### Hybrid LM + DiT Architecture -ACE-Step 1.5は、楽曲構造を計画する言語モデル（LM）と、音声合成を担当する拡散トランスフォーマー（DiT）を組み合わせたハイブリッドアーキテクチャを採用しており、両者がComfyUI上でネイティブに動作します。 +ACE-Step 1.5 combines a Language Model that plans the song structure with a Diffusion Transformer (DiT) that handles audio synthesis, all running natively in ComfyUI. -## ComfyUIへの今後の追加予定機能 +## Coming Soon to ComfyUI -以下の機能はACE-Step 1.5には既に実装済みですが、現時点ではComfyUIではまだサポートされていません： +These features are available in ACE-Step 1.5 but not yet supported in ComfyUI: -- **カバー（Cover）**：任意の楽曲を入力として与え、新しいプロンプトおよび歌詞とともに提示すると、モデルが全く異なるスタイルでその楽曲を再解釈・再構成します -- **リペイント（Repaint）**：音声の一部区間を選択し、その部分のみを再生成します。モデルは他の部分を一切変更せずに、再生成した区間をシームレスに挿入・統合します +- **Cover**: Give the model any song as input along with a new prompt and lyrics, and it will reimagine the track in a completely different style +- **Repaint**: Select a segment, regenerate just that section, and the model stitches it back in while keeping everything else untouched -## ACE-Step 1.5関連のComfyUIリソース +## ACE-Step 1.5 ComfyUI Related Resources -- [プロジェクトページ](https://ace-step.github.io/) +- [Project Page](https://ace-step.github.io/) - [Hugging Face](https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files) - [GitHub](https://github.com/ace-step/ACE-Step) -- [ブログ記事](https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui) +- [Blog Post](https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui) diff --git a/ja/tutorials/audio/ace-step/ace-step-v1.mdx b/ja/tutorials/audio/ace-step/ace-step-v1.mdx index d558882bf..2a2e468ce 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "ACE-Step Related Resources": f74e2db9 --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ACE-Step は、中国のチーム StepFun と ACE Studio が共同開発したオープンソースの音楽生成基盤モデルであり、音楽クリエイターに効率的で柔軟性が高く、高品質な音楽生成および編集ツールを提供することを目的としています。 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx index a0c3bcfa4..57309ad74 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -15,37 +15,42 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" **関連リンク**： - [GitHub: Stability-AI/stable-audio-open-1.0](https://github.com/Stability-AI/stable-audio-open-1.0) -## ワークフロー +## Workflow - - JSON をダウンロードするか、テンプレートライブラリで"Stable Audio 1.0"を検索 - +### Stable Audio 1.0: Text to Audio (`audio_stable_audio_example`) - - Comfy Cloud で開く - +Generate audio from text prompts using Stable Audio. + +Stable Audio 1.0 text to audio workflow preview -![Stable Audio 1.0 ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_example-1.webp) + + + Open in Comfy Cloud + + + Download JSON or search "Stable Audio 1.0: Text to Audio" in Template Library + + -**標準の ComfyUI ノード**のみを使用し、カスタムノードは不要です。Stable Audio 1.0 チェックポイントを読み込み、CLIP テキストエンコーダーでプロンプトをエンコードし、KSampler で潜在空間をノイズ除去し、VAE で音声にデコードします。 +The workflow uses **standard ComfyUI nodes** — no custom nodes required. It loads the Stable Audio 1.0 checkpoint, encodes your prompt via a CLIP text encoder (t5-base), denoises the latent audio with a KSampler, and decodes it to audio through the model's VAE. -**使用方法**： -1. **モデルを読み込む** — `CheckpointLoaderSimple` ノードで `stable-audio-open-1.0.safetensors` を使用 -2. **プロンプトを書く** — `CLIPTextEncode` ノードに説明を入力（例："heaven church electronic dance music"） -3. **再生時間を設定** — `EmptyLatentAudio` ノードの長さを調整（デフォルト 47.6 秒） -4. **実行**（`Ctrl/Cmd + Enter`）をクリックして生成。音声は `ComfyUI/output/audio/` に保存されます +**How to use**: +1. **Load the checkpoint** — The `CheckpointLoaderSimple` node uses `stable-audio-open-1.0.safetensors` +2. **Write a prompt** — Enter your description in the `CLIPTextEncode` node (e.g. "heaven church electronic dance music") +3. **Set duration** — Adjust the `EmptyLatentAudio` node's length value (default 47.6 seconds) +4. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -## モデルダウンロード +## Model download -ワークフローを読み込むと、モデルがない場合に ComfyUI がダウンロードリンクを提示します。手動で設定する場合、以下のファイルをダウンロードして適切なフォルダに配置してください。 +When loading the workflow, ComfyUI will prompt you with download links for any missing models. To set up manually, download the files below and place them in the correct folders. -### チェックポイント +### Checkpoint - - 2.3GB。models/checkpoints/ に配置 + + 2.3GB. Place in models/checkpoints/ -以下のように配置します： +Place the checkpoint in: ``` 📂 ComfyUI/ @@ -54,13 +59,13 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" │ └── stable-audio-open-1.0.safetensors ``` -### テキストエンコーダー +### Text encoder - - プロンプト処理用テキストエンコーダー。models/text_encoders/ に配置 + + Text encoder for prompt conditioning. Place in models/text_encoders/ -以下のように配置します： +Place the text encoder in: ``` 📂 ComfyUI/ @@ -69,4 +74,4 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" │ └── t5-base.safetensors ``` -配置後、ComfyUI で **R** キーを押してノード定義をリフレッシュすると、最新のモデルが利用可能になります。 +After placing the files, press **R** in ComfyUI to refresh nodes and load the latest models. diff --git a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx index e638d2d56..1eec88452 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Model download": d729490a --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" @@ -27,65 +26,73 @@ Stable Audio 3 には3つのバリエーションがあります： - [Hugging Face (Comfy-Org/stable-audio-3)](https://huggingface.co/Comfy-Org/stable-audio-3) - [ブログ：リリース告知](https://blog.comfy.org/p/stable-audio-3-day-0-support) -## 利用可能なワークフロー +## Available workflows -### Stable Audio 3 Medium +### Stable Audio 3.0 Medium (`audio_stable_audio_3_medium`) - - JSON をダウンロードするか、テンプレートライブラリで"Stable Audio 3 Medium"を検索 - +Input a short text idea, optional duration, seed, and category. Generate stereo audio (music, SFX, or instruments) using Stable Audio 3 with optional AI-driven text expansion. - - Comfy Cloud で開く +Stable Audio 3 Medium workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Stable Audio 3.0 Medium" in Template Library + -![Stable Audio 3 Medium ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium-1.webp) +The **Stable Audio 3 Medium** workflow is a full-featured text-to-audio generation pipeline. You provide a short text idea, optional duration, seed, and category — the workflow expands your prompt using Qwen with a **category-aware reprompt template**, then generates stereo audio via the Stable Audio 3 checkpoint. -**Stable Audio 3 Medium** ワークフローは、完全なテキストから音声生成パイプラインです。短いテキストアイデア、任意の再生時間、シード、カテゴリを入力すると、Qwen を使用したカテゴリ認識リプロンプトテンプレートでプロンプトを拡張し、Stable Audio 3 チェックポイントでステレオ音声を生成します。 +**How to use**: +1. **Text idea** — Enter a short description of the sound, music, or effect you want (e.g. "upbeat electronic dance track with heavy bass") +2. **Duration** — Set the desired clip length in seconds (default varies) +3. **Seed** — Control reproducibility by adjusting the seed value +4. **Category** — Choose a reprompt preset: **Music**, **Instrument**, **SFX**, or **One-shot** +5. **Enable reprompt** — Toggle `use_reprompt` on to let Qwen expand your short idea into a detailed prompt before generation +6. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -**使用方法**： -1. **テキストアイデア** — 生成したい音声の簡単な説明を入力（例：「重いベースのアップテンポなエレクトロニックダンスミュージック」） -2. **再生時間** — クリップの長さ（秒）を設定 -3. **シード** — 再現性を制御 -4. **カテゴリ** — リプロンプトプリセットを選択：**Music**（音楽）、**Instrument**（楽器）、**SFX**（効果音）、**One-shot**（単発音） -5. **リプロンプトを有効化** — `use_reprompt` をオンにして Qwen が短いアイデアを詳細なプロンプトに拡張 -6. **実行**（`Ctrl/Cmd + Enter`）をクリックして生成。音声は `ComfyUI/output/audio/` に保存されます +### Stable Audio 3.0 Medium Base (`audio_stable_audio_3_medium_base`) -### Stable Audio 3 Medium Base +Input a short text description of a sound, music, or effect. The workflow expands your prompt with Qwen and generates a stereo audio clip from Stable Audio 3. - - JSON をダウンロードするか、テンプレートライブラリで"Stable Audio 3 Medium Base"を検索 - +Stable Audio 3 Medium Base workflow preview - - Comfy Cloud で開く + + + Open in Comfy Cloud + + Download JSON or search "Stable Audio 3.0 Medium Base" in Template Library + + -![Stable Audio 3 Medium Base ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium_base-1.webp) - -Qwen リプロンプト機能を省いたシンプルなバージョン。完全なテキストプロンプトを直接 Stable Audio 3 モデルに渡します。すでに詳細なプロンプトがある場合や、高速に生成したい場合に使用します。 +A simplified version of Stable Audio 3 without Qwen reprompt expansion. It expects a complete text prompt and passes it directly to the model. Use this when you already have a detailed prompt and want faster generation. -**使用方法**： -1. **テキストプロンプト** — 生成したい音声の詳細な説明を入力 -2. **再生時間** — クリップの長さ（秒）を設定 -3. **シード** — 再現性を制御 -4. **実行**（`Ctrl/Cmd + Enter`）をクリックして生成 +**How to use**: +1. **Text prompt** — Enter a detailed description of the audio you want +2. **Duration** — Set the clip length in seconds +3. **Seed** — Control reproducibility +4. Click **Run** (`Ctrl/Cmd + Enter`) to generate -## モデルダウンロード +## Model download -ワークフローを読み込むと、モデルがない場合に ComfyUI がダウンロードリンクを提示します。手動で設定する場合、以下のファイルをダウンロードして適切なフォルダに配置してください。 +When loading the workflow, ComfyUI will prompt you with download links for any missing models. To set up manually, download the files below and place them in the correct folders. -### チェックポイント +### Checkpoints - - Medium ワークフロー用。models/checkpoints/ に配置 + + + For the Medium workflow. Place in models/checkpoints/ - - Medium Base ワークフロー用。models/checkpoints/ に配置 + + For the Medium Base workflow. Place in models/checkpoints/ + -以下のように配置します： +Place checkpoints in: ``` 📂 ComfyUI/ @@ -95,17 +102,19 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 │ └── stable_audio_3_medium_base.safetensors ``` -### テキストエンコーダー +### Text encoders - - すべての Stable Audio 3 ワークフローで必要。models/text_encoders/ に配置 + + + Required for all Stable Audio 3 workflows. Place in models/text_encoders/ - - Medium ワークフローで必要（Qwen リプロンプト）。models/text_encoders/ に配置 + + Required for the Medium workflow (Qwen reprompt). Place in models/text_encoders/ + -以下のように配置します： +Place text encoders in: ``` 📂 ComfyUI/ @@ -115,4 +124,4 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 │ └── qwen3.5_2b_bf16.safetensors ``` -配置後、ComfyUI で **R** キーを押してノード定義をリフレッシュすると、最新のモデルが利用可能になります。 +After placing the files, press **R** in ComfyUI to refresh nodes and load the latest models. diff --git a/ja/tutorials/basic/image-to-image.mdx b/ja/tutorials/basic/image-to-image.mdx index d1d04333b..5fbb304ea 100644 --- a/ja/tutorials/basic/image-to-image.mdx +++ b/ja/tutorials/basic/image-to-image.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Try It Yourself": 6b2e3528 --- - ## 画像から画像へとは 「画像から画像へ（Image to Image）」は、ComfyUIにおけるワークフローの一種で、ユーザーが入力画像を指定し、その画像に基づいて新しい画像を生成することを可能にします。 diff --git a/ja/tutorials/basic/inpaint.mdx b/ja/tutorials/basic/inpaint.mdx index d97d65da2..96dc17058 100644 --- a/ja/tutorials/basic/inpaint.mdx +++ b/ja/tutorials/basic/inpaint.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "VAE Encoder (for Inpainting) Node": 222c8a2c --- - この記事では、AI画像生成における「局部再描画（インペインティング）」の概念を紹介し、ComfyUIでインペインティングワークフローを構築する手順を解説します。以下のような内容をカバーします： - インペインティングワークフローを用いた画像の編集方法 - ComfyUIのマスクエディタを用いたマスク作成方法 diff --git a/ja/tutorials/basic/lora.mdx b/ja/tutorials/basic/lora.mdx index c6e0fc6b1..05becec99 100644 --- a/ja/tutorials/basic/lora.mdx +++ b/ja/tutorials/basic/lora.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Try It Yourself": 9f276881 --- - **LoRA（Low-Rank Adaptation：低ランク適応）** は、Stable Diffusion などの大規模生成モデルを微調整するための効率的な手法です。 事前に学習済みのモデルに、学習可能な低ランク行列を導入することで、モデル全体の再学習ではなく、一部のパラメータのみを調整し、特定のタスクに対する最適化を、比較的低い計算コストで実現します。 SD1.5 などのベースモデルと比較して、LoRA モデルはファイルサイズが小さく、学習も容易です。 diff --git a/ja/tutorials/basic/multiple-loras.mdx b/ja/tutorials/basic/multiple-loras.mdx index 7e5e19e89..88ddeae12 100644 --- a/ja/tutorials/basic/multiple-loras.mdx +++ b/ja/tutorials/basic/multiple-loras.mdx @@ -33,7 +33,6 @@ translationFrom: tutorials/basic/multiple-loras.mdx 以下の画像をダウンロードし、**ComfyUI へドラッグ＆ドロップ**してワークフローを読み込んでください： ![ComfyUI ワークフロー：複数 LoRA](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/multiple_loras.png) - メタデータにワークフローの JSON を含む画像は、ComfyUI へ直接ドラッグ＆ドロップするか、メニューから `Workflows` → `Open (Ctrl+O)` を選択して読み込むことができます。 diff --git a/ja/tutorials/basic/outpaint.mdx b/ja/tutorials/basic/outpaint.mdx index 9a96f5004..741becf54 100644 --- a/ja/tutorials/basic/outpaint.mdx +++ b/ja/tutorials/basic/outpaint.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "ComfyUI Outpainting Workflow Example Explanation": e06f66e2 --- - このガイドでは、AI画像生成における拡張描画（Outpainting）の概念と、ComfyUIで拡張描画ワークフローを作成する方法について解説します。以下のような内容をカバーします： - 拡張描画ワークフローを用いた画像の拡張 - ComfyUIにおける拡張描画関連ノードの理解と活用 diff --git a/ja/tutorials/basic/text-to-image.mdx b/ja/tutorials/basic/text-to-image.mdx index 7e244fc5f..483390f52 100644 --- a/ja/tutorials/basic/text-to-image.mdx +++ b/ja/tutorials/basic/text-to-image.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Introduction to SD1.5 Model": cd91e138 --- - 本ガイドでは、ComfyUIにおけるテキストから画像へ生成するワークフローの基本的な仕組みを紹介し、さまざまなComfyUIノードの機能と使い方について理解を深めます。 本ドキュメントでは、以下の内容を学びます： diff --git a/ja/tutorials/basic/upscale.mdx b/ja/tutorials/basic/upscale.mdx index 2a21a2375..8468e285c 100644 --- a/ja/tutorials/basic/upscale.mdx +++ b/ja/tutorials/basic/upscale.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Additional Tips": 5352d8c0 --- - ## 画像アップスケールとは？ 画像アップスケール（Image Upscaling）とは、アルゴリズムを用いて低解像度画像を高解像度画像に変換するプロセスです。 @@ -54,7 +53,6 @@ translationBlockHashes: - ### ワークフローとアセット 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして、基本的なアップスケールワークフローを読み込みます： diff --git a/ja/tutorials/controlnet/controlnet.mdx b/ja/tutorials/controlnet/controlnet.mdx index 42acac85e..351c5b68a 100644 --- a/ja/tutorials/controlnet/controlnet.mdx +++ b/ja/tutorials/controlnet/controlnet.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Start Your Exploration": c5020328 --- - AI による画像生成において、画像生成を正確に制御することは、単にワンクリックで実現できるものではありません。 通常は、満足のいく画像を得るまで、多数の生成試行を繰り返す必要があります。しかし、**ControlNet** の登場により、この課題は効果的に解決されました。 diff --git a/ja/tutorials/controlnet/depth-controlnet.mdx b/ja/tutorials/controlnet/depth-controlnet.mdx index 19637bc3d..af7dff1de 100644 --- a/ja/tutorials/controlnet/depth-controlnet.mdx +++ b/ja/tutorials/controlnet/depth-controlnet.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Combining Depth Control with Other Techniques": 9e3bcc3d --- - ## 深度マップと Depth ControlNet の概要 深度マップ（Depth Map）は、シーン内の物体と観察者またはカメラとの距離をグレースケール値で表現する特殊な画像です。深度マップでは、グレースケール値が距離に反比例しており：明るい領域（白に近い）ほど物体が近く、暗い領域（黒に近い）ほど物体が遠くなります。 diff --git a/ja/tutorials/controlnet/depth-t2i-adapter.mdx b/ja/tutorials/controlnet/depth-t2i-adapter.mdx index f6ae26c4c..483b43c5e 100644 --- a/ja/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/ja/tutorials/controlnet/depth-t2i-adapter.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Characteristics of T2I Adapter Usage": d485c810 --- - ## T2I Adapter とは [T2I-Adapter](https://huggingface.co/TencentARC/T2I-Adapter) は、[Tencent ARC ラボ](https://github.com/TencentARC) が開発した軽量なアダプターであり、テキストから画像を生成するモデル（例：Stable Diffusion）の構造・色・スタイル制御能力を強化することを目的としています。 diff --git a/ja/tutorials/controlnet/mixing-controlnets.mdx b/ja/tutorials/controlnet/mixing-controlnets.mdx index a9990bd85..d2b669703 100644 --- a/ja/tutorials/controlnet/mixing-controlnets.mdx +++ b/ja/tutorials/controlnet/mixing-controlnets.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Multi-dimensional Control Applications for a Single Subject": 5790d753 --- - AI による画像生成において、単一の制御条件では複雑なシーンの要件を満たすことが困難な場合が多くあります。複数の ControlNet を混合することで、画像の異なる領域や異なる側面を同時に制御でき、より精密な画像生成制御が可能になります。 特定のシナリオでは、異なる制御条件の特性を活かした ControlNet の混合により、さらに細かい条件付き制御を実現できます： diff --git a/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx b/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx index aacef9020..9a032927a 100644 --- a/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/ja/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Advantages of 2-Pass Image Generation": fba5db79 --- - ## OpenPose について [OpenPose](https://github.com/CMU-Perceptual-Computing-Lab/openpose) は、カーネギーメロン大学（CMU）が開発したオープンソースのリアルタイム多人物ポーズ推定システムであり、コンピュータビジョン分野における重要な技術的ブレイクスルーです。このシステムは、1枚の画像内に複数の人物を同時に検出し、以下の情報を抽出できます： diff --git a/ja/tutorials/flux/flux-1-controlnet.mdx b/ja/tutorials/flux/flux-1-controlnet.mdx index 2b3dd526a..66e49a945 100644 --- a/ja/tutorials/flux/flux-1-controlnet.mdx +++ b/ja/tutorials/flux/flux-1-controlnet.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Community Versions of Flux Controlnets": 29728dcf --- - - ![Flux.1 Canny Controlnet](/images/tutorial/flux/flux-1-canny-controlnet.png) ![Flux.1 Depth Controlnet](/images/tutorial/flux/flux-1-depth-controlnet.png) @@ -48,18 +46,36 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -## FLUX.1-Canny-dev 完全版ワークフロー +### Flux.1 Canny Model (`flux_canny_model_example`) + +Generate images guided by edge detection using Flux.1 Canny. + +Flux.1 Canny ワークフロープレビュー - - Download JSON or search "Flux.1 Canny" in Template Library - Comfy Cloud で開く + + Download JSON or search "Flux.1 Canny" in Template Library + -### 1. ワークフローおよび関連アセット +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_canny_model_example_input_image.png` + + + +
+ flux_canny_model_example_input_image.png +
+ +**## 1. ワークフローおよび関連アセット 下記のワークフロー画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 @@ -119,18 +135,34 @@ ComfyUI/ - [ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet) - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -## FLUX.1-Depth-dev-lora ワークフロー +### Flux.1 Depth Lora (`flux_depth_lora_example`) + +Generate images guided by depth information using Flux.1 LoRA. + +Flux.1 Depth LoRA ワークフロープレビュー - - Download JSON or search "Flux.1 Depth LoRA" in Template Library - Comfy Cloud で開く + + Download JSON or search "Flux.1 Depth LoRA" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_depth_lora_example_input_image.png` + -LoRA 版ワークフローは、完全版ワークフローに LoRA モデルを追加したものであり、[Flux ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image) と比較して、対応する LoRA モデルを読み込むためのノードが追加されています。 +
+ flux_depth_lora_example_input_image.png +
### 1. ワークフローおよび関連アセット diff --git a/ja/tutorials/flux/flux-1-fill-dev.mdx b/ja/tutorials/flux/flux-1-fill-dev.mdx index 56b8f065b..f1d2e397e 100644 --- a/ja/tutorials/flux/flux-1-fill-dev.mdx +++ b/ja/tutorials/flux/flux-1-fill-dev.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Flux.1 Fill dev Outpainting Workflow": dd827926 --- - - ![Flux.1 fill dev](/images/tutorial/flux/flux-fill-dev-demo.jpeg) ## Flux.1 fill dev モデルの概要 @@ -60,36 +58,41 @@ ComfyUI/ ## Flux.1 Fill dev を用いた Inpainting ワークフロー -### 1. Inpainting ワークフローおよび関連アセット +### Flux.1 Inpaint (`flux_fill_inpaint_example`) + +Fill missing parts of images using Flux.1 Fill Inpainting. + +Flux.1 inpaint ワークフロープレビュー - - Download JSON or search "flux_fill_inpaint" in Template Library - Comfy Cloud で開く + + Download JSON or search "Flux.1 Inpaint" in Template Library + +**入力素材** + +Upload this file to the matching `LoadImage` node: + - - Download JSON or search "flux_fill_inpaint" in Template Library - - - Comfy Cloud で開く + + `LoadImage` node 17 · `flux_fill_inpaint_example_input_image.png` -以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 -![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) +
+ flux_fill_inpaint_example_input_image.png +
-以下の画像をダウンロードし、入力画像として使用します。 -![ComfyUI Flux.1 inpaint input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input.png) +**出力例** - -この画像には既にアルファチャンネルが含まれているため、別途マスクを描画する必要はありません。 -独自のマスクを作成したい場合は、[こちら](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input_original.png)からマスクなしの画像を取得し、[ComfyUI レイアウト Inpainting の使用例](/ja/tutorials/basic/inpaint#using-the-mask-editor) の「MaskEditor の使用方法」セクションを参照して、`Load Image` ノード内でマスクを描画する方法を学んでください。 - +
+ 入力画像 + Flux.1 inpaint 出力例 +
### 2. ワークフローの実行手順 diff --git a/ja/tutorials/flux/flux-1-kontext-dev.mdx b/ja/tutorials/flux/flux-1-kontext-dev.mdx index d09d96fbb..d1ff8ef69 100644 --- a/ja/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ja/tutorials/flux/flux-1-kontext-dev.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Flux.1 Kontext Dev Workflow": ffae9c25 --- - - import PromptTechniques from "/snippets/ja/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -79,18 +77,43 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： │ └── t5xxl_fp16.safetensors または t5xxl_fp8_e4m3fn_scaled.safetensors ``` -## Flux.1 Kontext Dev ワークフロー +### Flux Kontext Dev Image Edit (`flux_kontext_dev_basic`) + +Smart image editing that keeps characters consistent, edits specific parts without affecting others, and preserves original styles. + +Flux Kontext Dev ワークフロープレビュー - - Download JSON or search "Flux Kontext Dev" in Template Library - Comfy Cloud で開く + + Download JSON or search "Flux Kontext Dev" in Template Library + -このワークフローでは、編集対象の画像を読み込むために `Load Image(from output)` ノードを採用しており、編集後の画像を容易に取得・再利用できるため、複数回の反復編集がよりスムーズに行えます。 +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 190 · `flux_kontext_dev_basic_input_image.jpg` + + + +
+ flux_kontext_dev_basic_input_image.jpg +
+ +**出力例** + +
+ 入力画像 + Flux Kontext Dev 出力例 +
+ +This workflow uses the `Load Image(from output)` node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing. ### 1. ワークフローおよび入力画像のダウンロード diff --git a/ja/tutorials/flux/flux-1-text-to-image.mdx b/ja/tutorials/flux/flux-1-text-to-image.mdx index 555c78bb9..35e896838 100644 --- a/ja/tutorials/flux/flux-1-text-to-image.mdx +++ b/ja/tutorials/flux/flux-1-text-to-image.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Flux.1 FP8 Checkpoint Version Text-to-Image Example": daf5c58e --- - - ![Flux](/images/tutorial/flux/flux_example.png) Flux は、現時点で最も大規模なオープンソースのテキストから画像へ生成するモデルの一つであり、120億（12B）パラメータを有し、オリジナルファイルサイズは約23GBです。このモデルは、元 Stable Diffusion チームのメンバーによって設立された [Black Forest Labs](https://blackforestlabs.ai/) が開発しました。 Flux は、優れた画像品質と高い柔軟性で知られており、高品質かつ多様な画像を生成できます。 @@ -47,36 +45,41 @@ Flux は、優れた画像品質と高い柔軟性で知られており、高品 ![Flux Agreement](/images/tutorial/flux/flux_agreement.jpg) -### Flux.1 Dev +### Flux.1 Dev fp8: Text to Image (`flux_dev_checkpoint_example`) -#### 1. ワークフロー・ファイル +Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices with limited VRAM, requires only one model file, but image quality is slightly reduced compared to the full version. -下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 -![Flux Dev オリジナルバージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) +Flux.1 Dev fp8 ワークフロープレビュー Comfy Cloud でこのワークフローを実行 - Download JSON or search "Flux.1 Dev" in Template Library + Download JSON or search "Flux.1 Dev fp8" in Template Library -#### 2. モデルの手動インストール +**出力例** + +![Flux.1 Dev fp8 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_checkpoint_example.png) + +#### 1. Workflow File + +#### 2. Manual Model Installation -- `flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) のライセンス契約に同意する必要があります。 -- VRAM が少ない環境では、`t5xxl_fp16.safetensors` の代わりに [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用することを検討してください。 +- The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) agreement before downloading via browser. +- If your VRAM is low, you can try using [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) to replace the `t5xxl_fp16.safetensors` file. -以下のモデルファイルをダウンロードしてください： +Please download the following model files: - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) （VRAM が32GBを超える環境では推奨） +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) - [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) -保存先のディレクトリ構成： +Storage location: ``` ComfyUI/ ├── models/ @@ -89,106 +92,73 @@ ComfyUI/ │ └── flux1-dev.safetensors ``` -#### 3. ワークフロー実行手順 +#### 3. Steps to Run the Workflow -以下の画像を参照し、すべてのモデルファイルが正しく読み込まれていることを確認してください。 +Please refer to the image below to ensure all model files are loaded correctly -![ComfyUI Flux Dev ワークフロー](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) +![ComfyUI Flux Dev Workflow](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) -1. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認してください： +1. Ensure the `DualCLIPLoader` node has the following models loaded: - clip_name1: t5xxl_fp16.safetensors - clip_name2: clip_l.safetensors -2. `Load Diffusion Model` ノードに `flux1-dev.safetensors` が読み込まれていることを確認してください -3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認してください -4. `Queue` ボタンをクリックするか、ショートカット `Ctrl（Cmd） + Enter` を押してワークフローを実行してください +2. Ensure the `Load Diffusion Model` node has `flux1-dev.safetensors` loaded +3. Make sure the `Load VAE` node has `ae.safetensors` loaded +4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -Flux の優れたプロンプト追従能力により、負のプロンプト（ネガティブプロンプト）は不要です。 +Thanks to Flux's excellent prompt following capability, we don't need any negative prompts -### Flux.1 Schnell -#### 1. ワークフロー・ファイル +### Flux.1 Schnell FP8 (`flux_schnell`) -下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 +Quickly generate images with Flux.1 Schnell fp8 quantized version. Ideal for low-end hardware, requires only 4 steps to generate images. -![Flux Schnell バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) +Flux.1 Schnell FP8 checkpoint ワークフロープレビュー Comfy Cloud でこのワークフローを実行 - - Download JSON or search "Flux.1 Schnell" in Template Library + + Download JSON or search "Flux.1 Schnell FP8" in Template Library -#### 2. モデルの手動インストール - - -本ワークフローでは、Flux1 Dev バージョンのワークフローと異なるモデルファイルは2つだけです。t5xxl については、より良い結果を得るために引き続き fp16 バージョンを使用できます。 -- **t5xxl_fp16.safetensors** → **t5xxl_fp8.safetensors** -- **flux1-dev.safetensors** → **flux1-schnell.safetensors** - - -完全なモデルファイル一覧： -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) - -ファイルの保存先ディレクトリ構成： -``` -ComfyUI/ -├── models/ -│ ├── text_encoders/ -│ │ ├── clip_l.safetensors -│ │ └── t5xxl_fp8_e4m3fn.safetensors -│ ├── vae/ -│ │ └── ae.safetensors -│ └── diffusion_models/ -│ └── flux1-schnell.safetensors -``` - -#### 3. ワークフロー実行手順 +Please download the image below and drag it into ComfyUI to load the workflow. -![Flux Schnell バージョンワークフロー](/images/tutorial/flux/flow_diagram_flux_schnell_t5fp8.jpg) +Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. -1. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認してください： - - clip_name1: t5xxl_fp8_e4m3fn.safetensors - - clip_name2: clip_l.safetensors -2. `Load Diffusion Model` ノードに `flux1-schnell.safetensors` が読み込まれていることを確認してください -3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認してください -4. `Queue` ボタンをクリックするか、ショートカット `Ctrl（Cmd） + Enter` を押してワークフローを実行してください +Ensure that the corresponding `Load Checkpoint` node loads `flux1-schnell-fp8.safetensors`, and you can try to run the workflow. ## Flux.1 FP8 Checkpoint バージョンによるテキストから画像へ生成の例 FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したものです。 ある程度、このバージョンの品質は fp16 バージョンよりも劣りますが、その一方で必要な VRAM 量が少なくなり、試行運用のためにインストールするモデルファイルは1つだけで済みます。 -### Flux.1 Dev +### Flux.1 Dev: Text to Image (`flux_dev_full_text_to_image`) -下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 +Generate high-quality images with Flux Dev full version. Requires larger VRAM and multiple model files, but provides the best prompt following capability. -![Flux Dev fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) +Flux.1 Dev text-to-image ワークフロープレビュー Comfy Cloud でこのワークフローを実行 - Download JSON or search "Flux.1 Dev FP8" in Template Library - - - Comfy Cloud でこのワークフローを実行 - - - Download JSON or search "Flux.1 Schnell FP8" in Template Library + Download JSON or search "Flux.1 Dev FP8" in Template Library -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 +**出力例** + +![Flux.1 Dev FP8 checkpoint 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_full_text_to_image.png) + +Please download the image below and drag it into ComfyUI to load the workflow. + +Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. -対応する `Load Checkpoint` ノードが `flux1-dev-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 +Ensure that the corresponding `Load Checkpoint` node loads `flux1-dev-fp8.safetensors`, and you can try to run the workflow. ### Flux.1 Schnell diff --git a/ja/tutorials/flux/flux-1-uso.mdx b/ja/tutorials/flux/flux-1-uso.mdx index f5de85629..d4848a85b 100644 --- a/ja/tutorials/flux/flux-1-uso.mdx +++ b/ja/tutorials/flux/flux-1-uso.mdx @@ -25,33 +25,41 @@ USO は以下の3つの主要なアプローチをサポートします： -### 1. ワークフローと入力 +### Flux.1 Dev USO Reference Image Generation (`flux1_dev_uso_reference_image_gen`) -以下の画像をダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込みます。 +Use reference images to control both style and subject. Keep your character's face while changing artistic style, or apply artistic styles to new scenes. -![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - - -

JSON ワークフローをダウンロード

- +Flux.1 Dev USO reference image ワークフロープレビュー - - Download the workflow JSON and drag it into ComfyUI - Comfy Cloud でこのワークフローを実行 + + Download the workflow JSON and drag it into ComfyUI + -以下の画像を入力画像として使用します。 +**入力素材** -![入力](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/input.png) +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 47 · `flux1_dev_uso_reference_image_gen_input_image.png` + + + +
+ flux1_dev_uso_reference_image_gen_input_image.png +
+ +**出力例** + +
+ 入力画像 + Flux.1 Dev USO 出力例 +
### 2. モデルのダウンロードリンク @@ -85,7 +93,6 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', │ │ └── sigclip_vision_patch14_384.safetensors ``` - ### 3. ワークフローの操作手順 ![ワークフローの操作手順](/images/tutorial/flux/flux1_uso_reference_image_gen.jpg) diff --git a/ja/tutorials/flux/flux-2-dev.mdx b/ja/tutorials/flux/flux-2-dev.mdx index f0c32e103..2ea935dd8 100644 --- a/ja/tutorials/flux/flux-2-dev.mdx +++ b/ja/tutorials/flux/flux-2-dev.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Model links": f4c1677b --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' diff --git a/ja/tutorials/flux/flux-2-klein.mdx b/ja/tutorials/flux/flux-2-klein.mdx index e360d1c0d..73197ea0c 100644 --- a/ja/tutorials/flux/flux-2-klein.mdx +++ b/ja/tutorials/flux/flux-2-klein.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Flux.2 Klein 9B Model Downloads": 0f38ad48 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' diff --git a/ja/tutorials/flux/flux1-krea-dev.mdx b/ja/tutorials/flux/flux1-krea-dev.mdx index 8acbbb818..bdc7e94ce 100644 --- a/ja/tutorials/flux/flux1-krea-dev.mdx +++ b/ja/tutorials/flux/flux1-krea-dev.mdx @@ -22,88 +22,81 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **モデルのライセンス** 本モデルは、[flux-1-dev-non-commercial-license](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/LICENSE.md) の下で公開されています。 -## Flux.1 Krea Dev ComfyUI ワークフロー +### Flux.1 Krea Dev (`flux1_krea_dev`) - +A fine-tuned FLUX model pushing photorealism to the max. -#### 1. ワークフロー ファイル - -以下の画像または JSON ファイルをダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 -![Flux Krea Dev ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) +Flux.1 Krea Dev ワークフロープレビュー Comfy Cloud でこのワークフローを実行 - Download JSON or search "Flux.1 Krea Dev" in Template Library + Download JSON or search "Flux.1 Krea Dev" in Template Library - - - Comfy Cloud でこのワークフローを実行 - - - Download JSON or search "Flux.1 Krea Dev" in Template Library - - +**出力例** -#### 2. 手動によるモデルのインストール +![Flux.1 Krea Dev 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux1_krea_dev.png) -以下のモデルファイルをダウンロードしてください： +#### 1. Workflow Files -**Diffusion モデル** -以下のいずれか 1 つのバージョンを選択してください： +#### 2. Manual Model Installation + +Please download the following model files: +**Diffusion model** - [flux1-krea-dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/FLUX.1-Krea-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-krea-dev_fp8_scaled.safetensors) -より高品質な出力を求め、かつ十分な VRAM をお持ちの場合、オリジナルの重みファイルも試すことができます： +If you want to pursue higher quality and have enough VRAM, you can try the original model weights - [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) -`flux1-dev.safetensors` ファイルは、ブラウザからダウンロードする前に、[black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) の利用規約に同意する必要があります。 +The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) agreement before downloading via browser. -以前に Flux 関連のワークフローをご利用済みの場合、以下のモデルは既に存在するため、再ダウンロードの必要はありません。 +If you have used Flux related workflows before, the following models are the same and don't need to be downloaded again -**テキストエンコーダー** -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) — VRAM が 32GB を超える環境で推奨 -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) — 低 VRAM 環境向け +**Text encoders** +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM -**VAE** -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +**VAE** +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -ファイルの保存先： +File save location: ``` ComfyUI/ ├── models/ │ ├── diffusion_models/ -│ │ └── flux1-krea-dev_fp8_scaled.safetensors または flux1-krea-dev.safetensors +│ │ └── flux1-krea-dev_fp8_scaled.safetensors or flux1-krea-dev.safetensors │ ├── text_encoders/ │ │ ├── clip_l.safetensors -│ │ └── t5xxl_fp16.safetensors または t5xxl_fp8_e4m3fn.safetensors +│ │ └── t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors │ ├── vae/ │ │ └── ae.safetensors + ``` -#### 3. ワークフローが正しく動作することを確認する手順 +#### 3. Step-by-step Verification to Ensure Workflow Runs Properly -低 VRAM 環境では、このモデルがスムーズに動作しない可能性があります。FP8 や GGUF 形式のバージョンがコミュニティから提供されるのをお待ちください。 + For low VRAM users, this model may not run smoothly on your device, you can wait for the community to provide FP8 or GGUF version. -以下の画像を参考に、すべてのモデルファイルが正しく読み込まれていることをご確認ください。 +Please refer to the image below to ensure all model files have been loaded correctly -![ComfyUI Flux Krea Dev ワークフロー](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) +![ComfyUI Flux Krea Dev Workflow](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) -1. `Load Diffusion Model` ノードに `flux1-krea-dev_fp8_scaled.safetensors` または `flux1-krea-dev.safetensors` のいずれかが読み込まれていることを確認してください。 -　　- 低 VRAM 環境には `flux1-krea-dev_fp8_scaled.safetensors` の使用を推奨します。 -　　- `flux1-krea-dev.safetensors` はオリジナルの重みであり、24GB などの十分な VRAM をお持ちの場合は、より高品質な結果を得るためにご利用いただけます。 -2. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認してください： -　　- `clip_name1`: `t5xxl_fp16.safetensors` または `t5xxl_fp8_e4m3fn.safetensors` -　　- `clip_name2`: `clip_l.safetensors` -3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認してください。 -4. `Queue` ボタンをクリックするか、ショートカットキー `Ctrl`（macOS の場合は `Cmd`）＋`Enter` を押して、ワークフローを実行してください。 \ No newline at end of file +1. Ensure that `flux1-krea-dev_fp8_scaled.safetensors` or `flux1-krea-dev.safetensors` is loaded in the `Load Diffusion Model` node + - `flux1-krea-dev_fp8_scaled.safetensors` is recommended for low VRAM users + - `flux1-krea-dev.safetensors` is the original weights, if you have enough VRAM like 24GB you can use it for better quality +2. Ensure the following models are loaded in the `DualCLIPLoader` node: + - clip_name1: t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors + - clip_name2: clip_l.safetensors +3. Ensure that `ae.safetensors` is loaded in the `Load VAE` node +4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow diff --git a/ja/tutorials/image/anima/anima.mdx b/ja/tutorials/image/anima/anima.mdx index e78fb3dca..15521cb70 100644 --- a/ja/tutorials/image/anima/anima.mdx +++ b/ja/tutorials/image/anima/anima.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Limitations": 0adbf360 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Anima** は [CircleStone Labs](https://huggingface.co/circlestone-labs/Anima) が Comfy Org と協力して開発したオープンなテキストから画像生成モデルです。**20 億** パラメータを備え、高品質な **アニメおよび非フォトリアリスティック** な画像を生成するように設計されており、キャラクター、シーン、コンセプトアートに最適です。 @@ -36,47 +35,57 @@ Anima は 2 つのワークフローを提供しています——標準的な このワークフローは Subgraph ノードを使用してモジュール化された処理を実現しています。Subgraph のドキュメントを参照して、ワークフローのカスタマイズと拡張方法を学んでください。 -### Anima Base v1：テキストから画像へ +### Anima Base v1: Text to Image (`image_anima_base_v1`) - - JSON をダウンロードするか、テンプレートライブラリで "Anima Base v1" を検索 - +Input a text prompt describing an anime or artistic illustration. Generate a non-photorealistic image focused on anime concepts, characters, or styles. - +Anima Base v1 text-to-image ワークフロープレビュー + + + Comfy Cloud で開く + + Download JSON or search "Anima Base v1" in Template Library + + -#### はじめに +**出力例** -1. ComfyUI を最新バージョンに更新するか、[Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) を使用してください -2. **テンプレート** に移動し、**Anima Base v1** を検索します -3. **Anima Base v1: Text to Image** ワークフローを選択します -4. 不足しているモデルをダウンロードし（[モデルのダウンロード](#anima-モデルのダウンロード)を参照）、プロンプトを入力して **実行** をクリックします +![Anima Base v1 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_base_v1.png) -#### サンプル出力 +#### Get started -Anima Base v1 サンプル出力 +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) +2. Go to **Template** and search for **Anima Base v1** +3. Select the **Anima Base v1: Text to Image** workflow +4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** -### Anima Preview：アニメテキストから画像生成 +### Anima Preview: Anime Text-to-Image Generation (`image_anima_preview`) - - JSON をダウンロードするか、テンプレートライブラリで "Anima Preview" を検索 - +Input a text prompt to generate an anime-style image using the Anima model. Configure settings like steps and CFG scale to control the output. + +Anima Preview text-to-image ワークフロープレビュー - + + Comfy Cloud で開く + + Download JSON or search "Anima Preview" in Template Library + + -#### はじめに +**出力例** -1. ComfyUI を最新バージョンに更新するか、[Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) を使用してください -2. **テンプレート** に移動し、**Anima Preview** を検索します -3. **Anima Anime Text-to-Image Generation** ワークフローを選択します -4. 不足しているモデルをダウンロードし（[モデルのダウンロード](#anima-モデルのダウンロード)を参照）、プロンプトを入力して **実行** をクリックします +![Anima Preview 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_preview.png) -#### サンプル出力 +#### Get started -Anima Preview サンプル出力 +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) +2. Go to **Template** and search for **Anima Preview** +3. Select the **Anima Anime Text-to-Image Generation** workflow +4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** ## Anima モデルのダウンロード diff --git a/ja/tutorials/image/boogu/boogu-image-0.1.mdx b/ja/tutorials/image/boogu/boogu-image-0.1.mdx index 7b75c26ed..5d0e1b811 100644 --- a/ja/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ja/tutorials/image/boogu/boogu-image-0.1.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Boogu-Image-0.1-Edit image editing workflow": 3c1752bb --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" **Boogu-Image-0.1** は Apache-2.0 オープンソースの統合画像生成・編集モデルファミリーです。理解と生成を統合するシステムにより、写真、テキストレンダリング、スタイライゼーション、画像編集タスクで競争力のあるパフォーマンスを発揮します。 diff --git a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index 7b1f7a47c..c9f7a17f2 100644 --- a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -36,7 +36,6 @@ Hugging Face: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/cosmos/predict2/cosmos_predict2_2B_t2i.png) - ### 2. 手動によるモデルインストール モデルの自動ダウンロードが失敗した場合、以下の手順で手動でダウンロードできます。 @@ -55,7 +54,6 @@ Hugging Face: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - ファイルの保存場所 ``` 📂 ComfyUI/ diff --git a/ja/tutorials/image/ernie-image/ernie-image.mdx b/ja/tutorials/image/ernie-image/ernie-image.mdx index b02fa0f45..86667f861 100644 --- a/ja/tutorials/image/ernie-image/ernie-image.mdx +++ b/ja/tutorials/image/ernie-image/ernie-image.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Examples": 0e2eb115 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **ERNIE-Image** は百度が開発したオープンなテキストから画像生成モデルで、Apache-2.0 ライセンスで公開されています。**8B** パラメータの拡散トランスフォーマー（DiT）をベースに構築されており、精密なテキストレンダリング、高い命令追従性、構造化された視覚生成を実現する高品質な画像生成が可能です。 @@ -31,18 +30,28 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - [Hugging Face（ERNIE-Image）](https://huggingface.co/Baidu/ERNIE-Image) - [Hugging Face（ComfyUI サポート）](https://huggingface.co/Comfy-Org/ERNIE-Image) +## ERNIE-Image テキストから画像生成ワークフロー -## ERNIE-Image テキストから画像へのワークフロー + - - ERNIE-Image テキストから画像へのワークフロー JSON ファイルをダウンロードします。 - +### Ernie Image: Text to Image (`image_ernie_image`) + +Generate images from text prompts using the ERNIE-Image model. Input a text description to produce detailed, structured visuals with a broad stylistic range. + +ERNIE-Image text-to-image ワークフロープレビュー - - このワークフローを Comfy Cloud 上で直接実行します。 + + + Comfy Cloud でこのワークフローを直接実行 + + Download the ERNIE-Image text-to-image workflow JSON file + + - +**出力例** + +![ERNIE-Image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image.png) ### はじめに @@ -55,21 +64,20 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' リパッケージされたすべてのモデルファイルは、Hugging Face の [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) で入手できます。 - + + ERNIE-Image 用拡散モデル。 - - + ERNIE-Image 用テキストエンコーダー。 - - + ERNIE-Image 用プロンプトエンハンサーテキストエンコーダー。 - - + ERNIE-Image 用 VAE。 + **モデルの保存場所** @@ -87,33 +95,43 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## ERNIE-Image-Turbo -[ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo) は DMD と RL で最適化された高速バリアントで、標準モデルに必要な約 50 ステップに対し、わずか **8 ステップ** で画像を生成します。 +[ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo) は DMD と RL で最適化された高速バリアントで、標準モデルの約 50 ステップに対し、わずか **8 ステップ**で画像を生成します。 - - ERNIE-Image-Turbo テキストから画像へのワークフロー JSON ファイルをダウンロードします。 - +### Ernie Image Turbo: Text To Image (`image_ernie_image_turbo`) + +Generate images from text prompts using the ERNIE-Image turbo model. Input a text description and receive a high-quality image with precise text rendering. - - このワークフローを Comfy Cloud 上で直接実行します。 +ERNIE-Image-Turbo text-to-image ワークフロープレビュー + + + + Comfy Cloud でこのワークフローを直接実行 + + + Download the ERNIE-Image-Turbo text-to-image workflow JSON file + + +**出力例** + +![ERNIE-Image-Turbo 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image_turbo.png) ### ERNIE-Image-Turbo モデルのダウンロード - + + ERNIE-Image-Turbo 用拡散モデル。 - - + ERNIE-Image-Turbo 用テキストエンコーダー。 - - + ERNIE-Image-Turbo 用プロンプトエンハンサーテキストエンコーダー。 - - + ERNIE-Image-Turbo 用 VAE。 + **モデルの保存場所** diff --git a/ja/tutorials/image/hidream/hidream-e1.mdx b/ja/tutorials/image/hidream/hidream-e1.mdx index bd9a0ee43..4f4ad1432 100644 --- a/ja/tutorials/image/hidream/hidream-e1.mdx +++ b/ja/tutorials/image/hidream/hidream-e1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ![HiDream-E1 デモ](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -72,19 +70,36 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン │ └── hidream_e1_full_bf16.safetensors ``` - ## HiDream E1.1 の ComfyUI ネイティブ ワークフローの例 -E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メガピクセル解像度** をサポートしています。ワークフローでは `Scale Image to Total Pixels` ノードを用いて、入力画像を自動的に 100万ピクセルにスケーリングします。 +### HiDream E1.1 Image Editing (`hidream_e1_1`) - -テスト時の VRAM 使用量の参考値は以下の通りです： -1. A100 40GB（VRAM 使用率 95％）：初回生成 211秒、2回目以降の生成 73秒 +Edit images with HiDream E1.1. Superior image quality and editing accuracy compared to HiDream-E1-Full. -2. RTX 4090D 24GB（VRAM 使用率 98％） -　- Full バージョン：メモリ不足（Out of memory） -　- FP8_e4m3fn_fast（VRAM 使用率 98％）：初回生成 120秒、2回目以降の生成 91秒 - +HiDream E1.1 image editing ワークフロープレビュー + + + + Comfy Cloud で開く + + + Download JSON or search "HiDream E1.1" in Template Library + + + +**入力素材** + +このファイルを一致する `LoadImage` ノードにアップロード： + + + + `LoadImage` node 13 · `hidream_e1_1_input_image.jpg` + + + +
+ hidream_e1_1_input_image.jpg +
### 1. HiDream E1.1 ワークフローおよび関連素材 @@ -116,18 +131,44 @@ E1.1 は 2025年7月16日にリリースされた更新版で、**動的な 1メ - HiDream E1.1 は「合計ピクセル数が 100万」の動的入力をサポートするため、ワークフローでは `Scale Image to Total Pixels` ノードを用いてすべての入力画像を処理・変換します。このため、元の入力画像と比較してアスペクト比が変化する場合があります。 - fp16 版モデルを使用する場合、A100 40GB および RTX 4090D 24GB における実際のテストでは、Full バージョンでメモリ不足（Out of memory）が発生しました。そのため、ワークフローはデフォルトで `fp8_e4m3fn_fast` を推論に使用するよう設定されています。 - ## HiDream E1 の ComfyUI ネイティブ ワークフローの例 +### HiDream E1 Image Edit (`hidream_e1_full`) + +Edit images with HiDream E1. Professional natural language image editing model. + +HiDream E1 image editing ワークフロープレビュー + - + Comfy Cloud で開く - Download JSON or search "HiDream E1 Full" in Template Library + Download JSON or search "HiDream E1 Full" in Template Library +**入力素材** + +このファイルを一致する `LoadImage` ノードにアップロード： + + + + `LoadImage` node 13 · `hidream_e1_full_input_image.jpg` + + + +
+ hidream_e1_full_input_image.jpg +
+ +**出力例** + +
+ 入力画像 + HiDream E1 出力例 +
+ E1 は 2025年4月28日にリリースされたモデルで、**768×768 の固定解像度のみ** をサポートします。 diff --git a/ja/tutorials/image/hidream/hidream-i1.mdx b/ja/tutorials/image/hidream/hidream-i1.mdx index 4d78b1d4d..a63e8d9bb 100644 --- a/ja/tutorials/image/hidream/hidream-i1.mdx +++ b/ja/tutorials/image/hidream/hidream-i1.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Other Related Resources": e794d9ef --- - - ![HiDream-I1 デモ](https://raw.githubusercontent.com/HiDream-ai/HiDream-I1/main/assets/demo.jpg) HiDream-I1 は、HiDream-ai 社が 2025 年 4 月 7 日に公式にオープンソース化したテキストから画像へ変換するモデルです。このモデルは 17B（170 億）パラメータを有し、[MIT ライセンス](https://github.com/HiDream-ai/HiDream-I1/blob/main/LICENSE) の下で公開されており、個人プロジェクト、学術研究、商用利用のすべてに対応しています。現在、複数のベンチマークテストにおいて優れた性能を発揮しています。 @@ -100,134 +98,155 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ │ └── ... # 対応するバージョンのワークフローでインストール方法をご案内します ``` -### HiDream-I1 Full バージョンのワークフロー +### HiDream I1 Full (`hidream_i1_full`) + +Generate images with HiDream I1 Full. Complete version with 50 inference steps for highest quality output. + +HiDream I1 Full ワークフロープレビュー - セットアップ不要で Comfy Cloud で実行 + Comfy Cloud でセットアップ不要で実行 - ワークフロー JSON ファイルをダウンロード + Download the workflow JSON file -#### 1. モデルファイルのダウンロード +**出力例** + +![HiDream I1 Full 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_full.png) + +#### 1. Model File Download -ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 +Please select the appropriate version based on your hardware. Click the link and download the corresponding model file to save it to the `ComfyUI/models/diffusion_models/` folder. -- FP8 バージョン：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 version: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) requires more than 27GB of VRAM -#### 2. ワークフローファイルのダウンロード +#### 2. Workflow File Download -以下の画像をダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 -![HiDream-I1 Full バージョンのワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) +Please download the image below and drag it into ComfyUI to load the corresponding workflow +![HiDream-I1 Full Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) -#### 3. ワークフローの実行手順（ステップバイステップ） +#### 3. Complete the Workflow Step by Step -![HiDream-I1 Full バージョンのフローダイアグラム](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) +![HiDream-I1 Full Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) -ワークフローの実行を以下の手順で完了してください： -1. `Load Diffusion Model` ノードが `hidream_i1_full_fp8.safetensors` ファイルを使用していることを確認してください。 -2. `QuadrupleCLIPLoader` 内の 4 つのテキストエンコーダが正しく読み込まれていることを確認してください： +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_full_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` ノードが `ae.safetensors` ファイルを使用していることを確認してください。 -4. **Full** バージョンでは、`ModelSamplingSD3` ノードの `shift` パラメータを `3.0` に設定してください。 -5. `Ksampler` ノードでは、以下の設定を行ってください： - - `steps` を `50` に設定 - - `cfg` を `5.0` に設定 - - （任意）`sampler` を `lcm` に設定 - - （任意）`scheduler` を `normal` に設定 -6. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（Cmd）+ Enter` を押して、画像生成を実行してください。 +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **full** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `50` + - Set `cfg` to `5.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation -### HiDream-I1 Dev バージョンのワークフロー +### HiDream I1 Dev (`hidream_i1_dev`) + +Generate images with HiDream I1 Dev. Balanced version with 28 inference steps, suitable for medium-range hardware. + +HiDream I1 Dev ワークフロープレビュー - セットアップ不要で Comfy Cloud で実行 + Comfy Cloud でセットアップ不要で実行 - ワークフロー JSON ファイルをダウンロード + Download the workflow JSON file -#### 1. モデルファイルのダウンロード +**出力例** -ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 +![HiDream I1 Dev 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_dev.png) -- FP8 バージョン：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +#### 1. Model File Download +Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -#### 2. ワークフローファイルのダウンロード +- FP8 version: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) requires more than 27GB of VRAM -以下の画像をダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 -![HiDream-I1 Dev バージョンのワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) +#### 2. Workflow File Download +Please download the image below and drag it into ComfyUI to load the corresponding workflow -#### 3. ワークフローの実行手順（ステップバイステップ） +![HiDream-I1 Dev Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) -![HiDream-I1 Dev バージョンのフローダイアグラム](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) +#### 3. Complete the Workflow Step by Step -ワークフローの実行を以下の手順で完了してください： -1. `Load Diffusion Model` ノードが `hidream_i1_dev_fp8.safetensors` ファイルを使用していることを確認してください。 -2. `QuadrupleCLIPLoader` 内の 4 つのテキストエンコーダが正しく読み込まれていることを確認してください： +![HiDream-I1 Dev Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_dev_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` ノードが `ae.safetensors` ファイルを使用していることを確認してください。 -4. **Dev** バージョンでは、`ModelSamplingSD3` ノードの `shift` パラメータを `6.0` に設定してください。 -5. `Ksampler` ノードでは、以下の設定を行ってください： - - `steps` を `28` に設定 - - （重要）`cfg` を `1.0` に設定 - - （任意）`sampler` を `lcm` に設定 - - （任意）`scheduler` を `normal` に設定 -6. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（Cmd）+ Enter` を押して、画像生成を実行してください。 +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **dev** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `6.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `28` + - (Important) Set `cfg` to `1.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation + +### HiDream I1 Fast (`hidream_i1_fast`) -### HiDream-I1 Fast バージョンのワークフロー +Generate images quickly with HiDream I1 Fast. Lightweight version with 16 inference steps, ideal for rapid previews on lower-end hardware. + +HiDream I1 Fast ワークフロープレビュー - セットアップ不要で Comfy Cloud で実行 + Comfy Cloud でセットアップ不要で実行 - ワークフロー JSON ファイルをダウンロード + Download the workflow JSON file -#### 1. モデルファイルのダウンロード +**出力例** + +![HiDream I1 Fast 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_fast.png) -ハードウェア環境に応じて適切なバージョンを選択し、対応するリンクをクリックして、`ComfyUI/models/diffusion_models/` フォルダにダウンロードしてください。 +#### 1. Model File Download +Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -- FP8 バージョン：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（VRAM 16GB 以上が必要） -- 完全版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（VRAM 27GB 以上が必要） +- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM -#### 2. ワークフローファイルのダウンロード +#### 2. Workflow File Download +Please download the image below and drag it into ComfyUI to load the corresponding workflow -以下の画像をダウンロードし、ComfyUI にドラッグ＆ドロップして、対応するワークフローを読み込んでください。 -![HiDream-I1 Fast バージョンのワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) +![HiDream-I1 Fast Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) -#### 3. ワークフローの実行手順（ステップバイステップ） +#### 3. Complete the Workflow Step by Step -![HiDream-I1 Fast バージョンのフローダイアグラム](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) +![HiDream-I1 Fast Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) -ワークフローの実行を以下の手順で完了してください： -1. `Load Diffusion Model` ノードが `hidream_i1_fast_fp8.safetensors` ファイルを使用していることを確認してください。 -2. `QuadrupleCLIPLoader` 内の 4 つのテキストエンコーダが正しく読み込まれていることを確認してください： +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_fast_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` ノードが `ae.safetensors` ファイルを使用していることを確認してください。 -4. **Fast** バージョンでは、`ModelSamplingSD3` ノードの `shift` パラメータを `3.0` に設定してください。 -5. `Ksampler` ノードでは、以下の設定を行ってください： - - `steps` を `16` に設定 - - （重要）`cfg` を `1.0` に設定 - - （任意）`sampler` を `lcm` に設定 - - （任意）`scheduler` を `normal` に設定 -6. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（Cmd）+ Enter` を押して、画像生成を実行してください。 +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **fast** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `16` + - (Important) Set `cfg` to `1.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation ## その他の関連リソース diff --git a/ja/tutorials/image/hidream/hidream-o1.mdx b/ja/tutorials/image/hidream/hidream-o1.mdx index 6295428f3..70f06e710 100644 --- a/ja/tutorials/image/hidream/hidream-o1.mdx +++ b/ja/tutorials/image/hidream/hidream-o1.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Additional Notes": 762758eb --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -33,50 +32,24 @@ HiDream-O1-Image は [MIT ライセンス](https://github.com/HiDream-ai/HiDream ## HiDream-O1-Image Full ワークフロー -### 1. ワークフローをダウンロード +### HiDream O1 Full: Image generation (`image_hidream_o1`) -ComfyUI を最新バージョンにアップデートし、メニューから `ワークフロー` → `テンプレートを閲覧` → `Image` に進み、"HiDream O1 Full: Image generation" を見つけてワークフローを読み込んでください。 +Input a text prompt and optionally upload reference images. Generate a high-resolution image up to 2048x2048 with text-to-image, editing, or subject-driven personalization. -![HiDream-O1-Image Full ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) +HiDream O1 Full ワークフロープレビュー - - Download workflow + + + Comfy Cloud で開く - - - Open in cloud + + Download JSON or search "HiDream O1 Full" in Template Library + -### 2. モデルをダウンロード - -**チェックポイント** — 再パッケージおよび量子化済み。すべてのバージョンで最悪の外れ値は bf16 で保持され、未使用の deepstack 層は削除されています： - -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量子化版 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 フル精度版（最大ファイル） - -**テキストエンコーダ（プロンプト補強）** — 全バージョン共通： +**出力例** -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) - -**LoRA（オプション）** — Dev 蒸留は LoRA として Full モデルにも適用でき、蒸留強度を調整できます（[Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： - -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — フルランク -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — プルーニング版 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 代替 Checkpoint ベースの蒸留 - -``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 checkpoints/ -│ │ ├── hidream_o1_image_fp8_scaled.safetensors -│ │ ├── hidream_o1_image_mxfp8.safetensors -│ │ └── hidream_o1_image_bf16.safetensors -│ ├── 📂 loras/ -│ │ └── hidream_o1_dev_lora_rank_64_bf16.safetensors -│ └── 📂 text_encoders/ -│ └── gemma4_e4b_it_fp8_scaled.safetensors -``` +![HiDream O1 Full 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) ### 3. ワークフローの使い方 @@ -89,42 +62,41 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` ## HiDream-O1-Image Dev ワークフロー -### 1. ワークフローをダウンロード +### HiDream O1 Dev (`image_hidream_o1_dev`) -メニューから `ワークフロー` → `テンプレートを閲覧` → `Image` に進み、"HiDream O1 Dev" を見つけてください。 +Input a text prompt and optional reference images. Generate a high-resolution image (up to 2048x2048) with support for text-to-image, image editing, and subject-driven personalization. -![HiDream-O1-Image Dev ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1_dev.png) +HiDream O1 Dev ワークフロープレビュー - - Download workflow + + + Comfy Cloud で開く - - - Open in cloud + + Download JSON or search "HiDream O1 Dev" in Template Library + -### 2. モデルをダウンロード +**入力素材** -**チェックポイント（Dev）** — 再パッケージおよび量子化済み。すべてのバージョンで最悪の外れ値は bf16 で保持され、未使用の deepstack 層は削除されています： +Upload this file to the matching `LoadImage` node: -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量子化版。対応ハードウェア上で安全な MLP 層に fp8/mxfp8 行列乗算を使用し高速化 -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量子化版 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 フル精度版（最大ファイル） + + + `LoadImage` node 213 · `noir_portrait.png` + + -**テキストエンコーダ（プロンプト補強）** — 全バージョン共通： +
+ noir_portrait.png +
-- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +**出力例** -``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 checkpoints/ -│ │ ├── hidream_o1_image_dev_fp8_scaled.safetensors -│ │ ├── hidream_o1_image_dev_mxfp8.safetensors -│ │ └── hidream_o1_image_dev_bf16.safetensors -│ └── 📂 text_encoders/ -│ └── gemma4_e4b_it_fp8_scaled.safetensors -``` +
+ 入力画像 + HiDream O1 Dev 出力例 +
### 3. ワークフローの使い方 diff --git a/ja/tutorials/image/ideogram/ideogram-v4.mdx b/ja/tutorials/image/ideogram/ideogram-v4.mdx index 2c44d5ea6..4363d0eb8 100644 --- a/ja/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ja/tutorials/image/ideogram/ideogram-v4.mdx @@ -10,25 +10,30 @@ translationBlockHashes: "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; Ideogram 4.0 は、Ideogram がオープンソースモデルとして公開した最新のテキストから画像へのモデルで、ローカルで完全に動作します。優れたフォトリアリスティックな品質、正確なテキストレンダリング、精密なスタイル制御を備えています。自然言語または **構造化 JSON プロンプト** を使用して、レイアウト、色、画像内テキストを細かく制御できます。 -## Ideogram 4.0 テキストから画像へのワークフロー +### Ideogram v4: Text to Image (`image_ideogram4_t2i`) + +Input a text prompt or structured JSON description. Generate an image with precise layout, color, and style control using Ideogram 4.0. - +Ideogram 4.0 text-to-image ワークフロープレビュー + + + Comfy Cloud で開く - - - JSON をダウンロードするか、テンプレートライブラリで"Ideogram v4: Text to Image"を検索 + + Download JSON or search "Ideogram v4: Text to Image" in Template Library + + +**出力例** ![Ideogram 4.0 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) -*Ideogram 4.0 オープンソースモデルの出力例* ### プロンプト形式 @@ -46,25 +51,23 @@ Ideogram 4.0 は、Ideogram がオープンソースモデルとして公開し Hugging Face の [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) ですべての再パッケージ化されたモデルファイルを見つけることができます。 - + + Ideogram 4.0 拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 - - + Ideogram 4.0 条件なし拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 - - + Ideogram 4.0 テキストエンコーダー（~8 GB）。models/text_encoders/ に配置 - - + Ideogram 4.0 テキストエンコーダー（~2 GB）。models/text_encoders/ に配置 - - + Ideogram 4.0 VAE（~335 MB）。models/vae/ に配置 + **モデルの保存場所** diff --git a/ja/tutorials/image/krea/krea-2.mdx b/ja/tutorials/image/krea/krea-2.mdx index 1ee36b678..6c038ea2d 100644 --- a/ja/tutorials/image/krea/krea-2.mdx +++ b/ja/tutorials/image/krea/krea-2.mdx @@ -12,9 +12,6 @@ translationBlockHashes: "Krea-2 Turbo style reference workflow": 5e966a16 --- - - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -39,7 +36,6 @@ Krea 2には2つのバリアントがあります： - **Krea 2 RAW**：フルステップサンプリング（52ステップ）のベースモデル。多様性と柔軟性のために設計されており、ファインチューニングやLoRAトレーニングに最適です。 - **Krea 2 Turbo**：8ステップの蒸留チェックポイントで、高速かつ高品質な生成を実現。RAWで学習したLoRAをシームレスにTurboに適用できます。 - **関連リンク**: - [Krea 2 on Hugging Face (RAW)](https://huggingface.co/krea/Krea-2-Raw) - [Krea 2 on Hugging Face (Turbo)](https://huggingface.co/krea/Krea-2-Turbo) @@ -47,30 +43,32 @@ Krea 2には2つのバリアントがあります： - [公式GitHubリポジトリ](https://github.com/krea-ai/krea-2) - [テクニカルレポート](https://www.krea.ai/blog/krea-2-technical-report) -## Krea-2 Turbo テキストから画像へのワークフロー +### Krea-2: Text to Image (`image_krea2_turbo_t2i`) -Krea-2 Turbo テキストから画像ワークフロー +Generate images from text prompts using Krea 2, a foundation model built for aesthetic quality and creative control. It focuses on rendering expressive, stylistically diverse images. + +Krea-2 Turbo text-to-image ワークフロープレビュー Comfy Cloud で開く - JSON をダウンロードするか、テンプレートライブラリで "Krea-2" を検索 + Download JSON or search "Krea-2" in Template Library -ワークフローは以下の部分で構成されています： +The workflow is organized into a few parts: -1. **Text to Image (Krea-2 Turbo) サブグラフ**：モデル読み込み、プロンプト処理、サンプリング、VAEデコードを含むコア生成パイプライン -2. **ResolutionSelector**：希望する出力解像度を選択します。Krea 2は1Kから2Kまでの出力をサポートしており、メガピクセル値を2.0に設定すると2K解像度が得られます。 -3. **CustomCombo (LoRAセレクター)**：利用可能なスタイルLoRA用のトリガーワードセレクターがあらかじめ組み込まれています。追加のLoRAをダウンロードした場合、このセレクターをカスタマイズし、対応するLoRAファイルと組み合わせて使用できます。 -4. **SaveImage**：生成された画像を保存します。 +1. **Text to Image (Krea-2 Turbo) subgraph**: the core generation pipeline, containing model loading, prompt handling, sampling, and VAE decode +2. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K — set the megapixels value to 2.0 to get 2K resolution. +3. **CustomCombo (LoRA selector)**: a pre-built trigger word selector for the available style LoRAs. If you download additional LoRAs, you can customize this selector and pair them with the corresponding LoRA files accordingly. +4. **SaveImage**: saves the generated image - - このワークフローはサブグラフノードを使用してモジュール化されています。サブグラフのドキュメントを確認して、ワークフローをカスタマイズおよび拡張する方法を学びましょう。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ### ワンクリック生成 @@ -146,27 +144,47 @@ Krea 2向けのスタイルLoRAコレクションもKreaから公開されてい │ └── krea2_softwatercolor.safetensors (and other style LoRAs) ``` -## Krea-2 Turbo スタイル参照ワークフロー +### Krea-2 Int8: Image Style Reference (`image_krea2_turbo_int8_image_style_reference`) + +Generate images with the Krea-2 Turbo model while referencing the style of 1–2 uploaded images, using the high-performance Int8 Convrot format for fast inference. -Krea-2 Turbo スタイル参照ワークフロー +Krea-2 Turbo style reference ワークフロープレビュー Comfy Cloud で開く - JSONをダウンロードするか、テンプレートライブラリで「Krea-2 スタイル参照」を検索してください + Download JSON or search "Krea-2 Style Reference" in Template Library -スタイル参照ワークフローは、Krea-2 Turbo パイプラインに参照画像の条件付けを追加したものです。1つ以上の参照画像をアップロードして、生成出力の美的スタイル、雰囲気、ビジュアルの方向性に影響を与えます。 +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 69 · `krea2_reference_image.png` + + + +
+ krea2_reference_image.png +
+ +**出力例** + +![Krea-2 style reference 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_krea2_turbo_int8_image_style_reference.png) + +The style reference workflow builds on the Krea-2 Turbo pipeline by adding reference image conditioning. Upload one or more reference images to influence the aesthetic style, mood, and visual direction of the generated output. -ワークフローは以下のいくつかの部分で構成されています： +The workflow is organized into a few parts: -1. **画像スタイル参照（Krea-2 Turbo）サブグラフ**：スタイル参照に対応したコア生成パイプラインです。モデルの読み込み、参照画像の条件付け、プロンプト処理、サンプリングを含みます。 -2. **LoadImage**：スタイル参照画像をアップロードします。 -3. **ResolutionSelector**：希望する出力解像度を選択します。Krea 2 は 1K から 2K までの出力をサポートします。 -4. **SaveImage**：生成画像を保存します。 +1. **Image Style Reference (Krea-2 Turbo) subgraph**: the core generation pipeline with style reference support, containing model loading, reference image conditioning, prompt handling, and sampling +2. **LoadImage**: upload your style reference images +3. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K. +4. **SaveImage**: saves the generated image ### スタイル参照の使用方法 diff --git a/ja/tutorials/image/lens/lens.mdx b/ja/tutorials/image/lens/lens.mdx index edd441dec..a092229fd 100644 --- a/ja/tutorials/image/lens/lens.mdx +++ b/ja/tutorials/image/lens/lens.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Available models": 5876b860 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Lens** は **Microsoft** によるオープンなテキスト画像生成モデルで、MIT ライセンスで提供されています。**38億** パラメータを持ち、**デュアルストリーム MMDiT** アーキテクチャに **GPT-OSS-20B** テキストエンコーダーの特徴量と **FLUX.2 セマンティック VAE** を組み合わせ、より大規模なモデルよりも少ない学習計算量で競争力のある画質を実現します。 @@ -37,57 +36,63 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' このワークフローは Subgraph ノードを使用したモジュール処理を採用しています。Subgraph のドキュメントを参照して、ワークフローのカスタマイズと拡張方法を学んでください。 -### Lens +### Lens: Text to Image (`image_lens_t2i`) + +Input a text prompt and select resolution and aspect ratio. Generate a high-quality image using the efficient Lens text-to-image model. + +Lens text-to-image ワークフロープレビュー - JSON をダウンロード、またはテンプレートライブラリで "Lens" を検索 + Download JSON or search "Lens" in Template Library - {/* TODO: Lens が Comfy Cloud で利用可能になったら有効化 */} + {/* TODO: Enable Cloud template when Lens is available on Comfy Cloud */} {/**/} {/* Comfy Cloud で開く*/} {/**/} - +**出力例** + +![Lens text-to-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_t2i.png) -#### はじめ方 + -1. ComfyUI を最新バージョンに更新 - {/* TODO: Cloud テンプレートが利用可能になったら Cloud オプションを追加 */} -2. **テンプレート** で **Lens** を検索 -3. **Lens** ワークフローを選択 -4. 不足しているモデルをダウンロードして（[モデルダウンロード](#モデルダウンロード) 参照）、プロンプトを入力し **実行** をクリック +#### Get started -#### 出力例 +1. Update ComfyUI to the latest version + {/* TODO: Add Cloud option when template is available */} +2. Go to **Template** and search for **Lens** +3. Select the **Lens** workflow +4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** -Lens テキスト画像生成の出力例 +### Lens Turbo: Text to Image (`image_lens_turbo_t2i`) -### Lens Turbo +Input a text prompt and select resolution, aspect ratio, and inference steps. Generate a high-quality image using the Lens text-to-image model. -Lens Turbo は蒸留版で、より少ないサンプリングステップで画像を生成し、高速な推論を実現します。 +Lens Turbo text-to-image ワークフロープレビュー - JSON をダウンロード、またはテンプレートライブラリで "Lens Turbo" を検索 + Download JSON or search "Lens Turbo" in Template Library - {/* TODO: Lens Turbo が Comfy Cloud で利用可能になったら有効化 */} + {/* TODO: Enable Cloud template when Lens Turbo is available on Comfy Cloud */} {/**/} {/* Comfy Cloud で開く*/} {/**/} -#### はじめ方 +**出力例** -1. ComfyUI を最新バージョンに更新 - {/* TODO: Cloud テンプレートが利用可能になったら Cloud オプションを追加 */} -2. **テンプレート** で **Lens Turbo** を検索 -3. **Lens Turbo** ワークフローを選択 -4. 不足しているモデルをダウンロードして（[モデルダウンロード](#モデルダウンロード) 参照）、プロンプトを入力し **実行** をクリック +![Lens Turbo text-to-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_turbo_t2i.png) -#### 出力例 +#### Get started -Lens Turbo テキスト画像生成の出力例 +1. Update ComfyUI to the latest version + {/* TODO: Add Cloud option when template is available */} +2. Go to **Template** and search for **Lens Turbo** +3. Select the **Lens Turbo** workflow +4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** ## モデルダウンロード diff --git a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 024ee3d5b..ad3ad9abf 100644 --- a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Prompt format": 5b819c6c --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **NewBie-image-Exp0.1** は、NewBieAI Lab が開発した 35 億パラメータの DiT（Diffusion Transformer）モデルで、アニメスタイルの文生成画像タスク専用に設計されています。Next-DiT アーキテクチャを採用しており、非常に詳細で視覚的に印象的なアニメ画像を生成できます。 @@ -29,26 +27,25 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - [Hugging Face](https://huggingface.co/NewBie-AI/NewBie-image-Exp0.1) - [はじめにガイド](https://ai.feishu.cn/wiki/NZl9wm7V1iuNzmkRKCUcb1USnsh) -## NewBie-image 文生成画像ワークフロー +### NewBie Exp0.1: Anime Generation (`image_newbieimage_exp0_1-t2i`) - - - Download JSON or search "NewBie-image" in Template Library - - - Open in cloud - - +Generate detailed anime-style images with NewBie Exp0.1's Next-DiT architecture. Supports XML structured prompts for better multi-character scenes and attribute binding. + +NewBie-image text-to-image ワークフロープレビュー - - Download JSON or search "NewBie-image" in Template Library + + Comfy Cloud で開く - - クラウドで開く + + Download JSON or search "NewBie-image" in Template Library +**出力例** + +![NewBie-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_newbieimage_exp0_1-t2i.png) + ## モデルのダウンロードリンク diff --git a/ja/tutorials/image/omnigen/omnigen2.mdx b/ja/tutorials/image/omnigen/omnigen2.mdx index 87674bfb1..0d41c40e1 100644 --- a/ja/tutorials/image/omnigen/omnigen2.mdx +++ b/ja/tutorials/image/omnigen/omnigen2.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "ComfyUI OmniGen2 Image Editing Workflow": 1e06072e --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## OmniGen2 について @@ -65,15 +63,24 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 ## ComfyUI OmniGen2 テキストから画像へ（Text-to-Image）ワークフロー -### 1. ワークフローファイルのダウンロード +### OmniGen2: Text to Image (`image_omnigen2_t2i`) + +Generate high-quality images from text prompts using OmniGen2's unified 7B multimodal model with dual-path architecture. + +OmniGen2 text-to-image ワークフロープレビュー - Open and run this workflow directly in Comfy Cloud. + Comfy Cloud で直接開いて実行 + + + Download JSON or search "OmniGen2" in Template Library -![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) +**出力例** + +![OmniGen2 text-to-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_omnigen2_t2i.png) ### 2. ワークフローの手順通り実行 @@ -95,18 +102,41 @@ OmniGen2 は、総パラメータ数約 **70 億**（テキストモデル 30 OmniGen2 は豊富な画像編集機能を備えており、画像へのテキスト追加もサポートします。 -### 1. ワークフローファイルのダウンロード +### OmniGen2 Image Edit (`image_omnigen2_image_edit`) + +Edit images with natural language instructions using OmniGen2's advanced image editing capabilities and text rendering support. + +OmniGen2 image edit ワークフロープレビュー - Open and run this workflow directly in Comfy Cloud. + Comfy Cloud で直接開いて実行 + + + Download JSON or search "OmniGen2 Image Edit" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 16 · `image_omnigen2_image_edit_input_image.png` -![テキストから画像へワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) +
+ image_omnigen2_image_edit_input_image.png +
+ +**出力例** -以下の画像をダウンロードし、この画像を入力として使用します。 -![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/input_fairy.png) +
+ 入力画像 + OmniGen2 image edit 出力例 +
### 2. ワークフローの手順通り実行 diff --git a/ja/tutorials/image/ovis/ovis-image.mdx b/ja/tutorials/image/ovis/ovis-image.mdx index 2dd410dd0..5250029a2 100644 --- a/ja/tutorials/image/ovis/ovis-image.mdx +++ b/ja/tutorials/image/ovis/ovis-image.mdx @@ -20,23 +20,18 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - [GitHub](https://github.com/AIDC-AI/Ovis-Image) - [Hugging Face](https://huggingface.co/AIDC-AI/Ovis-Image-7B) -## Ovis-Image のテキストから画像を生成するワークフロー +### Ovis-Image Text to Image (`image_ovis_text_to_image`) - - - Comfy Cloud で開く - - - Download JSON or search "Ovis image" in Template Library - - +Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering in generated images. Designed to operate efficiently under stringent computational constraints. + +Ovis-Image text-to-image ワークフロープレビュー - + Comfy Cloud で開く - Download JSON or search "Ovis image" in Template Library + Download JSON or search "Ovis image" in Template Library diff --git a/ja/tutorials/image/pixeldit/pixeldit.mdx b/ja/tutorials/image/pixeldit/pixeldit.mdx index f6d92b263..9d80b3f34 100644 --- a/ja/tutorials/image/pixeldit/pixeldit.mdx +++ b/ja/tutorials/image/pixeldit/pixeldit.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Model downloads": e4bafb0a --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **PixelDiT** は NVIDIA が開発したピクセル空間拡散トランスフォーマーで、**1024px** のテキストから画像への生成を行います。従来の潜在空間で動作する拡散モデルとは異なり、PixelDiT はデュアルレベル DiT アーキテクチャを使用してピクセル空間で直接画像を生成します——パッチレベル DiT とピクセルレベル DiT を組み合わせ、MM-DiT フュージョンによるテキストと画像トークン間の joint attention を実現します。 diff --git a/ja/tutorials/image/qwen/qwen-image-2512.mdx b/ja/tutorials/image/qwen/qwen-image-2512.mdx index 0c7bbfce2..b8bfd9dcc 100644 --- a/ja/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ja/tutorials/image/qwen/qwen-image-2512.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Qwen-Image-2512 ComfyUI Native Workflow Example": 50fbe75c --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image-2512** は、Qwen-Image のテキストから画像を生成する基盤モデルの12月アップデート版です。8月にリリースされたベース版 Qwen-Image モデルと比較して、Qwen-Image-2512 は画像品質およびリアリズムにおいて大幅な向上を実現しています。 @@ -56,8 +54,6 @@ ComfyUI を更新した後、テンプレートからワークフローファイ - **Text to Image (Qwen-Image 2512)**: 標準的な50ステップ生成 - **Text to Image (Qwen-Image 2512 4steps)**: Lightning LoRA を用いた高速4ステップ生成 - - ### 2. モデルのダウンロード **テキストエンコーダー** diff --git a/ja/tutorials/image/qwen/qwen-image-edit.mdx b/ja/tutorials/image/qwen/qwen-image-edit.mdx index 06a485c07..dbcbeb21d 100644 --- a/ja/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ja/tutorials/image/qwen/qwen-image-edit.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Qwen-Image-Edit ComfyUI Native Workflow Example": 6703460f --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image-Edit** は、Qwen-Image の画像編集専用バージョンです。20B規模の Qwen-Image モデルを基に追加学習が行われており、Qwen-Image の特徴的なテキストレンダリング能力を編集タスクへと成功裏に拡張し、高精度なテキスト編集を実現しています。さらに、Qwen-Image-Edit では入力画像を Qwen2.5-VL（視覚的意味制御用）および VAE エンコーダ（視覚的外観制御用）の両方に同時に供給することで、意味と外観の両方を独立して制御可能な「二重編集機能」を実現しています。 @@ -43,31 +41,41 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' -### 1. ワークフローファイル +### Qwen Image Edit (`image_qwen_image_edit`) -ComfyUI を更新後、テンプレートからワークフローファイルを取得するか、下記のワークフローを ComfyUI へドラッグ＆ドロップして読み込むことができます。 -![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) +Edit images with precise bilingual text editing and dual semantic/appearance editing capabilities using Qwen-Image-Edit's 20B MMDiT model. + +Qwen-Image-Edit ワークフロープレビュー - - Download JSON or search "image_qwen_image_edit" in Template Library - - Run this workflow on Cloud GPUs with zero setup + Cloud GPU でセットアップ不要で実行 + + + Download JSON or search "Qwen Image Edit" in Template Library +**入力素材** + +Upload this file to the matching `LoadImage` node: + - - Download JSON or search "image_qwen_image_edit" in Template Library - - - Run this workflow on Cloud GPUs with zero setup + + `LoadImage` node 78 · `image_qwen_image_edit_input_image.png` -以下の画像を入力画像としてダウンロードしてください。 -![Qwen-image テキストから画像生成のワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) +
+ image_qwen_image_edit_input_image.png +
+ +**出力例** + +
+ 入力画像 + Qwen-Image-Edit 出力例 +
### 2. モデルのダウンロード diff --git a/ja/tutorials/image/qwen/qwen-image-layered.mdx b/ja/tutorials/image/qwen/qwen-image-layered.mdx index 63b66cfb3..7f1606f80 100644 --- a/ja/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ja/tutorials/image/qwen/qwen-image-layered.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Workflow settings": 098636f1 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image-Layered** は、アリババ社の通義千問（Qwen）チームが開発したモデルで、入力画像を複数の RGBA レイヤーに分解することができます。このレイヤー化された表現により、本質的な編集可能性が実現されます：各レイヤーを他のコンテンツに影響を与えることなく独立して操作できます。 diff --git a/ja/tutorials/image/qwen/qwen-image.mdx b/ja/tutorials/image/qwen/qwen-image.mdx index 0d7391946..d1d49190e 100644 --- a/ja/tutorials/image/qwen/qwen-image.mdx +++ b/ja/tutorials/image/qwen/qwen-image.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Qwen Image Union ControlNet LoRA Workflow": a08d8e37 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Qwen-Image** は、アリババのQwenチームがリリースした初の画像生成基盤モデルです。これは、Apache 2.0ライセンスのもとでオープンソース化された20BパラメータのMMDiT（マルチモーダル拡散トランスフォーマー）モデルです。このモデルは、**複雑なテキストレンダリング**および**精密な画像編集**において顕著な進展を遂げており、英語や中国語など複数の言語において高忠実度の出力を実現しています。 @@ -54,40 +52,46 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' allowFullScreen > +### Qwen-Image: Text to Image (`image_qwen_image`) -## Qwen-Imageネイティブワークフローの例 +Generate images with exceptional multilingual text rendering and editing capabilities using Qwen-Image's 20B MMDiT model. - +Qwen-Image text-to-image ワークフロープレビュー + + + + Comfy Cloud で開く + + + Download JSON or search "Qwen-Image" in Template Library + + - - Comfy Cloudで実行 - +**出力例** -本ドキュメントに添付されたワークフローでは、以下の3種類の異なるモデルが使用されています： -1. Qwen-Imageオリジナルモデル（fp8_e4m3fn） -2. 8ステップ高速化版：Qwen-Imageオリジナルモデル（fp8_e4m3fn）＋lightx2v製8ステップLoRA -3. 蒸留版：Qwen-Image蒸留モデル（fp8_e4m3fn） +![Qwen-Image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_qwen_image.png) -**VRAM使用量の参考値** -GPU：RTX4090D（24GB） +There are three different models used in the workflow attached to this document: +1. Qwen-Image original model fp8_e4m3fn +2. 8-step accelerated version: Qwen-Image original model fp8_e4m3fn with lightx2v 8-step LoRA +3. Distilled version: Qwen-Image distilled model fp8_e4m3fn -| 使用モデル | VRAM使用量 | 初回生成時間 | 2回目以降の生成時間 | -| --------------------------------------- | ---------- | ------------ | ------------------ | -| fp8_e4m3fn | 86% | ≈ 94秒 | ≈ 71秒 | -| fp8_e4m3fn（lightx2v 8ステップLoRA使用） | 86% | ≈ 55秒 | ≈ 34秒 | -| 蒸留版 fp8_e4m3fn | 86% | ≈ 69秒 | ≈ 36秒 | +**VRAM Usage Reference** +GPU: RTX4090D 24GB +| Model Used | VRAM Usage | First Generation | Second Generation | +| --------------------------------------- | ---------- | --------------- | ---------------- | +| fp8_e4m3fn | 86% | ≈ 94s | ≈ 71s | +| fp8_e4m3fn with lightx2v 8-step LoRA | 86% | ≈ 55s | ≈ 34s | +| Distilled fp8_e4m3fn | 86% | ≈ 69s | ≈ 36s | ### 1. ワークフローファイル ComfyUIを更新後、テンプレートからワークフローファイルを検索するか、以下のワークフローをComfyUIにドラッグ＆ドロップして読み込むことができます。 ![Qwen-image テキストから画像へ変換するワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/qwen/qwen-image.png) - - 蒸留版 - ### 2. モデルのダウンロード **ComfyUIで利用可能なモデル** @@ -157,15 +161,36 @@ Qwen_image_distill（蒸留版） 蒸留モデルとlightx2v製8ステップ高速化LoRAは、同時に使用できない可能性があります。両者を組み合わせた動作を確認するために、さまざまな組み合わせを試すことができます。 -## Qwen Image InstantX ControlNetワークフロー +### Qwen-Image InstantX Union ControlNet (`image_qwen_image_instantx_controlnet`) -これはControlNetモデルであるため、通常のControlNetとして使用できます。 +Generate images with Qwen-Image InstantX ControlNet, supporting canny, soft edge, depth, and pose. - - Comfy Cloudで実行 - +Qwen-Image InstantX ControlNet ワークフロープレビュー -### 1. ワークフローおよび入力画像 + + + Comfy Cloud で開く + + + Download JSON or search "Qwen-Image InstantX ControlNet" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 71 · `image_qwen_image_instantx_controlnet_input_image.jpg` + + + +
+ image_qwen_image_instantx_controlnet_input_image.jpg +
+ +**出力例*. ワークフローおよび入力画像 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) @@ -219,25 +244,40 @@ ComfyUI/ 3. このサブグラフはLotus Depthモデルを使用しています。テンプレートからLotus Depthを検索するか、サブグラフを編集して詳細を確認し、すべてのモデルが正しく読み込まれていることを確認してください 4. `Run`ボタンをクリックするか、ショートカット`Ctrl(cmd) + Enter`でワークフローを実行してください -## Qwen Image ControlNet DiffSynth-ControlNetsモデルパッチワークフロー +### Qwen-Image ControlNet Model Patch (`image_qwen_image_controlnet_patch`) - - Comfy Cloudで実行 - +Control image generation using Qwen-Image ControlNet models. Supports canny, depth, and inpainting controls through model patching. -このモデルは実際にはControlNetではなく、Canny、Depth、Inpaintの3種類の異なる制御モードをサポートする「モデルパッチ」です。 +Qwen-Image ControlNet model patch ワークフロープレビュー -オリジナルモデルのURL：[DiffSynth-Studio/Qwen-Image ControlNet](https://www.modelscope.cn/collections/Qwen-Image-ControlNet-6157b44e89d444) -Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/tree/main/split_files/model_patches) + + + Comfy Cloud で開く + + + Download JSON or search "Qwen-Image ControlNet Patch" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 71 · `image_qwen_image_controlnet_patch_input_image.png` + + +
+ image_qwen_image_controlnet_patch_input_image.png +
### 1. ワークフローおよび入力画像 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして対応するワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) - - 以下の画像を入力としてダウンロードしてください： ![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/input.png) @@ -250,7 +290,6 @@ Comfy Org再ホストURL：[Qwen-Image-DiffSynth-ControlNets/model_patches](http - [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) - [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) - ### 3. ワークフローの使用方法 現在、diffsynthにはCanny、Depth、Inpaintの3種類のパッチモデルがあります。 @@ -288,21 +327,50 @@ Inpaintモデルでは、[マスクエディター](/ja/interface/maskeditor)を 5. 必要に応じて、`QwenImageDiffsynthControlnet`ノードの`strength`を調整して、対応する制御強度を制御できます 6. `Run`ボタンをクリックするか、ショートカット`Ctrl(cmd) + Enter`でワークフローを実行してください -## Qwen Image Union ControlNet LoRAワークフロー +### Qwen-Image Union Control (`image_qwen_image_union_control_lora`) + +Generate images with precise structural control using Qwen-Image's unified ControlNet LoRA. Supports multiple control types including canny, depth, lineart, softedge, normal, and openpose. + +Qwen-Image Union Control ワークフロープレビュー + + + + Comfy Cloud で開く + + + Download JSON or search "Qwen-Image Union Control" in Template Library + + - - Comfy Cloudで実行 - +**入力素材** -オリジナルモデルのURL：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) -Comfy Org再ホストURL：[qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors)：Canny、Depth、Pose、Lineart、Softedge、Normal、Openposeをサポートする画像構造制御用LoRA +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 73 · `image_qwen_image_union_control_lora_input_image.png` + + + +
+ image_qwen_image_union_control_lora_input_image.png +
+ +**出力例** + +
+ 入力画像 + Qwen-Image Union Control 出力例 +
+ +Original model address: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) +Comfy Org rehost address: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): Image structure control LoRA supporting canny, depth, pose, lineart, softedge, normal, openpose ### 1. ワークフローおよび入力画像 以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップしてワークフローを読み込んでください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - 以下の画像を入力としてダウンロードしてください ![ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/input.png) diff --git a/ja/tutorials/image/z-image/z-image-turbo.mdx b/ja/tutorials/image/z-image/z-image-turbo.mdx index 13004d220..5be36630b 100644 --- a/ja/tutorials/image/z-image/z-image-turbo.mdx +++ b/ja/tutorials/image/z-image/z-image-turbo.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Z-Image-Turbo Fun Union ControlNet workflow": 34191a15 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -33,17 +31,24 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - [GitHub](https://github.com/Tongyi-MAI/Z-Image) - [Hugging Face](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo) -## Z-Image-Turbo 文字から画像へのワークフロー +### Z-Image-Turbo: Text to Image (`image_z_image_turbo`) + +An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer, supports English & Chinese. + +Z-Image-Turbo text-to-image ワークフロープレビュー - - Z-Image-Turbo 文字から画像へのワークフローのJSONファイルをダウンロードします。 + + Comfy Cloud でこのワークフローを直接実行 - - このワークフローを ComfyUI Cloud 上で直接実行します。 + + Download the Z-Image-Turbo text-to-image workflow JSON file - + +**出力例** + +![Z-Image-Turbo 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_z_image_turbo.png) ### Z-Image-Turbo モデルのダウンロード @@ -73,26 +78,32 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## Z-Image-Turbo Fun Union ControlNet ワークフロー -このワークフローでは、Z-Image-Turbo Fun Union ControlNet モデルを用いて、ControlNet を活用した画像生成を行います。参照画像に対してCannyエッジ検出を適用し、その結果をControlNetによる生成プロセスのガイドとして活用します。 +### Z-Image-Turbo Fun Union ControlNet (`image_z_image_turbo_fun_union_controlnet`) - - Z-Image-Turbo Fun Union ControlNet ワークフローのJSONファイルをダウンロードします。 +Multi-control ControlNet supporting Canny, HED, Depth, Pose, and MLSD for Z-Image-Turbo. + +Z-Image-Turbo Fun Union ControlNet ワークフロープレビュー + + + + Comfy Cloud でこのワークフローを直接実行 -### ControlNet 用の追加モデル + + Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file + + + +**入力素材** - - Z-Image-Turbo 専用のControlNetモデルパッチです。 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 58 · `image_z_image_turbo_fun_union_controlnet_input_image.png` -**モデルの保存場所** + + +
+ image_z_image_turbo_fun_union_controlnet_input_image.png +
-``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 text_encoders/ -│ │ └── qwen_3_4b.safetensors -│ ├── 📂 diffusion_models/ -│ │ └── z_image_turbo_bf16.safetensors -│ ├── 📂 vae/ -│ │ └── ae.safetensors -│ └── 📂 model_patches/ -│ └── Z-Image-Turbo-Fun-Controlnet-Union.safetensors diff --git a/ja/tutorials/llm/gemma4/gemma4.mdx b/ja/tutorials/llm/gemma4/gemma4.mdx index 0234a4b0e..f35857439 100644 --- a/ja/tutorials/llm/gemma4/gemma4.mdx +++ b/ja/tutorials/llm/gemma4/gemma4.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Model Download": 9f2919ea --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -35,58 +34,79 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - [Google AI for Developers](https://ai.google.dev/gemma) - [ComfyUI ソースコード (nodes_textgen.py)](https://github.com/Comfy-Org/ComfyUI/blob/master/comfy_extras/nodes_textgen.py) -## 利用可能なワークフロー +## Available workflow -### Gemma 4：テキスト生成 +### Gemma4: Text Generation (`llm_gemma4_text_gen`) - - JSON をダウンロード、またはテンプレートライブラリで "Gemma 4 Text Generation" を検索 - +Input your text prompt and optionally an image, audio, or video. Generate text output with configurable reasoning, coding, and multilingual support. + +Gemma 4 text generation workflow preview - - Comfy Cloud で開く + + + Open in Comfy Cloud + + Download JSON or search "Gemma4: Text Generation" in Template Library + + + +**入力素材** -![Gemma 4 テキスト生成ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_gemma4_text_gen-1.webp) +Upload these optional files to the matching nodes: + + + + `LoadImage` node 2 · `the_lily_veil.png` + + + `LoadAudio` node 5 · `voice_demo.mp3` + + + `LoadVideo` node 6 · `video_wan_vace_inpainting_input_video.mp4` + + -このワークフローは Gemma 4 のコアとなる**テキスト生成**機能を示しています。テキストプロンプトに加えて、オプションで画像、音声、動画を追加コンテキストとして入力でき、推論・コーディング・多言語プロンプトに対応した自然言語出力を生成します。 +This workflow demonstrates the core **text generation** capabilities of Gemma 4. It accepts an optional image, audio file, or video as additional context alongside your text prompt, and generates natural language output — with support for reasoning, coding, and multilingual prompts. -**入力**： -- **テキストプロンプト** — 質問または指示 -- **画像**（オプション）— 視覚理解タスク用（OCR、物体検出、チャート読み取りなど） -- **音声**（オプション）— 音声認識・文字起こし用 -- **動画**（オプション）— フレーム単位の動画理解用（内部で 1 FPS にサブサンプリング） +**Inputs**: +- **Text prompt** — your question or instruction +- **Image** (optional) — for visual understanding tasks (OCR, object detection, chart reading, etc.) +- **Audio** (optional) — for speech recognition or transcription +- **Video** (optional) — for video understanding across frames (subsampled to 1 FPS internally) -**主要な制御パラメータ**： -- **Max length** — 生成する最大トークン数（デフォルト 256） -- **Sampling mode** — サンプリングのオン/オフ、temperature、top-k、top-p、繰り返しペナルティ、シードを調整 -- **Thinking mode** — 最終回答前の段階的推論を有効化 -- **Use default template** — モデル内蔵のシステムプロンプトテンプレートを使用 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Thinking mode** — enable step-by-step reasoning before the final answer +- **Use default template** — apply the built-in system prompt for the model -**出力**： -- **Generated text** — モデルが生成したテキスト応答 +**Output**: +- **Generated text** — the model's response as a plain text string - - このワークフローは Subgraph ノードを使用してモジュール処理を実現しています。Subgraph ドキュメントでカスタマイズと拡張の方法を確認してください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## モデルのダウンロード +## Model Download -Gemma 4 モデルは ComfyUI ではテキストエンコーダー（text encoder）として読み込まれます。該当するモデルファイルをダウンロードし、正しいディレクトリに配置してください： +Gemma 4 models are loaded as text encoders in ComfyUI. Download the relevant model file and place it in the correct directory: - - 高速・軽量、コンシューマー GPU に最適。 + + + Fast, lightweight. Recommended for consumer GPUs. - - バランスの取れた性能。ワークフローのデフォルトモデル。 + + Balanced performance. The default model in the workflow. - - すべての Gemma 4 モデルウェイトを閲覧。 + + Browse all Gemma 4 model weights. + -ダウンロードした `.safetensors` ファイルを以下のディレクトリに配置してください： +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ diff --git a/ja/tutorials/llm/qwen/qwen3.mdx b/ja/tutorials/llm/qwen/qwen3.mdx index 0a2e148a1..b4edb8061 100644 --- a/ja/tutorials/llm/qwen/qwen3.mdx +++ b/ja/tutorials/llm/qwen/qwen3.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Model Download": d1d0cfd2 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -25,7 +24,6 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - **ComfyUI ネイティブ** — 組み込み `TextGenerate` ノードで動作、カスタムノード不要 - **軽量** — Qwen3.5 と同じテキストエンコーダー形式を共有、2B/4B/9B の3種類から選択可能 - ## 使用例 Qwen 3.0 は、ComfyUI ワークフロー内で構造化テキスト生成とインテリジェントな推論を必要とするタスクに適しています： @@ -35,59 +33,64 @@ Qwen 3.0 は、ComfyUI ワークフロー内で構造化テキスト生成とイ - **テキスト分析** — テキスト入力から情報を抽出し、コンテンツを分類したり、構造化レポートを生成します。 - **連鎖推論** — 思考モードを有効にして、最終出力の前の中間推論が必要な複雑な多段階タスクを処理します。 -## 利用可能なワークフロー +## Available workflow -### Qwen 3.0：テキスト生成 +### Qwen3.0: Text Generation (`llm_qwen3_text_gen`) - - JSON をダウンロードするか、テンプレートライブラリで "Qwen 3.0 Text Generation" を検索 - +Input a text prompt to generate detailed, reasoned responses using the Qwen3-4B-Thinking model. - - Comfy Cloud で開く - +Qwen 3.0 text generation workflow preview -![Qwen 3.0 テキスト生成ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_text_gen-1.webp) + + + Open in Comfy Cloud + + + Download JSON or search "Qwen3.0: Text Generation" in Template Library + + -このワークフローは Qwen 3.0 のコア**テキスト生成**機能を示しています。テキストプロンプトを受け取り、モデルの推論機能を使用して詳細な構造化レスポンスを生成します。 +This workflow demonstrates the core **text generation** capabilities of Qwen 3.0. It accepts a text prompt and generates detailed, structured responses using the model's built-in reasoning capabilities. -**入力**： -- **テキストプロンプト** — 質問、指示、タスクの説明 +**Inputs**: +- **Text prompt** — your question, instruction, or task description -**主要なコントロール**： -- **Max length** — 生成する最大トークン数（デフォルト 256） -- **Sampling mode** — サンプリングのオン/オフ切り替え、temperature、top-k、top-p、繰り返しペナルティ、シードを調整 -- **Thinking mode** — 最終回答の前に段階的推論を有効化 -- **Use default template** — モデル組み込みシステムプロンプトを適用 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Thinking mode** — enable step-by-step reasoning before the final answer +- **Use default template** — apply the built-in system prompt for the model -**出力**： -- **Generated text** — モデルのレスポンス（プレーンテキスト） +**Output**: +- **Generated text** — the model's response as a plain text string - - このワークフローは Subgraph ノードを使用してモジュール化されています。Subgraph ドキュメントを参照してワークフローのカスタマイズと拡張方法を学んでください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## モデルのダウンロード +## Model Download -Qwen 3.0 モデルは ComfyUI でテキストエンコーダーとして読み込まれます。モデルファイルは Qwen3.5 と共有されています。ハードウェアに合わせて適切なバージョンを選択してください： +Qwen 3.0 models are loaded as text encoders in ComfyUI. The model files are shared with Qwen3.5 — download the variant that best fits your hardware: - - 軽量版、約 4.5 GB。低 VRAM 環境や高速ダウンロードに最適。 + + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - - サイズと品質のバランス型。ほとんどのコンシューマー GPU に推奨。 + + Balanced size and quality. Recommended for most consumer GPUs. - - 最大版、約 19 GB。より高品質な出力、より多くの VRAM が必要。 + + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + -ダウンロードした `.safetensors` ファイルを以下の場所に配置： +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ -│ └── qwen3.5_4b_bf16.safetensors # または 2b / 9b バージョン +│ └── qwen3.5_4b_bf16.safetensors # or 2b / 9b variant ``` diff --git a/ja/tutorials/llm/qwen/qwen3_5.mdx b/ja/tutorials/llm/qwen/qwen3_5.mdx index 0a8733d93..66ff36a4d 100644 --- a/ja/tutorials/llm/qwen/qwen3_5.mdx +++ b/ja/tutorials/llm/qwen/qwen3_5.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Model Download": 84a7c321 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -26,7 +25,6 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - **ComfyUI ネイティブ** — 組み込みの `TextGenerate` ノードを使用、カスタムノード不要 - **軽量** — 4B パラメータモデル、コンシューマー GPU に適切 - ## 使用例 Qwen3.5 は、視覚的理解とテキスト生成の組み合わせが ComfyUI ワークフローに付加価値をもたらすシナリオで力を発揮します： @@ -37,59 +35,78 @@ Qwen3.5 は、視覚的理解とテキスト生成の組み合わせが ComfyUI - **ビジュアル質問応答** — 画像コンテンツに関する質問（「このシーンには何がある？」「背景は何色？」）に対して構造化されたテキスト回答を得ます。 - **テキスト読み取り** — 適切なプロンプトを使用すると、モデルが画像内のテキストやラベルを読み取ろうとすることがありますが、信頼性は文字の品質と鮮明さに依存します。 -## 利用可能なワークフロー +## Available workflow + +### Qwen3.5: Text Generation (`llm_qwen3_5_text_gen`) + +Use the Qwen3.5 model to analyze an input image and generate descriptive text prompts. This workflow performs image captioning and reverse prompt engineering. -### Qwen3.5: テキスト生成 +Qwen3.5 text generation workflow preview - - JSON をダウンロードするか、テンプレートライブラリで "Qwen3.5 Text Generation" を検索 + + + Open in Comfy Cloud + + Download JSON or search "Qwen3.5: Text Generation" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: - - Comfy Cloud で開く + + + `LoadImage` node 2 · `man_with_red_hat.png` + -![Qwen3.5 テキスト生成ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_5_text_gen-1.webp) +
+ Input image +
-このワークフローは Qwen3.5 の**テキスト生成と画像理解**機能を示します。テキストプロンプトとオプションの画像を受け取り、入力に基づいて説明テキストや構造化分析を生成します。 +This workflow demonstrates the **text generation and image understanding** capabilities of Qwen3.5. It accepts a text prompt and an optional image, and generates descriptive text or structured analysis based on the input. -**入力**： -- **テキストプロンプト** — 質問、指示、タスクの説明 -- **画像**（オプション） — 視覚的理解タスク用（画像キャプション作成、リバースプロンプトエンジニアリング、プロンプト最適化等） +**Inputs**: +- **Text prompt** — your question, instruction, or task description +- **Image** (optional) — for visual understanding tasks (image captioning, reverse prompt engineering, prompt optimization, etc.) -**主要なコントロール**： -- **最大長** — 生成する最大トークン数（デフォルト 256） -- **サンプリングモード** — サンプリングのオン/オフ切り替え、温度、top-k、top-p、繰り返しペナルティ、シードの調整 -- **デフォルトテンプレートを使用** — モデル内蔵のシステムプロンプトを適用 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Use default template** — apply the built-in system prompt for the model -**出力**： -- **生成されたテキスト** — プレーンテキスト文字列 +**Output**: +- **Generated text** — the model's response as a plain text string - - このワークフローはサブグラフノードを使用してモジュール処理を行います。サブグラフのドキュメントを確認してワークフローをカスタマイズ・拡張してください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## モデルのダウンロード +## Model Download -Qwen3.5 モデルは ComfyUI でテキストエンコーダーとして読み込まれます。ハードウェアに合わせて適切なバージョンを選択してください： +Qwen3.5 models are loaded as text encoders in ComfyUI. Choose the variant that best suits your hardware: - - 軽量版、約 4.5 GB。低 VRAM 環境や高速ダウンロードに最適。 + + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - - サイズと品質のバランス型。ほとんどのコンシューマー GPU に推奨。 + + Balanced size and quality. Recommended for most consumer GPUs. - - 最大版、約 19 GB。より高品質な出力、より多くの VRAM が必要。 + + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + -ダウンロードした `.safetensors` ファイルを以下の場所に配置： +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ -│ └── qwen3.5_4b_bf16.safetensors # または 2b / 9b バージョン +│ └── qwen3.5_4b_bf16.safetensors # or 2b / 9b variant ``` diff --git a/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx index be3f1ac6f..3b0fd9c55 100644 --- a/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/ja/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Beeble SwitchX: Video Edit": 166c2e59 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx b/ja/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx index d4be79d18..35f530df3 100644 --- a/ja/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx +++ b/ja/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Flux 1.1[pro] Image-to-Image Tutorial": 6b3cfbea --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx b/ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx index d5ca915b7..1dbf4bcce 100644 --- a/ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx +++ b/ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Flux.1 Kontext Max Image Partner Nodes Workflow": df399d73 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import PromptTechniques from "/snippets/ja/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bria/background-removal.mdx b/ja/tutorials/partner-nodes/bria/background-removal.mdx index 784ec4095..c8d4cde13 100644 --- a/ja/tutorials/partner-nodes/bria/background-removal.mdx +++ b/ja/tutorials/partner-nodes/bria/background-removal.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Video Background Processing": aab02df5 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index 3b1be6be6..5ce957ffd 100644 --- a/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 1ae5cf45 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 0c9e28ba8..abee2c4d9 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Available workflows": 16803a18 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index 6e5d3508d..598bcb16d 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Using real-person and AI-generated portraits in ComfyUI for Seedance 2.0": d9966fae --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index d046f9d6d..232170940 100644 --- a/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/ja/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Get started": d14874f4 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx index 1a3a4274c..9784a7db2 100644 --- a/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/ja/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Get started": 64517938 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/google/nano-banana-2-lite.mdx b/ja/tutorials/partner-nodes/google/nano-banana-2-lite.mdx index f7b517f61..ae9a50546 100644 --- a/ja/tutorials/partner-nodes/google/nano-banana-2-lite.mdx +++ b/ja/tutorials/partner-nodes/google/nano-banana-2-lite.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 5a827068 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/google/nano-banana-2.mdx b/ja/tutorials/partner-nodes/google/nano-banana-2.mdx index cb416c0b5..f81503d88 100644 --- a/ja/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/ja/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Get started": f6189d9e --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/google/nano-banana-pro.mdx b/ja/tutorials/partner-nodes/google/nano-banana-pro.mdx index 44c555d74..971147442 100644 --- a/ja/tutorials/partner-nodes/google/nano-banana-pro.mdx +++ b/ja/tutorials/partner-nodes/google/nano-banana-pro.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 38fdf056 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 6d9909590..8a4ddb261 100644 --- a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "HappyHorse 1.0 video edit": 5b531078 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index 6a9cfef2d..36b5e58c0 100644 --- a/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/ja/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Getting started": 27bbb438 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 66fa89cb2..bab93e98b 100644 --- a/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/ja/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Advanced features": 6b37a964 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index 3977901aa..6548d2657 100644 --- a/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/ja/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Multi-view-to-3D workflow": ff251560 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 4a651b5ce..d753672f8 100644 --- a/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/ja/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx index 7e0f2fdc7..ac4d986d0 100644 --- a/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ja/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Tips for better results": b6fdceb5 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/krea2/krea2-t2i.mdx b/ja/tutorials/partner-nodes/krea2/krea2-t2i.mdx index 9d6880290..3fa61bdf4 100644 --- a/ja/tutorials/partner-nodes/krea2/krea2-t2i.mdx +++ b/ja/tutorials/partner-nodes/krea2/krea2-t2i.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Additional Notes": 7099b5ae --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx b/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx index b376af763..12e84efea 100644 --- a/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/ja/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -22,8 +22,6 @@ translationBlockHashes: "Key takeaway": 77d51b1d --- - - **ComfyUI** では、Luma **Uni-1** は **パートナー API ノード**として利用します。**Create** はプロンプト（と任意の参照画像）から新規画像を生成し、**Modify** は入力画像を編集します。**Load Image** / **Save Image** などとノードを接続し、プロンプト・シード・アスペクト比・参照スロットを Luma ノード上で設定してグラフをキューするか、**Comfy Cloud** のテンプレートから開いて試せます。 Luma は Uni-1 を拡散モデルではないデコーダのみの自己回帰モデルとして説明しており、生成前にプロンプトを推論します。キャンバス上では **Create / Modify の選択**、参照画像の役割の明示、シードによる反復が実務上の要点です。 diff --git a/ja/tutorials/partner-nodes/meshy/meshy-6.mdx b/ja/tutorials/partner-nodes/meshy/meshy-6.mdx index fd0ec7942..7ebfa8cb4 100644 --- a/ja/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/ja/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Multi-view to Model Workflow": 0cf6bb73 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 951239804..2ca597581 100644 --- a/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/ja/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -12,12 +12,10 @@ translationBlockHashes: "Moonvalley Video-to-Video Workflow": 5de07f68 --- - **サービス利用不可**：Moonvalley API サービスは現在提供されていません。これらのノードは非推奨となっており、正常に動作しない可能性があります。 - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -58,7 +56,6 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として

JSON 形式のワークフロー ファイルをダウンロード

- ### 2. ワークフロー実行手順 ![テキストから動画への生成ワークフロー](/images/tutorial/api_nodes/moonvalley/api_moonvalley_text_to_video.jpg) @@ -69,7 +66,6 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として 4. `Run` ボタンをクリックするか、ショートカット `Ctrl（Cmd）+ Enter` を押して動画生成を開始します 5. API から結果が返却された後、`Save Video` ノードで生成された動画を確認できます。また、動画は `ComfyUI/output/` ディレクトリにも保存されます - ## Moonvalley 画像から動画への生成（イメージ・トゥ・ビデオ）ワークフロー ### 1. ワークフロー ファイルのダウンロード @@ -99,7 +95,6 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として 5. `Run` ボタンをクリックするか、ショートカット `Ctrl（Cmd）+ Enter` を押して動画生成を開始します 6. API から結果が返却された後、`Save Video` ノードで生成された動画を確認できます。また、動画は `ComfyUI/output/` ディレクトリにも保存されます - ## Moonvalley 動画から動画への生成（ビデオ・トゥ・ビデオ）ワークフロー `Moonvalley Marey Video to Video` ノードでは、参照動画を入力して動画の再描画（リダーリング）を行うことができます。参照動画の動きや登場人物のポーズを活用して動画を生成できます。 @@ -124,7 +119,6 @@ Moonvalley Marey Realism v1.5 は、映画レベルの制作を目的として src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video_input.mp4" > - ### 2. ワークフロー実行手順 ![動画から動画への生成ワークフロー](/images/tutorial/api_nodes/moonvalley/api_moonvalley_video_to_video.jpg) diff --git a/ja/tutorials/partner-nodes/openai/dall-e-2.mdx b/ja/tutorials/partner-nodes/openai/dall-e-2.mdx index 765d3370a..0f56ae7be 100644 --- a/ja/tutorials/partner-nodes/openai/dall-e-2.mdx +++ b/ja/tutorials/partner-nodes/openai/dall-e-2.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/ja/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/openai/dall-e-3.mdx b/ja/tutorials/partner-nodes/openai/dall-e-3.mdx index e84d63407..d27e8db37 100644 --- a/ja/tutorials/partner-nodes/openai/dall-e-3.mdx +++ b/ja/tutorials/partner-nodes/openai/dall-e-3.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/ja/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -57,7 +56,6 @@ DALL·E 3 は OpenAI の最新の画像生成モデルであり、テキスト ![ComfyUI openai-dall-e-3 ワークフロー](/images/tutorial/api_nodes/openai/openai-dall-e-3/text2image.jpg) - 1. ComfyUI に **OpenAI DALL·E 3** ノードを追加します 2. プロンプトのテキストボックスに、生成したい画像の説明を入力します 3. 必要に応じて、オプションパラメーター（quality、style、size など）を調整します diff --git a/ja/tutorials/partner-nodes/openai/gpt-image-1.mdx b/ja/tutorials/partner-nodes/openai/gpt-image-1.mdx index be8c924fb..7bdca09b2 100644 --- a/ja/tutorials/partner-nodes/openai/gpt-image-1.mdx +++ b/ja/tutorials/partner-nodes/openai/gpt-image-1.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/ja/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx b/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx index 1cdaa3e1a..2ae29cd94 100644 --- a/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/ja/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Hybrid Pipelines": 9c6479f0 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/openrouter/llm.mdx b/ja/tutorials/partner-nodes/openrouter/llm.mdx index 4bfd0e510..c5300ce5e 100644 --- a/ja/tutorials/partner-nodes/openrouter/llm.mdx +++ b/ja/tutorials/partner-nodes/openrouter/llm.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Example workflow (`api_openrouter_llm`)": f1dbfe6a --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/overview.mdx b/ja/tutorials/partner-nodes/overview.mdx index ea50c540e..fcab7c974 100644 --- a/ja/tutorials/partner-nodes/overview.mdx +++ b/ja/tutorials/partner-nodes/overview.mdx @@ -17,7 +17,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Requirements from "/snippets/ja/tutorials/partner-nodes/requirements.mdx"; import Faq from "/snippets/ja/tutorials/partner-nodes/faq.mdx"; @@ -40,7 +39,6 @@ ComfyUI アカウント API キーによるログイン方法を学ぶ ![Comfy API キーによるログインを選択](/images/interface/setting/user/user-login-api-1.jpg) - ## 有料モデルのパートナー ノードを呼び出すための ComfyUI アカウント API キー統合の利用 現在、ComfyUI アカウント API キー統合を介して当社のサービスにアクセスし、有料モデルのパートナー ノードを呼び出すことをサポートしています。有料モデルのパートナー ノードを呼び出すために API キー統合を利用する方法については、API キー統合セクションをご参照ください。 @@ -53,8 +51,6 @@ ComfyUI アカウント API キーによるログイン方法を学ぶ 有料モデルのパートナー ノードを呼び出すために API キー統合を利用する方法については、API キー統合セクションをご参照ください - - ## パートナー ノードの利点 パートナー ノードは、ComfyUI ユーザーにとって以下の重要な利点を提供します： @@ -64,7 +60,6 @@ ComfyUI アカウント API キーによるログイン方法を学ぶ - **簡素化された体験**: API キーの管理や複雑な API リクエストの処理が不要 - **コストのコントロール**: プレペイド方式により、予期しない課金を防ぎ、支出を完全に管理可能 - ## 料金体系 @@ -94,7 +89,6 @@ python main.py --disable-api-nodes pause ``` - ## 利用例 パートナー ノードの強力な応用例の一つは、外部モデルの出力をローカルのノードと組み合わせることです。例えば： diff --git a/ja/tutorials/partner-nodes/pricing.mdx b/ja/tutorials/partner-nodes/pricing.mdx index 9640a750c..99f43cb06 100644 --- a/ja/tutorials/partner-nodes/pricing.mdx +++ b/ja/tutorials/partner-nodes/pricing.mdx @@ -45,7 +45,6 @@ translationBlockHashes: "Cloud GPU": 103c55e6 --- - 以下の表は、現在のパートナーノードの料金を一覧表示したものです。すべての価格はクレジット単位です。 ## Anthropic diff --git a/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx b/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx index 68c4e9254..865366c52 100644 --- a/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/ja/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Additional notes": d5f3f109 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/reve/reve-image.mdx b/ja/tutorials/partner-nodes/reve/reve-image.mdx index c7eeb597f..6d762001b 100644 --- a/ja/tutorials/partner-nodes/reve/reve-image.mdx +++ b/ja/tutorials/partner-nodes/reve/reve-image.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Available nodes": 5927bf70 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/rodin/model-generation.mdx b/ja/tutorials/partner-nodes/rodin/model-generation.mdx index dc22a566f..28df13336 100644 --- a/ja/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ja/tutorials/partner-nodes/rodin/model-generation.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Other Related Nodes": c885ce31 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/runway/image-generation.mdx b/ja/tutorials/partner-nodes/runway/image-generation.mdx index 1dbd1174f..4c0f73bf3 100644 --- a/ja/tutorials/partner-nodes/runway/image-generation.mdx +++ b/ja/tutorials/partner-nodes/runway/image-generation.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Runway Image Reference-to-Image Workflow": cc84e030 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/runway/video-generation.mdx b/ja/tutorials/partner-nodes/runway/video-generation.mdx index e85367f8e..39e44ecde 100644 --- a/ja/tutorials/partner-nodes/runway/video-generation.mdx +++ b/ja/tutorials/partner-nodes/runway/video-generation.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "First-Last Frame Video Generation Workflow": b7f1102b --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/sonilo/video-to-music.mdx b/ja/tutorials/partner-nodes/sonilo/video-to-music.mdx index d4b80f834..6d646e8c8 100644 --- a/ja/tutorials/partner-nodes/sonilo/video-to-music.mdx +++ b/ja/tutorials/partner-nodes/sonilo/video-to-music.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Use cases": 90983ea2 --- - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/tripo/model-generation.mdx b/ja/tutorials/partner-nodes/tripo/model-generation.mdx index 7231a5e78..9ba921558 100644 --- a/ja/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ja/tutorials/partner-nodes/tripo/model-generation.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Subsequent Task Processing for the Same Task": d24582e4 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx index 423619b6a..399d3996e 100644 --- a/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/ja/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Available Workflows": 5ec1d9c5 --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; @@ -78,4 +76,3 @@ Tripo 3.1 は、ComfyUI の Tripo パートナーノードで利用可能な最 | マテリアル出力 | PBR 対応 | 標準マテリアルマップ | | 最適化サポート | 対応 | 対応（Refine Draft は v1.4 のみ） | - diff --git a/ja/tutorials/partner-nodes/wan/wan2-7.mdx b/ja/tutorials/partner-nodes/wan/wan2-7.mdx index 453f1d8fb..e3464899e 100644 --- a/ja/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/ja/tutorials/partner-nodes/wan/wan2-7.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Wan2.7 video edit": 0d2511bb --- - - import ReqHint from "/snippets/ja/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; diff --git a/ja/tutorials/utility/depth-anything-3.mdx b/ja/tutorials/utility/depth-anything-3.mdx index d49627577..e7b81f645 100644 --- a/ja/tutorials/utility/depth-anything-3.mdx +++ b/ja/tutorials/utility/depth-anything-3.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Community Resources": 8c5d05ab --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # ComfyUI Depth Anything 3 概要 @@ -54,80 +53,103 @@ ComfyUI/ │ │ └── depth_anything_3_metric_large.safetensors ``` -## サンプルワークフロー +## Example Workflows ---- +### Depth Anything 3: Image Depth Estimation (`utility_depth_anything3_image_depth_estimation`) -## 1. 画像深度推定 +Upload one image and generate a depth map using Depth Anything 3. View a side-by-side comparison of the original image and depth output. -**機能説明：** 1 枚の画像をアップロードし、**Image Depth Estimation (Depth Anything 3)** を実行して深度マップを生成します。**Depth Preview** に元画像と深度出力のサイドバイサイド比較が表示されます。 +Depth Anything 3 image depth estimation workflow preview + + Open in Comfy Cloud + - JSON をダウンロード または テンプレートライブラリで "Depth Anything 3" を検索 + Download JSON or search "Depth Anything 3: Image Depth Estimation" in Template Library - - このワークフローのサンプル入力画像を取得 + + +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 85 · `retro_futuristic_home.png` -
- 画像深度推定出力 - 画像深度推定比較 +
+ Input image
-### 実行手順 +### Steps to Run -1. **LoadImage** — 入力画像を読み込む -2. **LoadDA3Model** — Depth Anything 3 バリアントを選択 -3. **実行** — Queue をクリックするか `Cmd+Enter` を押す -4. ワークフローが深度マップと並列比較を出力 +1. **LoadImage** — load your input image +2. **LoadDA3Model** — select a Depth Anything 3 variant +3. **Run** — click Queue or use `Cmd+Enter` +4. The workflow outputs a depth map and side-by-side comparison - - このワークフローはモジュール処理にサブグラフノードを使用しています。サブグラフのカスタマイズと拡張についてはサブグラフのドキュメントをご覧ください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ---- +### Depth Anything 3: Video Depth Estimation (`utility_depth_anything3_video_depth_estimation`) -## 2. 動画深度推定 +Upload a video to generate a per-frame depth sequence. Inside the subgraph, **GetVideoComponents** splits the input video into frames, **LoadDA3Model** loads the model, and **SetVideoComponents** reassembles the depth frames back into a video output. -**機能説明：** 動画をアップロードし、**Video Depth Estimation (Depth Anything 3)** を実行してフレームごとの深度シーケンスを生成します。サブグラフ内では **GetVideoComponents** が入力動画をフレームに分割し、**LoadDA3Model** がモデルを読み込み、**SetVideoComponents** が深度フレームを動画に再構成します。 +Depth Anything 3 video depth estimation workflow preview + + Open in Comfy Cloud + - JSON をダウンロード または テンプレートライブラリで "Depth Anything 3" を検索 + Download JSON or search "Depth Anything 3: Video Depth Estimation" in Template Library - - Comfy Cloud で開く + + +**入力素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 87 · `empty_room_assembly.mp4` -![動画深度推定プレビュー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_depth_anything3_video_depth_estimation-1.webp) +
+ +
-### 実行手順 +### Steps to Run -1. **LoadVideo** — 入力動画を読み込む -2. **モデルを選択** — **Small**、**Base**、**Mono-Large**、**Metric-Large** から選択 -3. **実行** — Queue をクリックするか `Cmd+Enter` を押す -4. ワークフローがフレームごとの深度マップ動画を出力 +1. **LoadVideo** — load your input video +2. **Select Model** — choose between **Small**, **Base**, **Mono-Large**, or **Metric-Large** +3. **Run** — click Queue or use `Cmd+Enter` +4. The workflow outputs a video with per-frame depth maps - - このワークフローはモジュール処理にサブグラフノードを使用しています。サブグラフのカスタマイズと拡張についてはサブグラフのドキュメントをご覧ください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## モデルバリアント +## Model Variants -| バリアント | head_type | 空検出 | 信頼度 | カメラデコーダ | 最適な用途 | -|-----------|-----------|:-------:|:------:|:--------------:|------------| -| **Small** | dualdpt | ❌ | ✅ | ✅ | 高速推論、モバイル/エッジ | -| **Base** | dualdpt | ❌ | ✅ | ✅ | バランスの取れた性能 | -| **Mono-Large** | dpt | ✅ | ❌ | ❌ | 空検出対応の単眼深度 | -| **Metric-Large** | dpt | ✅ | ❌ | ❌ | メートル単位の物理深度 | +| Variant | head_type | has_sky | has_confidence | camera_decoder | Best for | +|---------|-----------|:-------:|:--------------:|:--------------:|----------| +| **Small** | dualdpt | ❌ | ✅ | ✅ | Fast inference, mobile/edge | +| **Base** | dualdpt | ❌ | ✅ | ✅ | Balanced performance | +| **Mono-Large** | dpt | ✅ | ❌ | ❌ | Monocular depth with sky detection | +| **Metric-Large** | dpt | ✅ | ❌ | ❌ | Physical metric depth in metres | -- **Small** と **Base** は `dualdpt` ヘッドタイプを使用し、信頼度推定とカメラデコーダをサポート（多視点アプリケーション向け）。 -- **Mono-Large** と **Metric-Large** は `dpt` ヘッドタイプを使用し、空検出に対応。Metric-Large はメートル単位の生深度を出力。 +- **Small** and **Base** use the `dualdpt` head type with confidence estimation and camera decoder support for multi-view applications. +- **Mono-Large** and **Metric-Large** use the `dpt` head type with sky detection. Metric-Large outputs raw depth in metres. -## コミュニティリソース +## Community Resources -- [Depth Anything 3 GitHub (ByteDance-Seed)](https://github.com/ByteDance-Seed/Depth-Anything-3) — 研究論文とコード -- [Comfy-Org/Depth-Anything-3](https://huggingface.co/Comfy-Org/Depth-Anything-3) — 公式 ComfyUI モデル重み +- [Depth Anything 3 GitHub (ByteDance-Seed)](https://github.com/ByteDance-Seed/Depth-Anything-3) — Research paper and code +- [Comfy-Org/Depth-Anything-3](https://huggingface.co/Comfy-Org/Depth-Anything-3) — Official ComfyUI model weights diff --git a/ja/tutorials/utility/face-detection/mediapipe.mdx b/ja/tutorials/utility/face-detection/mediapipe.mdx index 0044222f6..b99e67718 100644 --- a/ja/tutorials/utility/face-detection/mediapipe.mdx +++ b/ja/tutorials/utility/face-detection/mediapipe.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Community Resources": ec7aa0f5 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -31,33 +30,48 @@ MediaPipe Face Detection は ComfyUI にネイティブ対応しています（P > **対象範囲：** 顔検出のみ — BlazeFace + FaceMesh v2 + ARKit blendshape。手、姿勢、全身検出は含まれません。 -## MediaPipe Face Detection ワークフロー +## MediaPipe Face Detection Workflow -### 1. ワークフローのダウンロード +### Mediapipe: Image Face Detection (`utility_face_detection_mediapipe`) -ComfyUI を最新バージョンにアップデートし、`Workflow` → `Browse Templates` に移動して、Utility カテゴリから "Mediapipe: Image Face Detection" を探してください。 +Input an image and detect up to 6 facial landmarks per face, enabling ultrafast multi-face detection. - - ワークフローをダウンロード - +Mediapipe image face detection workflow preview - - クラウドで開く + + + Open in Comfy Cloud + + Download JSON or search "Mediapipe: Image Face Detection" in Template Library + + + +**入力素材** - - このワークフローの入力サンプル画像を取得 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 5 · `soft_neon_girl.png` + + +
+ Input image +
+ +### 1. Download the Workflow -![MediaPipe Face Detection プレビュー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_face_detection_mediapipe-1.webp) +Update your ComfyUI to the latest version, then go to `Workflow` → `Browse Templates` and find "Mediapipe: Image Face Detection" under the Utility category. -### 2. モデルのダウンロード +### 2. Download the Model -MediaPipe Face Detection モデルは [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe) でホストされています。 +The MediaPipe Face Detection model is hosted on the [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe). -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) -以下のディレクトリ構造に配置してください： +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -66,54 +80,54 @@ MediaPipe Face Detection モデルは [Comfy-Org MediaPipe model repository](htt └── mediapipe_face_fp32.safetensors ``` -### 3. ワークフローの使用方法 +### 3. Using the Workflow -このワークフローは **subgraph**（サブグラフ）ノードを使用して、顔検出、可視化、マスク生成を調整します。サブグラフは以下の制御パラメータを公開しています： +This workflow uses a **subgraph** node that orchestrates face detection, visualization, and mask generation. The subgraph exposes the following controls: - - このワークフローは Subgraph ノードを使用したモジュール処理を採用しています。Subgraph のドキュメントを参照して、ワークフローのカスタマイズと拡張方法を学んでください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -**サブグラフの入力：** +**Subgraph inputs:** -| 入力 | 説明 | -|------|------| -| **image** | 解析する入力画像バッチ | -| **face_landmarker** | オプション。空の場合は内蔵モデルローダーを使用。外部 `FACE_DETECTION_MODEL` 出力を接続して上書き可能 | +| Input | Description | +|-------|-------------| +| **image** | Input image batch to analyze | +| **face_landmarker** | Optional. Leave empty to use the built-in model loader. Connect an external `FACE_DETECTION_MODEL` output to override | -**サブグラフのパラメータ：** +**Subgraph parameters:** -| パラメータ | デフォルト | 説明 | -|-----------|:---------:|------| -| **model_name** | `mediapipe_face_fp32.safetensors` | `ComfyUI/models/detection/` 内のチェックポイント。不足している場合は上記のモデルをダウンロード | -| **detector_variant** | `short` | **short** — 近距離/大きな顔向けに調整（約 2 m）。**full** — より小さい/遠い顔もカバー（約 5 m）、低速。**both** — 両方の検出器を実行し、フレームごとにより多くの顔を見つけた方を採用（約 2 倍のコスト） | -| **num_faces** | `1` | フレームごとに返す最大顔数。`0` = 制限なし（検出されたすべてを返す）。範囲：0–16 | -| **custom_face_oval** | `false` | マスク出力に顔の輪郭領域を含める | -| **custom_lips** | `false` | マスクに唇を含める（他の有効領域と結合） | -| **custom_left_eye** | `false` | マスクに左目領域を含める | -| **custom_right_eye** | `false` | マスクに右目領域を含める | -| **custom_irises** | `false` | マスクに虹彩領域を含める | +| Parameter | Default | Description | +|-----------|:-------:|-------------| +| **model_name** | `mediapipe_face_fp32.safetensors` | Checkpoint in `ComfyUI/models/detection/`. If missing, download the model above | +| **detector_variant** | `short` | **short** — tuned for close/large faces (~2 m range). **full** — covers smaller/farther faces (~5 m), slower. **both** — runs both detectors and keeps whichever found more faces per frame (~2× cost) | +| **num_faces** | `1` | Maximum faces to return per frame. `0` = no cap (return all detected). Range: 0–16 | +| **custom_face_oval** | `false` | Include face-outline region in the mask output | +| **custom_lips** | `false` | Include lips in the mask (union with other enabled regions) | +| **custom_left_eye** | `false` | Include left eye region in the mask | +| **custom_right_eye** | `false` | Include right eye region in the mask | +| **custom_irises** | `false` | Include iris regions in the mask | -マスク切り替えは内部的にカスタムモードを使用します：チェックされた領域のみが塗りつぶされ、複数の ON 領域はフレームごとに 1 つのマスクに結合されます。 +Mask toggles use custom mode internally: only checked regions are filled; multiple ON regions are **unioned** into one mask per frame. -**サブグラフの出力：** +**Subgraph outputs:** -| 出力 | タイプ | 説明 | -|------|-------|------| -| **face_landmarks** | `FACE_LANDMARKS` | フレームごとの顔データ。478 の 2D/3D ランドマーク、ARKit-52 blendshape、メッシュトポロジデータを含む — 可視化ノードとマスクノードに供給 | -| **bboxes** | `BOUNDING_BOX` | 顔境界ボックス — `DrawBBoxes` ノードと互換 | -| **mask** | `MASK` | 有効な領域切り替えから生成されたバイナリマスク | +| Output | Type | Description | +|--------|------|-------------| +| **face_landmarks** | `FACE_LANDMARKS` | Per-frame faces with 478 2D/3D landmarks, ARKit-52 blendshapes, mesh topology data — feeds into visualization and mask nodes | +| **bboxes** | `BOUNDING_BOX` | Face bounding boxes — compatible with `DrawBBoxes` node | +| **mask** | `MASK` | Binary mask from the enabled region toggles | -### 4. ワークフローの実行 +### 4. Run the Workflow -1. モデルファイルが `ComfyUI/models/detection/` に配置されていることを確認 -2. `Load Image` ノードで画像を読み込み -3. 必要に応じて検出パラメータを調整 -4. `Queue` をクリックするか `Ctrl(Cmd) + Enter` で実行 -5. ワークフローはメッシュオーバーレイ、境界ボックス、マスクプレビューを出力 +1. Ensure the model file is placed in `ComfyUI/models/detection/` +2. Load an image in the `Load Image` node +3. Adjust detection parameters as needed +4. Click `Queue` or use `Ctrl(Cmd) + Enter` to run +5. The workflow outputs the mesh overlay, bounding boxes, and mask preview -## コミュニティリソース +## Community Resources -- [MediaPipe GitHub](https://github.com/google-ai-edge/mediapipe) — MediaPipe アップストリームフレームワーク -- [Comfy-Org/mediapipe](https://huggingface.co/Comfy-Org/mediapipe) — 公式 ComfyUI モデルウェイト -- [ComfyUI Subgraph ガイド](https://docs.comfy.org/ja/interface/features/subgraph) — サブグラフの仕組みを学ぶ +- [MediaPipe GitHub](https://github.com/google-ai-edge/mediapipe) — Upstream MediaPipe framework +- [Comfy-Org/mediapipe](https://huggingface.co/Comfy-Org/mediapipe) — Official ComfyUI model weights +- [ComfyUI Subgraph Guide](https://docs.comfy.org/interface/features/subgraph) — Learn how subgraphs work diff --git a/ja/tutorials/utility/image-upscale.mdx b/ja/tutorials/utility/image-upscale.mdx index e924b4c65..591e91231 100644 --- a/ja/tutorials/utility/image-upscale.mdx +++ b/ja/tutorials/utility/image-upscale.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Tips": a966ce44 --- - 本ガイドでは、ComfyUI における画像アップスケーリングのワークフローについて解説します。ローカルモデルおよび各種ユースケースに応じたパートナー・ノードの選択肢も紹介します。 diff --git a/ja/tutorials/utility/moge.mdx b/ja/tutorials/utility/moge.mdx index b73155ff4..38c56e9bf 100644 --- a/ja/tutorials/utility/moge.mdx +++ b/ja/tutorials/utility/moge.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Community Resources": d9fcd8cd --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # ComfyUI MoGe の紹介 @@ -62,75 +60,118 @@ ComfyUI/ │ │ └── moge_1_vitl_fp16.safetensors ``` -## ワークフロー例 - ---- +## Example Workflows -## 1. 深度推定 +### MoGe: Depth Estimation (`utility_moge_depth_estimation`) -**機能：** 単一画像からメートルスケールの深度マップ、カラー深度プレビュー、マスクを生成——MoGe が1回の推論で推定するメートルスケール深度をそのまま出力します。シーンの深度リファレンスとしてコンポジットや深度エフェクトに有用で、メッシュ生成の前処理としても使えます。 +Upload a single RGB image and generate a colored depth preview and raw depth map. -MoGe は画像からカメラFOVも自動推定します。必要に応じて実際のFOVを入力するとさらに精度が向上します。 +MoGe depth estimation workflow preview - - JSONをダウンロード、またはテンプレートライブラリで "MoGe Depth Estimation" を検索 + + Open in Comfy Cloud - - このワークフローで使用するサンプル入力画像を取得 + + Download JSON or search "MoGe: Depth Estimation" in Template Library -### 1.2 実行手順 -1. `LoadMoGeModel` ノードが MoGe チェックポイントをロードしていることを確認 -2. `Load Image` ノードに画像をロード -3. `Queue` ボタン、またはショートカット `Ctrl(cmd) + Enter` で実行 -4. カラー深度プレビュー、生深度プレビュー、マスクが出力されます +**入力素材** ---- +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 9 · `alien_world.png` + + + +### 1.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded a MoGe checkpoint +2. Load an image in the `Load Image` node +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run +4. The workflow outputs colored depth preview, raw depth preview, and a mask + +### MoGe: Perspective Geometry Estimation (`3d_moge_perspective_to_mesh`) -## 2. 透視写真を3Dメッシュに変換 +Upload an image to estimate its perspective geometry. Generate a 3D depth map and surface normals from the input, then convert to a textured GLB mesh. -**機能：** 単一の透視写真をテクスチャ付きGLBメッシュに変換し、法線と深度のプレビューも生成します。MoGe が可視シーンからポイントマップ、深度、法線を推定し、メッシュに変換します。これは**単眼幾何推定**であり、オクルージョン領域や物体の裏側は欠落や断片が生じます。シーンのラピッドプロトタイピングや参照ジオメトリ、深度/法線のメッシュ可視化には有用ですが、マルチビュー3D再構築の代替にはなりません。 +MoGe perspective to mesh workflow preview - - JSONをダウンロード、またはテンプレートライブラリで "3D MoGe Perspective to Mesh" を検索 + + Open in Comfy Cloud - - このワークフローで使用するサンプル入力画像を取得 + + Download JSON or search "MoGe: Perspective Geometry Estimation" in Template Library -### 2.1 実行手順 -1. `LoadMoGeModel` ノードが MoGe チェックポイントをロードしていることを確認 -2. `Load Image` ノードに透視写真をロード -3. （オプション）OpenGL および DirectX の法線プレビューを表示 -4. `Queue` または `Ctrl(cmd) + Enter` で実行 +**入力素材** ---- +Upload this file to the matching `LoadImage` node: -## 3. パノラマをメッシュに変換 + + + `LoadImage` node 9 · `modern_living_room.png` + + -**機能：** 360°パノラマ（正距円筒図法）をテクスチャ付きGLBメッシュに変換します。`MoGePanoramaInference` がパノラマを12の視点に分割し、それぞれで独立して単眼幾何推定を実行、単一のメッシュに統合します。各セグメントは単一視点からの推定であるため、結果は大まかなシーン再構築——360°シーンの空間概要を得るには有用ですが、オクルージョン領域や表面背後は欠落や断片が生じます。 +
+ Input image +
+ +This is **monocular geometry estimation**: occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. + +### 2.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded a MoGe checkpoint +2. Load a perspective photo in the `Load Image` node +3. (Optional) View the OpenGL and DirectX normal previews +4. Click `Queue` or use `Ctrl(cmd) + Enter` to run + +### Moge: Panorama to Mesh (`3d_moge_panorama_to_mesh`) + +Upload an equirectangular 360° panorama image and generate a textured GLB mesh with vertex colors. + +MoGe panorama to mesh workflow preview - - JSONをダウンロード、またはテンプレートライブラリで "3D MoGe Panorama to Mesh" を検索 + + Open in Comfy Cloud - - このワークフローで使用するサンプル入力画像を取得 + + Download JSON or search "Moge: Panorama to Mesh" in Template Library -### 3.1 実行手順 -1. `LoadMoGeModel` ノードが MoGe チェックポイントをロードしていることを確認 -2. `Load Image` ノードに正距円筒図法のパノラマ画像をロード -3. `Queue` ボタン、または `Ctrl(cmd) + Enter` で実行 +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 28 · `lego_street_panorama.png` + + + +
+ Input panorama +
+ +The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, run monocular geometry estimation on each view independently, then merge them into a single mesh. + +### 3.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded one of the MoGe checkpoints +2. Load an equirectangular panorama image in the `Load Image` node +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run --- -## コミュニティリソース +## Community Resources -- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe) — 研究論文とコード -- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe) — 公式 ComfyUI モデルウェイト +- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe): Research paper and code +- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe): Official ComfyUI model weights diff --git a/ja/tutorials/utility/pose-detection-sdpose.mdx b/ja/tutorials/utility/pose-detection-sdpose.mdx index a63be80a2..4a7699880 100644 --- a/ja/tutorials/utility/pose-detection-sdpose.mdx +++ b/ja/tutorials/utility/pose-detection-sdpose.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": 409c3058 --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" @@ -31,69 +30,161 @@ SDPose + RT-DETRv4 は ComfyUI にネイティブサポートされています > **制限事項:** 検出精度は画像の解像度と被写体の視認性に依存します。極端に隠れている場合や非常に小さい被写体では、得られるキーポイントが少なくなることがあります。 -## SDPose ワークフロー +## SDPose Workflows -ユースケースに応じて4つのワークフローが利用可能です: +Four workflows are available depending on your use case: -| ワークフロー | 入力 | 出力 | 用途 | +| Workflow | Input | Output | Use Case | |----------|-------|--------|----------| -| 複数人 (画像) | 1枚の画像 | ポーズマップ + バウンディングボックス | 複数人が写った写真 | -| 複数人 (動画) | 動画 | フレーム単位のポーズマップ + バウンディングボックス | 動画のポーズトラッキング | -| OOD 画像からポーズ | 1枚の画像 | ポーズマップ | スタイル転送 / 画像からポーズ | -| OOD 動画からポーズマップ | 動画 | フレーム単位のポーズマップ | 動画からポーズアニメーション | +| Multi-Person (Image) | Single image | Pose map + BBoxes | Photos with multiple people | +| Multi-Person (Video) | Video | Per-frame pose map + BBoxes | Video pose tracking | +| OOD Image to Pose | Single image | Pose map | Style transfer / image-to-pose | +| OOD Video to Pose Map | Video | Per-frame pose map | Video-to-pose animation | + +### 1. Download Workflows + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find SDPose workflows under the Utility category. -### 1. ワークフローのダウンロード +### SDPose: Image Multi-Person Detection (`utility_sdpose_multi_person`) -ComfyUIを最新バージョンにアップデートし、`Workflow` → `Browse Templates` から、UtilityカテゴリにあるSDPoseワークフローを探してください。 +Upload an image to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose image multi-person detection workflow preview - - Run in Comfy Cloud + + Open in Comfy Cloud - - JSONをダウンロード + + Download JSON or search "SDPose: Image Multi-Person Detection" in Template Library +**入力素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 679 · `group_photo.png` + + + +
+ Input image +
+ +### SDPose: Video Multi-Person Detection (`utility_sdpose_multi_person_video`) + +Upload a video to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose video multi-person detection workflow preview + - - Run in Comfy Cloud + + Open in Comfy Cloud - - JSONをダウンロード + + Download JSON or search "SDPose: Video Multi-Person Detection" in Template Library +**入力素材** + +Upload this file to the matching `LoadVideo` node: + - - Run in Comfy Cloud + + `LoadVideo` node 694 · `man_playing_violin.mp4` - - JSONをダウンロード + + +
+ +
+ +### SDPose-OOD: Image to Pose Map (`utility_sdpose_ood_image_to_pose`) + +Upload an image to extract pose keypoints and generate a corresponding pose map using the SDPose-OOD model. + +SDPose-OOD image to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Image to Pose Map" in Template Library +**入力素材** + +Upload this file to the matching `LoadImage` node: + - - Run in Comfy Cloud + + `LoadImage` node 667 · `dancer.png` - - JSONをダウンロード + + +**出力例** + +
+ Input image + SDPose-OOD image to pose map example output +
+ +### SDPose-OOD: Video to Pose Map (`utility_sdpose_ood_video_to_pose_map`) + +Upload a video to extract pose keypoints and generate a pose map. The workflow supports multiple person detection and uses an enhanced SDPose model for accurate whole-body feature extraction. + +SDPose-OOD video to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Video to Pose Map" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 679 · `man_in_the_rain.mp4` -### 2. モデルのダウンロード +**出力例** + +
+ + +
+ +### 2. Download Models -SDPoseとRT-DETRv4のモデルチェックポイントは、[Comfy-Org SDPose モデルリポジトリ](https://huggingface.co/Comfy-Org/SDPose) で公開されています。 +The SDPose and RT-DETRv4 model checkpoints are hosted on the [Comfy-Org SDPose model repository](https://huggingface.co/Comfy-Org/SDPose). -**checkpoints** (SDPoseモデル): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +**checkpoints** (SDPose model): +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) -**diffusion_models** (RT-DETRv4検出器): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (推奨) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (完全精度、サイズ大) +**diffusion_models** (RT-DETRv4 detector): +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (recommended) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (full precision, larger) -以下のディレクトリ構成に配置してください: +Place them in the following directory structure: ``` 📂 ComfyUI/ @@ -105,44 +196,44 @@ SDPoseとRT-DETRv4のモデルチェックポイントは、[Comfy-Org SDPose └── rt_detr_v4-x-hgnet_fp32.safetensors ``` -### 3. ワークフローの使い方 +### 3. Using the Workflows -#### 複数人 (画像) +#### Multi-Person (Image) -- **入力** — `Load Image` ノードで画像を読み込みます。1人以上の人物が写った画像を使用してください (例: `group_photo.png`)。 -- **検出** — `Image to Pose Map (SDPose Multi-Person)` サブグラフが画像を処理し、以下を出力します: - - **IMAGE** — 画像に重ね合わされたポーズスケルトンの可視化 - - **keypoints** — 生の全身キーポイントデータ - - **bboxes** — バウンディングボックス座標 -- **描画オプション** — 描画する身体部位の設定: - - `draw_body`、`draw_hands`、`draw_face`、`draw_feet` — 表示の切り替え - - `stick_width`、`face_point_size` — 視覚スタイルの調整 - - `score_threshold` — キーポイント表示の最小信頼度 -- **検出オプション**: - - `resize_type.longer_size` — 検出前に長辺のサイズをスケーリング - - `max_detections` — 検出する最大人数 - - `detect_threshold` — 検出の信頼度しきい値 - - `detect_class` — 検出するオブジェクトクラス (デフォルト: person) +- **Input** — Load an image via the `Load Image` node. Use an image with one or more people (example: `group_photo.png`). +- **Detection** — The `Image to Pose Map (SDPose Multi-Person)` subgraph processes the image and outputs: + - **IMAGE** — pose skeleton visualization overlaid on the image + - **keypoints** — raw whole-body keypoint data + - **bboxes** — bounding box coordinates +- **Drawing Options** — Configure which body parts to draw: + - `draw_body`, `draw_hands`, `draw_face`, `draw_feet` — toggle visibility + - `stick_width`, `face_point_size` — adjust visual style + - `score_threshold` — minimum confidence for displaying keypoints +- **Detection Options**: + - `resize_type.longer_size` — scale the longer dimension before detection + - `max_detections` — maximum number of people to detect + - `detect_threshold` — detection confidence threshold + - `detect_class` — object class to detect (default: person) -#### 複数人 (動画) +#### Multi-Person (Video) -画像ワークフローと同様ですが、動画のフレームを順次処理します。動画ファイルの入力には `Load Video` を、結果の出力には `Save Video` を使用してください。 +Same as the image workflow but processes video frames sequentially. Use `Load Video` to input a video file and `Save Video` to export the result. -#### OOD 画像からポーズ +#### OOD Image to Pose -SDPoseモデルを使用して、バウンディングボックス表示なしで画像からクリーンなポーズマップを生成します。スタイル転送で、ある画像からスケルトンポーズを抽出して別の画像に適用したい場合に便利です。 +Uses the SDPose model to generate a clean pose map from an image, without bounding box visualization. This is useful for style transfer where you want to extract the skeleton pose from one image and apply it to another. -#### OOD 動画からポーズマップ +#### OOD Video to Pose Map -動画からフレームごとのポーズマップを生成します。出力は、抽出されたポーズスケルトンを含む各フレームの動画ファイルで、下流のアニメーションやControlNetワークフローに適しています。 +Generates per-frame pose maps from a video. The output is a video file where each frame contains the extracted pose skeleton, suitable for downstream animation or ControlNet workflows. - - これらのワークフローはモジュール処理のためにSubgraphノードを使用しています。ワークフローをカスタマイズして拡張する方法については、Subgraphのドキュメントをご覧ください。 + + These workflows use Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflows. -## 補足情報 +## Additional Notes -- **モデルディレクトリ**: SDPoseチェックポイントは `models/checkpoints/` に、RT-DETRv4検出器は `models/diffusion_models/` に配置します -- **入力画像の例**: テスト用に、ワークフローテンプレートの `input/` ディレクトリに `group_photo.png` ファイルが用意されています -- **キーポイント出力**: POSE_KEYPOINTタイプは、条件付き生成のためにポーズデータを受け付ける下流ノードに接続できます -- **アップデート必須**: SDPose + RT-DETRv4のサポートは最新版のComfyUIで利用可能です。ComfyUIが最新であることを確認してください +- **Model directory** — the SDPose checkpoint goes in `models/checkpoints/`, and the RT-DETRv4 detector goes in `models/diffusion_models/` +- **Input image example** — the `group_photo.png` file is available in the workflow template's `input/` directory for testing +- **Keypoint output** — the POSE_KEYPOINT type can be connected to downstream nodes that accept pose data for conditional generation +- **Update required** — SDPose + RT-DETRv4 support is available in recent ComfyUI versions. Make sure your ComfyUI is up to date. diff --git a/ja/tutorials/utility/preprocessors.mdx b/ja/tutorials/utility/preprocessors.mdx index 542b935df..be8e9277e 100644 --- a/ja/tutorials/utility/preprocessors.mdx +++ b/ja/tutorials/utility/preprocessors.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Normals extraction": 65c791bb --- - ## プリプロセッサーとは？ @@ -27,82 +26,175 @@ translationBlockHashes: - デバッグやチューニングが容易 - 画像および動画の結果がより予測可能に -## 深度推定 +## Depth estimation + +Depth estimation converts a flat image into a depth map representing relative distance within a scene. This structural signal is foundational for controlled generation, spatially aware edits, and relighting workflows. + +This workflow emphasizes: +- Clean, stable depth extraction +- Consistent normalization for downstream use +- Easy integration with ControlNet and image-edit pipelines + +Depth outputs can be reused across multiple passes, making it easier to iterate without re-running expensive upstream steps. + +### Video to Depth Map (`utility-depthAnything-v2-relative-video`) + +Convert a video to a temporally stable depth map. + +Video to Depth Map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Depth Map" in Template Library + + + +**入力素材** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 6 · `roller_coaster.mp4` + + + +
+ +
+ +## Lineart conversion + +Lineart preprocessors distill an image down to its essential edges and contours, removing texture and color while preserving structure. + +This workflow is designed to: +- Produce clean, high-contrast lineart +- Minimize broken or noisy edges +- Provide reliable structural guidance for stylization and redraw workflows + +Lineart pairs especially well with depth and pose, offering strong structural constraints without overconstraining style. + +### Video to Lineart / Canny (`utility-lineart-video`) + +Convert a video to lineart or Canny edges for control processors. + +Video to Lineart workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Lineart / Canny" in Template Library + + + +**入力素材** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 2 · `utility-lineart-video-input.mp4` + + + +
+ +
+ +## Pose detection + +Pose detection extracts body keypoints and skeletal structure from images, enabling precise control over human posture and movement. -深度推定は、平面画像をシーン内の相対的な距離を表す深度マップへと変換します。この構造的信号は、制御された生成、空間認識型の編集、再ライティング（relighting）ワークフローの基盤となります。 +This workflow focuses on: +- Clear, readable pose outputs +- Stable keypoint detection suitable for reuse across frames +- Compatibility with pose-based ControlNet and animation pipelines -本ワークフローでは以下の点に重点を置いています： -- クリーンで安定した深度抽出 -- 後続処理での利用を想定した一貫した正規化 -- ControlNet や画像編集パイプラインとの容易な統合 +By isolating pose extraction into a dedicated workflow, pose data becomes easier to inspect, refine, and reuse. -深度出力は複数回の処理パスで再利用可能であるため、高コストな上流ステップを再実行することなく、反復作業が容易になります。 +### Video to Pose Map - OpenPose (`utility-openpose-video`) - - Comfy Cloud で実行 - +Convert a video to a temporally stable pose control map. - - JSON をダウンロード - +Video to Pose Map workflow preview -## ラインアート変換 + + + Open in Comfy Cloud + + + Download JSON or search "Video to Pose Map - OpenPose" in Template Library + + -ラインアートプリプロセッサーは、画像をその基本的なエッジおよび輪郭にまで簡略化し、テクスチャや色を除去しつつ、構造を保持します。 +**入力素材** -本ワークフローは以下の目的で設計されています： -- クリーンで高コントラストなラインアートの生成 -- 切れたりノイズを含んだエッジの最小化 -- スタイライゼーションおよび再描画ワークフロー向けの信頼性の高い構造的ガイドの提供 +Upload this file to the matching `VHS_LoadVideo` node: -ラインアートは、特に深度やポーズと組み合わせると非常に効果的であり、スタイルを過剰に制約することなく、強固な構造的制約を提供します。 + + + `VHS_LoadVideo` node 2 · `pose_input.mp4` + + - - Comfy Cloud で実行 - +
+ +
- - JSON をダウンロード - +## Normals extraction -## ポーズ検出 +Normals estimation converts a flat image into a surface normal map—a per-pixel direction field that describes how each part of a surface is oriented (typically encoded as RGB). This signal is useful for relighting, material-aware stylization, and highly structured edits. -ポーズ検出は、画像から人体のキーポイントおよび骨格構造を抽出し、人間の姿勢や動きを精密に制御できるようにします。 +This workflow emphasizes: +- Clean, stable normal extraction with minimal speckling +- Consistent orientation and normalization for reliable downstream use +- ControlNet-ready outputs for relighting, refinement, and structure-preserving edits +- Reuse across passes so you can iterate without re-running earlier steps -本ワークフローでは以下の点に焦点を当てています： -- クリアで読みやすいポーズ出力 -- フレーム間で再利用可能な安定したキーポイント検出 -- ポーズベースの ControlNet やアニメーションパイプラインとの互換性 +Normal outputs can be used to: +- Drive relight/shading changes while preserving geometry +- Add a stronger 3D-like structure to stylization and redraw pipelines +- Improve consistency across frames when paired with pose/depth for animation work -ポーズ抽出を専用のワークフローに分離することで、ポーズデータの確認・微調整・再利用が容易になります。 +### Video to Normal Map (`utility-normal_crafter-video`) - - Comfy Cloud で実行 - +Convert a video to a temporally stable normal map. - - JSON をダウンロード - +Video to Normal Map workflow preview -## 法線抽出 + + + Open in Comfy Cloud + + + Download JSON or search "Video to Normal Map" in Template Library + + -法線推定は、平面画像を表面法線マップ（各ピクセルの向きを表す方向場。通常 RGB で符号化される）へと変換します。この信号は、再ライティング、マテリアル認識型のスタイライゼーション、および高度に構造化された編集に有用です。 +**入力素材** -本ワークフローでは以下の点に重点を置いています： -- スペックル（斑点ノイズ）を最小限に抑えたクリーンで安定した法線抽出 -- 後続処理での信頼性を確保するための一貫した向き・正規化 -- 再ライティング、精緻化、構造保持型編集に即時対応可能な ControlNet 対応出力 -- 処理パス間での再利用により、初期ステップを再実行せずに反復作業が可能 +Upload this file to the matching `VHS_LoadVideo` node: -法線出力は以下のような用途に活用できます： -- 幾何形状を維持したまま、再ライティング／シェーディングの変更を駆動 -- スタイライゼーションおよび再描画パイプラインに、より強い 3D 的構造を付与 -- アニメーション作業において、ポーズ／深度と併用することでフレーム間の一貫性を向上 + + + `VHS_LoadVideo` node 3 · `normals_input.mp4` + + - - Comfy Cloud で実行 - +
+ +
- - JSON をダウンロード - diff --git a/ja/tutorials/utility/remove-background-birefnet.mdx b/ja/tutorials/utility/remove-background-birefnet.mdx index c0f4cc3ff..b7388b942 100644 --- a/ja/tutorials/utility/remove-background-birefnet.mdx +++ b/ja/tutorials/utility/remove-background-birefnet.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": d07d5c84 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -30,27 +29,44 @@ BiRefNet は ComfyUI でネイティブサポートされており（PR [#12747] > **制限事項：** 背景が極めて雑然としている場合や、被写体が背景に溶け込んでいる場合は、マスクの精度が低下する可能性があります。1 回に 1 枚の画像を処理します。 -## BiRefNet 背景除去ワークフロー +## BiRefNet Background Removal Workflow + +### 1. Download Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "BiRefNet: Remove Background" under the Utility category. + +### BiRefNet: Remove Background (`utility_birefnet_remove_background`) -### 1. ワークフローをダウンロード +Upload an image with any background. Generate a version with the background removed and a precision segmentation mask. -ComfyUI を最新バージョンにアップデートし、メニューから `ワークフロー` -> `テンプレートを閲覧` に進み、Utility カテゴリから "BiRefNet: Remove Background" を見つけてください。 +BiRefNet remove background workflow preview - - ワークフローをダウンロード + + + Open in Comfy Cloud + + Download JSON or search "BiRefNet: Remove Background" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadImage` node: - - Open in cloud + + + `LoadImage` node 17 · `the_lily_veil.png` + -### 2. モデルをダウンロード +### 2. Download Models -BiRefNet モデルは [Comfy-Org BiRefNet モデルリポジトリ](https://huggingface.co/Comfy-Org/BiRefNet) でホストされています。 +The BiRefNet model is hosted on the [Comfy-Org BiRefNet model repository](https://huggingface.co/Comfy-Org/BiRefNet). -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) -以下のディレクトリ構造に配置してください： +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -59,21 +75,21 @@ BiRefNet モデルは [Comfy-Org BiRefNet モデルリポジトリ](https://hugg └── birefnet.safetensors ``` -### 3. ワークフローの使い方 +### 3. Using the Workflow -- **画像** — `Load Image` ノードで画像を読み込み（ComfyUI の `input/` フォルダに配置） -- `Remove Background (BiRefNet)` サブグラフが画像を処理し、以下を出力： - - **IMAGE** — 透明な背景を持つ RGBA 結果 - - **mask** — 抽出された前景マスク +- **Image** — Load an image via the `Load Image` node (place it in the ComfyUI `input/` folder) +- The `Remove Background (BiRefNet)` subgraph processes the image and outputs: + - **IMAGE** — the result with a transparent background (RGBA) + - **mask** — the extracted foreground mask -出力はプレビュー可能で、他のノードへの入力として合成、編集、保存に使用できます。 +Outputs can be previewed and used as inputs to other nodes for compositing, further editing, or saving. - - 本ワークフローは Subgraph ノードを使用してモジュール化された処理を実現しています。Subgraph ドキュメントを参照してワークフローのカスタマイズと拡張方法を学んでください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 補足説明 +## Additional Notes -- **モデルディレクトリ** — モデルは `ComfyUI/models/background_removal/` に配置する必要があります（`checkpoints` フォルダではありません） -- **更新が必要** — BiRefNet を利用するには最新バージョンの ComfyUI が必要です -- **RGBA 出力** — 透明背景の結果は新しい背景に直接合成でき、ダウンストリームワークフローでも使用可能です +- **Model directory** — the model must be placed in `ComfyUI/models/background_removal/`, not the `checkpoints` folder +- **Update required** — BiRefNet support is available in recent ComfyUI versions. Make sure your ComfyUI is up to date. +- **RGBA output** — the transparent background result can be directly composited onto new backgrounds or used in downstream workflows diff --git a/ja/tutorials/utility/seedvr2.mdx b/ja/tutorials/utility/seedvr2.mdx index fc739f1d9..63d921c71 100644 --- a/ja/tutorials/utility/seedvr2.mdx +++ b/ja/tutorials/utility/seedvr2.mdx @@ -57,79 +57,137 @@ ComfyUI/ │ └── seedvr2_ema_vae_fp16.safetensors ``` -## 1. 画像のアップスケール（3B INT8） +## Example Workflows -**機能:** 一枚の画像を SeedVR2 3B INT8 モデルを使用してアップスケールします。INT8 量子化バリアントは品質と VRAM 使用量のバランスに優れています。 +### SeedVR2 3B Int8: Upscale Image (`utility_seedvr2_3b_int8_upscale_image`) - - JSON をダウンロードするか、テンプレートライブラリで SeedVR2 3B Int8: Upscale Image を検索してください +Upscale images using SeedVR2 3B Int8, a one-step diffusion-based video restoration model that produces high-quality results with improved temporal consistency. + +SeedVR2 3B Int8 upscale image workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Image" in Template Library + + +**入力素材** + +Upload this file to the matching `LoadImage` node: - - このワークフローのサンプル入力画像を入手してください + + + `LoadImage` node 1 · `watch_macro_shot.png` + -![SeedVR2 3B INT8 アップスケールプレビュー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_3b_int8_upscale_image.png) +**出力例** -### 1.1 実行手順 +
+ Input image + SeedVR2 3B Int8 upscale example output +
-1. 画像を `ComfyUI/input/` フォルダーに配置し、`Load Image` ノードでその画像を選択します -2. SeedVR2 モデルローダーで `seedvr2_3b_int8_convrot.safetensors` チェックポイントを選択します -3. `Queue` をクリックするか、`Ctrl(cmd) + Enter` を使用して実行します +### 1.1 Steps to Run ---- +1. Place your image in the `ComfyUI/input/` folder and select it in the `Load Image` node +2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run + +### SeedVR2 7B Int8: Upscale Image (`utility_seedvr2_7b_int8_upscale_image`) -## 2. 画像のアップスケール（7B INT8） +Upscale images using SeedVR2 7B Int8, a one-step diffusion model that enhances resolution through adversarial training and adaptive window attention. -**機能：** SeedVR2 7B INT8 モデルを使用して単一画像をアップスケールします。より大きな 7B モデルは INT8 量子化により効率的な VRAM 使用で高品質な結果を提供します。 +SeedVR2 7B Int8 upscale image workflow preview - - JSON をダウンロードするか、テンプレートライブラリで SeedVR2 7B Int8: Upscale Image を検索してください + + + Open in Comfy Cloud + + Download JSON or search "SeedVR2 7B Int8: Upscale Image" in Template Library + + + +**入力素材** - - このワークフローのサンプル入力画像を入手してください +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 1 · `indoor_portrait.png` + -![SeedVR2 7B INT8 アップスケールプレビュー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_7b_int8_upscale_image.png) +**出力例** -### 2.1 実行手順 +
+ Input image + SeedVR2 7B Int8 upscale example output +
-1. 画像を `ComfyUI/input/` フォルダーに配置し、`Load Image` ノードでその画像を選択します -2. SeedVR2 モデルローダーで `seedvr2_7b_int8_convrot.safetensors` チェックポイントを選択します -3. `Queue` をクリックするか、`Ctrl(cmd) + Enter` を使用して実行します +### 2.1 Steps to Run ---- +1. Place your image in the `ComfyUI/input/` folder and select it in the `Load Image` node +2. Select the `seedvr2_7b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run -## 3. ビデオアップスケール（3B INT8） +### SeedVR2 3B Int8: Upscale Video (`utility_seedvr2_3b_int8_upscale_video`) -**機能：** SeedVR2 3B INT8 モデルを使用してビデオをアップスケールします。このワークフローは解像度を向上させながらフレーム間の時間的一貫性を維持します。古い映像や低解像度ビデオの復元に最適です。 +Upscale and restore video footage using SeedVR2 3B Int8, a one-step diffusion model that enhances resolution while maintaining temporal consistency across frames. - - JSON をダウンロードするか、テンプレートライブラリで SeedVR2 3B Int8: Upscale Video を検索してください +SeedVR2 3B Int8 upscale video workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Video" in Template Library + - - このワークフロー用のサンプル入力ビデオを入手する +**入力素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 73 · `grainy_perfume_shot_crf32.mp4` + + +**出力例** + +
+ + +
-### 3.1 実行手順 +### 3.1 Steps to Run -1. ビデオを `ComfyUI/input/` フォルダーに配置し、`Load Video` ノードでそれを選択します -2. SeedVR2 モデルローダーで `seedvr2_3b_int8_convrot.safetensors` チェックポイントを選択します -3. `Queue` をクリックするか、`Ctrl(cmd) + Enter` を使用して実行します +1. Place your video in the `ComfyUI/input/` folder and select it in the `Load Video` node +2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run -### パフォーマンス +### Performance -高い目標解像度では処理時間が長くなります。INT8 バリアントは FP16 と比較して VRAM 使用量を抑え、効率的な推論を提供します。 +Higher target resolutions require more processing time. The INT8 variant provides efficient inference with reduced VRAM usage compared to FP16. --- -## コミュニティリソース +## Community Resources -- [SeedVR2 プロジェクトページ](https://iceclear.github.io/projects/seedvr2/): 公式サイト -- [ByteDance SeedVR コードベース (GitHub)](https://github.com/ByteDance-Seed/SeedVR): 研究コードと論文 -- [Comfy-Org/SeedVR2 (HuggingFace)](https://huggingface.co/Comfy-Org/SeedVR2): ComfyUI モデルウェイト -- [ByteDance-Seed/SeedVR2-3B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-3B): オリジナル 3B モデルウェイト -- [ByteDance-Seed/SeedVR2-7B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-7B): オリジナル 7B モデルウェイト -- [論文 (arXiv)](https://arxiv.org/abs/2506.05301) +- [SeedVR2 Project Page](https://iceclear.github.io/projects/seedvr2/): Official project website +- [ByteDance SeedVR Codebase (GitHub)](https://github.com/ByteDance-Seed/SeedVR): Original research code and paper +- [Comfy-Org/SeedVR2 (HuggingFace)](https://huggingface.co/Comfy-Org/SeedVR2): Official ComfyUI model weights +- [ByteDance-Seed/SeedVR2-3B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-3B): Original 3B model weights +- [ByteDance-Seed/SeedVR2-7B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-7B): Original 7B model weights +- [Paper (arXiv)](https://arxiv.org/abs/2506.05301) diff --git a/ja/tutorials/utility/video-segment-sam3.mdx b/ja/tutorials/utility/video-segment-sam3.mdx index 6ac27d505..79318eab4 100644 --- a/ja/tutorials/utility/video-segment-sam3.mdx +++ b/ja/tutorials/utility/video-segment-sam3.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": c38c57a0 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -38,39 +37,79 @@ SAM 3.1 はテキストプロンプトに基づいて動画フレーム全体で > **制限事項：** テキストプロンプトのトークン上限は 32 トークンです。最良の結果を得るには、プロンプトを短く具体的に保ってください。 -## SAM 3.1 セグメンテーションワークフロー +## SAM 3.1 Segment Workflows + +### 1. Download Workflow -### 1. ワークフローをダウンロード +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find the SAM 3.1 workflows under the Utility category. -ComfyUI を最新バージョンにアップデートし、メニューから `ワークフロー` -> `テンプレートを閲覧` に進み、Utility カテゴリから SAM 3.1 ワークフローを見つけてください。 +### SAM3: Video Segmentation (`utility_video_segment_sam3`) -**動画セグメンテーション：** +Use the SAM3 model to segment the main subject or content from a video, isolating specific objects or regions. - - 動画ワークフローをダウンロード +SAM3 video segmentation workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SAM3: Video Segmentation" in Template Library + - - Open in cloud +**入力素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 115 · `drinking_unicorn.mp4` + + +**出力例** + + + +### SAM3: Image Segmentation (`utility_image_segment_sam3`) + +Use the SAM3 model to segment the main subject or content from a photo or image, isolating specific objects or regions. -**画像セグメンテーション：** +SAM3 image segmentation workflow preview - - 画像ワークフローをダウンロード + + + Open in Comfy Cloud + + Download JSON or search "SAM3: Image Segmentation" in Template Library + + + +**入力素材** - - Open in cloud +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 79 · `neon_guitarist.png` + + +
+ Input image +
-### 2. モデルをダウンロード +### 2. Download Models -SAM 3.1 モデルは [Comfy-Org SAM 3.1 モデルリポジトリ](https://huggingface.co/Comfy-Org/sam3.1) でホストされています。 +The SAM 3.1 model is hosted on the [Comfy-Org SAM 3.1 model repository](https://huggingface.co/Comfy-Org/sam3.1). -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) -以下のディレクトリ構造に配置してください： +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -79,36 +118,36 @@ SAM 3.1 モデルは [Comfy-Org SAM 3.1 モデルリポジトリ](https://huggin └── sam3.1_multiplex_fp16.safetensors ``` -### 3. ワークフローの使い方 +### 3. Using the Workflows -**画像セグメンテーション：** +**Image Segmentation:** -- **画像** — `Load Image` ノードで画像を読み込み（ComfyUI の `input/` フォルダに配置） -- **オブジェクトプロンプト** — セグメント対象の短いテキスト説明（例：`person`、`car`、`cat`） -- 出力は画像に適用されたマスクで、RGBA プレビューでセグメンテーション結果を確認 +- **Image** — Load an image via the `Load Image` node (place it in the ComfyUI `input/` folder) +- **Object Prompt** — A short text description of the object(s) to segment, e.g. `person`, `car`, `cat` +- The output is a mask applied to the image, with an RGBA preview showing the segmentation result -**動画セグメンテーション：** +**Video Segmentation:** -- **動画** — `Load Video` ノードで動画を読み込み -- **オブジェクトプロンプト** — 画像と同じ、フレーム間でトラッキング・セグメントする対象の短いテキスト -- 出力には各フレームのマスクとバウンディングボックスが含まれます +- **Video** — Load a video via the `Load Video` node +- **Object Prompt** — Same as image, a short text prompt describing what to track and segment across frames +- The output provides masks and bounding boxes for each frame -**プロンプト形式：** +**Prompt format:** -| プロンプト | 役割 | +| Prompt | Role | |--------|------| -| SAM3 オブジェクトプロンプト | セグメントする**対象**の短い説明。最大 32 トークン | +| SAM3 object prompt | Short description of **what** to segment. Max 32 tokens. | -複数の対象を個別にプロンプトするには、カンマで区切り、`:N` で各プロンプトの最大検出数を指定： +To prompt multiple subjects separately, separate with commas and use `:N` to specify the max amount of objects detected per prompt: `eye:2, window panels:4` - - 本ワークフローは Subgraph ノードを使用してモジュール化された処理を実現しています。Subgraph ドキュメントを参照してワークフローのカスタマイズと拡張方法を学んでください。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 補足説明 +## Additional Notes -- **プロンプトは短く具体的に** — モデルにはプロンプトあたり 32 トークンの制限があります -- **マルチオブジェクト検出** — カンマで異なるオブジェクトタイプを区切り、`:N` でタイプごとの検出数を制限 -- **セグメンテーションマスク** — 出力マスクは他のワークフロー（修復、背景除去など）の入力として使用可能 -- **更新が必要** — SAM 3.1 を利用するには ComfyUI を最新バージョンにアップデートしてください +- **Keep prompts short and specific** — the model has a 32-token limit per prompt +- **Multi-object detection** — use commas to separate different object types, and `:N` to cap detections per type +- **Segmentation masks** — the output mask can be used as input to other workflows (e.g., inpainting, background removal) +- **Update required** — make sure ComfyUI is updated to the latest version to access SAM 3.1 support diff --git a/ja/tutorials/utility/video-upscale.mdx b/ja/tutorials/utility/video-upscale.mdx index ea1a96206..a174e18d4 100644 --- a/ja/tutorials/utility/video-upscale.mdx +++ b/ja/tutorials/utility/video-upscale.mdx @@ -15,7 +15,6 @@ translationBlockHashes: "Tips": a5dcf8f3 --- - 本ガイドでは、ComfyUI における動画アップスケーリングのワークフローについて解説します。ローカルモデルおよびパートナーノードの選択肢を、さまざまなユースケースに応じて紹介します。 diff --git a/ja/tutorials/utility/void-video-inpainting.mdx b/ja/tutorials/utility/void-video-inpainting.mdx index 57a3967e5..9fb683e28 100644 --- a/ja/tutorials/utility/void-video-inpainting.mdx +++ b/ja/tutorials/utility/void-video-inpainting.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": a3d04a46 --- - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -45,44 +44,61 @@ VOID は ComfyUI でネイティブサポートされており（PR [#13403](htt > **制限事項：** 不明瞭なマスク、乱雑な動き、またはフレームを支配する対象物は、依然として最適とは言えない結果を生む可能性があります — プロンプトでは根本的に誤ったセグメンテーションを修正できません。 -## VOID ビデオ修復ワークフロー +## VOID Video Inpainting Workflow + +### 1. Download Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "VOID: Video Inpainting" under the Utility category. + +### VOID: Video Inpainting (`utility_void_video_inpainting`) -### 1. ワークフローをダウンロード +Upload a video and mask the object you want to remove. Generate a clean video with the object and its physical interactions deleted. -ComfyUI を最新バージョンにアップデートし、メニューから `ワークフロー` -> `テンプレートを閲覧` に進み、Utility カテゴリから "VOID: Video Inpainting" を見つけてください。 +VOID video inpainting workflow preview - - Download workflow + + + Open in Comfy Cloud + + Download JSON or search "VOID: Video Inpainting" in Template Library + + + +**入力素材** + +Upload this file to the matching `LoadVideo` node: - - Open in cloud + + + `LoadVideo` node 4 · `snowboarder.mp4` + -### 2. モデルをダウンロード +### 2. Download Models -すべてのモデルは [Comfy-Org VOID モデルリポジトリ](https://huggingface.co/Comfy-Org/void-model) でホストされています。 +All models are hosted on the [Comfy-Org VOID model repository](https://huggingface.co/Comfy-Org/void-model). -**拡散モデル** — 中核となる2パス修復モデル： +**Diffusion Models** — the core two-pass inpainting model: -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 精錬パス、時間的安定性に優れる -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 一次パス +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — Refinement pass, better temporal stability +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — Primary pass -**VAE：** +**VAE:** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) -**オプティカルフロー：** +**Optical Flow:** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) -**SAM3 セグメンテーション：** +**SAM3 Checkpoint** — for segmentation: -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) -**テキストエンコーダ：** +**Text Encoder:** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ @@ -100,34 +116,34 @@ ComfyUI を最新バージョンにアップデートし、メニューから ` │ └── void_pass1.safetensors ``` -### 3. ワークフローの使い方 +### 3. Using the Workflow -**入力パラメータ：** +**Inputs:** -- **ソース動画** — `Load Video` ノードで動画を読み込みます（ComfyUI `input/` フォルダに配置） -- **ポジティブプロンプト（修復フィル）** — オブジェクト削除**後**のシーンを説明します。削除されたものではなく、残ったものとその見え方に焦点を当てます - - 例：`empty kitchen counter, daylight, tiles visible` -- **ネガティブプロンプト** — オプションのアーティファクト防止リスト。空でも可 -- **SAM3 オブジェクトプロンプト** — **削除したいもの**の短いラベル。SAM3 は意味理解によりターゲットオブジェクトのセグメンテーションマスクを作成します - - 例：`person in blue jacket`、`red cup on table` - - SAM3 プロンプトの最大トークン数は **32**。複数の対象を個別に指定する場合はカンマで区切り、`:N` でプロンプトごとの最大検出数を指定：`eye:2, window panels:4` +- **Source video** — Load a video via the `Load Video` node (place it in the ComfyUI `input/` folder) +- **Positive prompt (inpaint fill)** — Describe the scene **after** removal. Focus on what remains and how it looks, not on what was removed + - Example: `empty kitchen counter, daylight, tiles visible` +- **Negative prompt** — Optional anti-artifact list; can be left empty +- **SAM3 object prompt** — A short label for **what** to mask out. SAM3 uses semantic understanding to create a segmentation mask for the target object. + - Example: `person in blue jacket`, `red cup on table` + - Max tokens for SAM3 prompts is **32**. To prompt multiple subjects separately, separate with commas and use `:N` to specify the max objects detected per prompt: `eye:2, window panels:4` -**プロンプトの役割分担：** +**Modes:** -| プロンプト | 役割 | +| Prompt | Role | |--------|------| -| SAM3 オブジェクト | **何を**削除するか（SAM3 が意味セグメンテーションでマスクを作成） | -| ポジティブ（修復） | **どのように**穴を埋めるか | +| SAM3 object | **What** is removed (SAM3 creates the mask via semantic segmentation) | +| Positive (inpaint) | **How** the hole is filled across time | -長いクリップやテクスチャ背景では **Pass 2**（精錬パス）を使用することで時間的安定性が向上します。**Pass 1** のみの方が高速ですが、ジッターが発生しやすくなります。 +Use **Pass 2** (refinement pass) for longer clips or textured backgrounds where temporal stability matters. **Pass 1** alone is faster but may show more jitter. - - このワークフローはモジュラーな動画処理に Subgraph ノードを使用しています。Subgraph ドキュメントを参照して、ワークフローのカスタマイズと拡張方法を学んでください。 + + This workflow uses Subgraph nodes for modular video processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 補足説明 +## Additional Notes -- **マスクの品質が重要** — 削除対象をしっかり囲むクリーンなマスクが最良の結果を生みます -- **プロンプト作成のコツ** — 削除後、シーンが自然に見えるべき姿を説明し、削除そのものを説明しないでください -- **ネガティブプロンプト** は、繰り返し現れる欠陥（ウォーターマーク、ぼやけ、余分な手足など）が見られる場合にのみ使用してください -- **2パスワークフロー** — テンプレートは Pass 1 から Pass 2 を自動実行します。テスト中は Pass 1 のみを実行して迅速に反復することもできます +- **Mask quality matters** — a clean, tight mask around the target object produces the best results +- **Prompt writing tip** — describe the scene as it should appear _naturally_ after removal, not the removal itself +- **Use negative prompt** only when you see repeating defects (watermarks, blur, extra limbs) +- **Two-pass workflow** — the template runs Pass 1 then Pass 2 automatically; you can also run just Pass 1 for faster iterations during testing diff --git a/ja/tutorials/video/bytedance/bernini-r.mdx b/ja/tutorials/video/bytedance/bernini-r.mdx index 224cde466..8446ecd27 100644 --- a/ja/tutorials/video/bytedance/bernini-r.mdx +++ b/ja/tutorials/video/bytedance/bernini-r.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Community Resources": 9b4a0aaf --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # ComfyUI Bernini-R 概要 @@ -82,19 +80,27 @@ ComfyUI/ **機能説明：** 照明が一致した編集画像を生成し、前後の比較を並べて表示します。ポートレートやプロダクトの再照明、写真コレクションの一貫した照明、E コマースカタログ撮影に最適です。 +Bernini-R 画像編集ワークフロープレビュー + - - JSON をダウンロード または テンプレートライブラリで "Bernini-R" を検索 - Comfy Cloud で開く + + JSON をダウンロード または テンプレートライブラリで "Bernini-R" を検索 + -
- Bernini-R 画像編集出力 - Bernini-R 画像編集比較 -
+**入力素材** + + + + デフォルトの入力画像をダウンロードするか、独自の画像を使用してください。 + + + デフォルトの参照画像をダウンロードするか、独自の画像を使用してください。 + + ### 実行手順 @@ -114,16 +120,27 @@ ComfyUI/ **機能説明：** Bernini-R で一貫した再照明の編集動画を生成します。ソース動画、オプションの参照画像や参照動画を接続し、タスクタイプを選択し、プロンプトを作成して実行します。 +Bernini-R 動画編集ワークフロープレビュー + - - JSON をダウンロード または テンプレートライブラリで "Bernini-R" を検索 - Comfy Cloud で開く + + JSON をダウンロード または テンプレートライブラリで "Bernini-R" を検索 + -![Bernini-R 動画編集プレビュー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_bernini_r_video_editing-1.webp) +**入力素材** + + + + デフォルトの入力動画をダウンロードするか、独自の動画を使用してください。 + + + デフォルトの参照画像をダウンロードするか、独自の画像を使用してください。 + + ### 実行手順 diff --git a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index e9fee4dff..a36959688 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Super-resolution upscaler": 89338328 --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; [HunyuanVideo 1.5](https://github.com/Tencent/HunyuanVideo) は、テンセントの Hunyuan チームによって開発された、軽量な 8.3B パラメータモデルです。消費者向け GPU（24GB VRAM）でフラッグシップ級の動画生成を実現し、品質を損なうことなく参入障壁を劇的に下げます。 diff --git a/ja/tutorials/video/hunyuan/hunyuan-video.mdx b/ja/tutorials/video/hunyuan/hunyuan-video.mdx index d6952e34e..a7a1b9ec5 100644 --- a/ja/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ja/tutorials/video/hunyuan/hunyuan-video.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Try It Yourself": 076fb43f --- - - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; +Workflow preview + Comfy Cloud で開く diff --git a/ja/tutorials/video/wan/wan2-2-fun-control.mdx b/ja/tutorials/video/wan/wan2-2-fun-control.mdx index 30c9883a2..f863a8219 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-control.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Wan2.2 Fun Control Video Generation Workflow Example": cbbb7456 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Wan2.2-Fun-Control** は、Alibaba PAI チームによってリリースされた次世代の動画生成・制御モデルです。革新的な Control Codes 機制を導入し、深層学習とマルチモーダル条件入力を組み合わせることで、预设された制御条件に準拠した高品質な動画を生成できます。本モデルは **Apache 2.0 ライセンス** でリリースされており、商用利用も可能です。 @@ -40,7 +38,6 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - このワークフローは 2 つのバージョンを提供します： 1. lightx2v による [Wan2.2-Lightning](https://huggingface.co/lightx2v/Wan2.2-Lightning) 4 ステップ LoRA を使用したバージョン：動画のダイナミクスにいくつかの損失が生じる可能性がありますが、速度は速くなります 2. 加速 LoRA を使用しない fp8_scaled バージョン @@ -64,6 +61,8 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/wan2.2_14B_fun_inp.mp4" > +Workflow preview + Comfy Cloud で開く @@ -142,7 +141,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. ワークフローガイド ![Wan2.2 Fun Control ワークフロー手順](/images/tutorial/video/wan/wan2_2/wan_2.2_14b_fun_control.jpg) diff --git a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx index c8d400a7e..433c5c5de 100644 --- a/ja/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ja/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 692b42a8 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Wan2.2-Fun-Inp** は、Alibaba PAI チームが開発・公開した首尾フレーム制御型動画生成モデルです。ユーザーは**開始フレーム画像と終了フレーム画像**を入力することで、それらの間を滑らかに遷移する中間動画を生成できます。これにより、クリエイターはより高度な創造的コントロールを実現できます。本モデルは **Apache 2.0 ライセンス**のもとで公開されており、商用利用も可能です。 @@ -62,6 +60,8 @@ ComfyUI を最新版に更新した後、メニュー `Workflow` → `Browse Tem または、ComfyUI を最新版に更新した上で、以下のリンクからワークフローファイルをダウンロードし、ComfyUI の画面にドラッグ＆ドロップして読み込んでください。 +Wan2.2 Fun Inp ワークフロープレビュー + Download JSON or search "Wan2.2 Fun Inp" in Template Library diff --git a/ja/tutorials/video/wan/wan2-2-s2v.mdx b/ja/tutorials/video/wan/wan2-2-s2v.mdx index 3fa4d0882..9daac98cf 100644 --- a/ja/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ja/tutorials/video/wan/wan2-2-s2v.mdx @@ -20,7 +20,6 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' Wan2.2 S2V ソースコード: [GitHub](https://github.com/aigc-apps/VideoX-Fun) Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) - ## Wan2.2 S2V の ComfyUI ネイティブワークフロー @@ -35,6 +34,8 @@ Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > +Workflow preview + Comfy Cloud で開く @@ -77,9 +78,6 @@ Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14 以下の画像および音声ファイルを入力としてダウンロードしてください: ![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - - - ### 2. モデルのダウンロードリンク すべてのモデルは、[当社のリポジトリ](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) から入手できます。 @@ -99,7 +97,6 @@ Wan2.2 S2V モデル: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14 **text_encoders** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - ``` ComfyUI/ ├───📂 models/ @@ -114,7 +111,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. ワークフローの操作手順 ![ワークフローの操作手順](/images/tutorial/video/wan/wan_2.2_14b_s2v.jpg) diff --git a/ja/tutorials/video/wan/wan2_2.mdx b/ja/tutorials/video/wan/wan2_2.mdx index b75e586ba..f09295522 100644 --- a/ja/tutorials/video/wan/wan2_2.mdx +++ b/ja/tutorials/video/wan/wan2_2.mdx @@ -16,8 +16,6 @@ translationBlockHashes: "Community Resources": 7463b48b --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - このチュートリアルでは [🤗 Comfy-Org/Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) バージョンを使用します。 @@ -103,6 +100,8 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > +Wan 2.2 5B 動画生成ワークフロープレビュー + Download JSON or search "Wan2.2 5B" in Template Library @@ -177,6 +176,8 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_t2v.mp4" > +Wan 2.2 14B テキストから動画ワークフロープレビュー + Download JSON or search "Wan2.2 14B T2V" in Template Library @@ -219,7 +220,6 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows **Text Encoder** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - ``` ComfyUI/ ├───📂 models/ @@ -256,6 +256,8 @@ ComfyUI を最新バージョンに更新し、メニュー `Workflow` -> `Brows src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_14B_i2v.mp4" > +Wan 2.2 14B 画像から動画ワークフロープレビュー + Download JSON or search "Wan2.2 14B I2V" in Template Library @@ -337,6 +339,8 @@ ComfyUI/ src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v.mp4" > +Wan 2.2 14B 最初と最後のフレームから動画ワークフロープレビュー + Download JSON or search "Wan2.2 14B FLF2V" in Template Library diff --git a/ja/tutorials/video/zai/scail2.mdx b/ja/tutorials/video/zai/scail2.mdx index 515be3201..7535f4c7c 100644 --- a/ja/tutorials/video/zai/scail2.mdx +++ b/ja/tutorials/video/zai/scail2.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Model Installation": d4fa2215 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **SCAIL-2** は、Wan2.1 上に構築されたエンドツーエンドのキャラクターアニメーションモデルです。駆動ビデオの動きを参照キャラクター画像に転送することで、キャラクターアニメーション（キャラクターに動きを実行させる）とビデオ内キャラクター置換（追跡された人物を参照キャラクターに置き換える）を可能にします。 @@ -31,6 +29,8 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## SCAIL-2 キャラクター置換ワークフロー +Workflow preview + Comfy Cloud で開く diff --git a/ko/tutorials/3d/hunyuan3D-2.mdx b/ko/tutorials/3d/hunyuan3D-2.mdx index d67fada7b..abc3d27ff 100644 --- a/ko/tutorials/3d/hunyuan3D-2.mdx +++ b/ko/tutorials/3d/hunyuan3D-2.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Hunyuan3D 2.0 Open-Source Model Series": 363ff037 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' # Hunyuan3D 2.0 소개 @@ -28,7 +26,6 @@ Hunyuan3D 2.0은 두 단계의 생성 방식을 채택하며, 먼저 텍스처 1. **기하학적 생성 모델(Hunyuan3D-DiT)**: 플로우 확산 트랜스포머 아키텍처를 기반으로 하며, 입력 조건과 정확히 일치하는 텍스처 없는 기하학적 모델을 생성합니다. 2. **텍스처 생성 모델(Hunyuan3D-Paint)**: 기하학적 조건과 다중뷰 확산 기법을 결합해 모델에 고해상도 텍스처를 추가하며, PBR 재료를 지원합니다. - **주요 장점** - **고정밀 생성**: 선명한 기하학적 구조, 풍부한 텍스처 색상, PBR 재료 생성 지원으로 사실에 가까운 라이팅 효과를 구현합니다. @@ -52,167 +49,242 @@ ComfyUI는 현재 Hunyuan3D-2mv를 기본적으로 지원하지만, 텍스처 이렇게 하면 해당 워크플로우가 로드되고 필요한 모델을 다운로드하라는 메시지가 표시됩니다. 생성된 `.glb` 형식의 모델은 `ComfyUI/output/mesh` 폴더에 출력됩니다. -## ComfyUI Hunyuan3D-2mv 워크플로우 예시 +## ComfyUI Hunyuan3D-2mv Workflow Example -Hunyuan3D-2mv 워크플로우에서는 다중뷰 이미지를 사용해 3D 모델을 생성합니다. 이 워크플로우에서는 다중뷰 이미지가 필수는 아니며, `front` 뷰 이미지만 사용해도 3D 모델을 생성할 수 있습니다. +In the Hunyuan3D-2mv workflow, we'll use multi-view images to generate a 3D model. Note that multiple view images are not mandatory in this workflow - you can use only the `front` view image to generate a 3D model. +### HY 3D 2.0 MV (`3d_hunyuan3d_multiview_to_model`) + +Generate 3D models from multiple views using Hunyuan3D 2.0 MV. + +HY 3D 2.0 MV workflow preview + - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 - 워크플로우 JSON 파일 다운로드 + + Run this workflow instantly on Comfy Cloud + + + Download JSON or search "HY 3D 2.0 MV" in Template Library + -### 1. 워크플로우 +**입력 자료** -아래 이미지를 다운로드해 ComfyUI로 드래그하여 워크플로우를 로드해 주세요. -![Hunyuan3D-2mv 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/hunyuan-3d-multiview-elf.webp) +Upload these files to the matching `LoadImage` nodes: -아래 이미지를 다운로드해 입력 이미지로 사용해 주세요. + + + `LoadImage` node 56 · front view + + + `LoadImage` node 78 · left view + + + `LoadImage` node 80 · back view + + + `LoadImage` node 87 · right view + + + +### 1. Workflow -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/front.png) -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/left.png) -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/back.png) +Please download the images below and drag into ComfyUI to load the workflow. +![Hunyuan3D-2mv workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/hunyuan-3d-multiview-elf.webp) +Download the images below we will use them as input images. + +![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/front.png) +![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/left.png) +![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/back.png) -이 예시에서 사용된 입력 이미지는 이미 사전 처리되어 불필요한 배경이 제거되었습니다. 실제 사용 시, [ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials)와 같은 커스텀 노드를 사용해 자동으로 불필요한 배경을 제거할 수 있습니다. +In this example, the input images have already been preprocessed to remove excess background. In actual use, you can use custom nodes like [ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) to automatically remove excess background. -### 2. 수동 모델 설치 +### 2. Manual Model Installation -아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv.safetensors // 이름 변경된 파일 +│ │ └── hunyuan3d-dit-v2-mv.safetensors // renamed file ``` -### 3. 워크플로우 실행 단계 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2mv](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv.jpg) -1. 이미지 전용 체크포인트 로더(img2vid 모델)가 다운로드하고 이름을 변경한 `hunyuan3d-dit-v2-mv.safetensors` 모델을 로드했는지 확인하세요. -2. 각 `Load Image` 노드에 해당 뷰 이미지를 로드하세요. -3. `Queue` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. +1. Ensure that the Image Only Checkpoint Loader(img2vid model) has loaded our downloaded and renamed `hunyuan3d-dit-v2-mv.safetensors` model +2. Load the corresponding view images in each of the `Load Image` nodes +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow + +If you need to add more views, make sure to load other view images in the `Hunyuan3Dv2ConditioningMultiView` node, and ensure that you load the corresponding view images in the `Load Image` nodes. + +## Hunyuan3D-2mv-turbo Workflow -뷰를 더 추가해야 하는 경우, `Hunyuan3Dv2ConditioningMultiView` 노드에 다른 뷰 이미지를 로드하고, `Load Image` 노드에도 해당 뷰 이미지를 로드했는지 확인하세요. +In the Hunyuan3D-2mv-turbo workflow, we'll use the Hunyuan3D-2mv-turbo model to generate 3D models. This model is a step distillation version of Hunyuan3D-2mv, allowing for faster 3D model generation. In this version of the workflow, we set `cfg` to 1.0 and add a `flux guidance` node to control the `distilled cfg` generation. -## Hunyuan3D-2mv-turbo 워크플로우 +### HY 3D 2.0 MV Turbo (`3d_hunyuan3d_multiview_to_model_turbo`) -Hunyuan3D-2mv-turbo 워크플로우에서는 Hunyuan3D-2mv-turbo 모델을 사용해 3D 모델을 생성합니다. 이 모델은 Hunyuan3D-2mv의 단계 증류 버전으로, 더 빠른 3D 모델 생성을 가능하게 합니다. 이번 버전의 워크플로우에서는 `cfg`를 1.0으로 설정하고, `flux guidance` 노드를 추가해 `증류된 cfg` 생성을 제어합니다. +Generate 3D models from multiple views using Hunyuan3D 2.0 MV Turbo. + +HY 3D 2.0 MV Turbo workflow preview - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 - 워크플로우 JSON 파일 다운로드 + + Run this workflow instantly on Comfy Cloud + + + Download JSON or search "HY 3D 2.0 MV Turbo" in Template Library + -### 1. 워크플로우 +**입력 자료** + +Upload these files to the matching `LoadImage` nodes: -아래 이미지를 다운로드해 ComfyUI로 드래그하여 워크플로우를 로드해 주세요. + + + `LoadImage` node 56 · front view + + + `LoadImage` node 82 · back view + + + `LoadImage` node 85 · left view + + + `LoadImage` node 87 · right view + + -![Hunyuan3D-2mv-turbo 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) +### 1. Workflow -아래 이미지를 다운로드해 입력 이미지로 사용해 주세요. +Please download the images below and drag into ComfyUI to load the workflow. +![Hunyuan3D-2mv-turbo workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) +Download the images below we will use them as input images. +![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) +![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) -### 2. 수동 모델 설치 +### 2. Manual Model Installation -아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2-mv-turbo.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv-turbo.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // 이름 변경된 파일 +│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // renamed file ``` -### 3. 워크플로우 실행 단계 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2mv_turbo](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv_turbo.jpg) -1. `Image Only Checkpoint Loader(img2vid 모델)` 노드가 이름을 변경한 `hunyuan3d-dit-v2-mv-turbo.safetensors` 모델을 로드했는지 확인하세요. -2. 각 `Load Image` 노드에 해당 뷰 이미지를 로드하세요. -3. `Queue` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. +1. Ensure that the `Image Only Checkpoint Loader(img2vid model)` node has loaded our renamed `hunyuan3d-dit-v2-mv-turbo.safetensors` model +2. Load the corresponding view images in each of the `Load Image` nodes +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow + +## Hunyuan3D-2 Single View Workflow + +In the Hunyuan3D-2 workflow, we'll use the Hunyuan3D-2 model to generate 3D models. This model is not a multi-view model. In this workflow, we use the `Hunyuan3Dv2Conditioning` node instead of the `Hunyuan3Dv2ConditioningMultiView` node. + +### HY 3D 2.0 (`3d_hunyuan3d_image_to_model`) + +Generate 3D models from single images using Hunyuan3D 2.0. + +HY 3D 2.0 workflow preview + + + + Run this workflow instantly on Comfy Cloud + + + Download JSON or search "HY 3D 2.0" in Template Library + + -## Hunyuan3D-2 단일뷰 워크플로우 +**입력 자료** -Hunyuan3D-2 워크플로우에서는 Hunyuan3D-2 모델을 사용해 3D 모델을 생성합니다. 이 모델은 다중뷰 모델이 아닙니다. 이번 워크플로우에서는 `Hunyuan3Dv2ConditioningMultiView` 노드 대신 `Hunyuan3Dv2Conditioning` 노드를 사용합니다. +Upload this file to the matching `LoadImage` node: - 이 워크플로우를 Comfy Cloud에서 즉시 실행하세요 - 워크플로우 JSON 파일 다운로드 + + `LoadImage` node 56 · `3d_hunyuan3d_image_to_model_input_image.png` + -### 1. 워크플로우 +### 1. Workflow -아래 이미지를 다운로드해 ComfyUI로 드래그하여 워크플로우를 로드해 주세요. +Please download the image below and drag it into ComfyUI to load the workflow. -![Hunyuan3D-2 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) +![Hunyuan3D-2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) -아래 이미지를 다운로드해 입력 이미지로 사용해 주세요. -![ComfyUI Hunyuan 3D 2 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan_3d_v2_non_multiview_train.png) +Download the image below we will use it as input image. +![ComfyUI Hunyuan 3D 2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan_3d_v2_non_multiview_train.png) -### 2. 수동 모델 설치 +### 2. Manual Model Installation -아래 모델을 다운로드해 해당 ComfyUI 폴더에 저장해 주세요. +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - 다운로드 후 `hunyuan3d-dit-v2.safetensors`로 이름을 변경할 수 있습니다. +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2.safetensors // 이름 변경된 파일 +│ │ └── hunyuan3d-dit-v2.safetensors // renamed file ``` -### 3. 워크플로우 실행 단계 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2_non_multiview.jpg) -1. `Image Only Checkpoint Loader(img2vid 모델)` 노드가 이름을 변경한 `hunyuan3d-dit-v2.safetensors` 모델을 로드했는지 확인하세요. -2. `Load Image` 노드에 이미지를 로드하세요. -3. `Queue` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. +1. Ensure that the `Image Only Checkpoint Loader(img2vid model)` node has loaded our renamed `hunyuan3d-dit-v2.safetensors` model +2. Load the image in the `Load Image` node +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -## 커뮤니티 리소스 +## Community Resources -아래는 Hunyuan3D-2 관련 ComfyUI 커뮤니티 리소스입니다. +Below are ComfyUI community resources related to Hunyuan3D-2 - [ComfyUI-Hunyuan3DWrapper](https://github.com/kijai/ComfyUI-Hunyuan3DWrapper) - [Kijai/Hunyuan3D-2_safetensors](https://huggingface.co/Kijai/Hunyuan3D-2_safetensors/tree/main) - [ComfyUI-3D-Pack](https://github.com/MrForExample/ComfyUI-3D-Pack) -## Hunyuan3D 2.0 오픈소스 모델 시리즈 +## Hunyuan3D 2.0 Open-Source Model Series -현재 Hunyuan3D 2.0은 완전한 3D 생성 과정을 포괄하는 여러 모델을 오픈소스로 공개했습니다. 자세한 정보는 [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2)를 방문해 확인하실 수 있습니다. +Currently, Hunyuan3D 2.0 has open-sourced multiple models covering the complete 3D generation process. You can visit [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2) for more information. -**Hunyuan3D-2mini 시리즈** +**Hunyuan3D-2mini Series** -| 모델 | 설명 | 날짜 | 파라미터 | Huggingface | -|-----------------------|--------------------|------------|----------|--------------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-mini | 미니 이미지 → 형상 모델 | 2025-03-18 | 0.6B | [방문](https://huggingface.co/tencent/Hunyuan3D-2mini/tree/main/hunyuan3d-dit-v2-mini) | +| Model | Description | Date | Parameters | Huggingface | +|-----------------------|---------------------------|------------|------------|--------------------------------------------------------------------------------------------| +| Hunyuan3D-DiT-v2-mini | Mini Image to Shape Model | 2025-03-18 | 0.6B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mini/tree/main/hunyuan3d-dit-v2-mini) | -**Hunyuan3D-2mv 시리즈** +**Hunyuan3D-2mv Series** -| 모델 | 설명 | 날짜 | 파라미터 | Huggingface | -|--------------------------|-----------------------------------------------------------------------------------------------------|------------|----------|------------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-mv-Fast | 지침 증류 버전, DIT 추론 시간 절반으로 줄일 수 있음 | 2025-03-18 | 1.1B | [방문](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv-fast) | -| Hunyuan3D-DiT-v2-mv | 다중뷰 이미지 → 형상 모델, 장면을 이해하는 데 여러 각도가 필요한 3D 생성에 적합 | 2025-03-18 | 1.1B | [방문](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv) | +| Model | Description | Date | Parameters | Huggingface | +|--------------------------|-------------------------------------------------------------------------------------------------------------|------------|------------|------------------------------------------------------------------------------------------| +| Hunyuan3D-DiT-v2-mv-Fast | Guidance Distillation Version, can halve DIT inference time | 2025-03-18 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv-fast) | +| Hunyuan3D-DiT-v2-mv | Multi-view Image to Shape Model, suitable for 3D creation requiring multiple angles to understand the scene | 2025-03-18 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv) | -**Hunyuan3D-2 시리즈** +**Hunyuan3D-2 Series** -| 모델 | 설명 | 날짜 | 파라미터 | Huggingface | -|-------------------------|----------------------|------------|----------|---------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-0-Fast | 지침 증류 모델 | 2025-02-03 | 1.1B | [방문](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast) | -| Hunyuan3D-DiT-v2-0 | 이미지 → 형상 모델 | 2025-01-21 | 1.1B | [방문](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0) | -| Hunyuan3D-Paint-v2-0 | 텍스처 생성 모델 | 2025-01-21 | 1.3B | [방문](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-paint-v2-0) | -| Hunyuan3D-Delight-v2-0 | 이미지 딜라이트 모델 | 2025-01-21 | 1.3B | [방문](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-delight-v2-0) | +| Model | Description | Date | Parameters | Huggingface | +|-------------------------|-----------------------------|------------|------------|---------------------------------------------------------------------------------------| +| Hunyuan3D-DiT-v2-0-Fast | Guidance Distillation Model | 2025-02-03 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast) | +| Hunyuan3D-DiT-v2-0 | Image to Shape Model | 2025-01-21 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0) | +| Hunyuan3D-Paint-v2-0 | Texture Generation Model | 2025-01-21 | 1.3B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-paint-v2-0) | +| Hunyuan3D-Delight-v2-0 | Image Delight Model | 2025-01-21 | 1.3B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-delight-v2-0) | diff --git a/ko/tutorials/3d/triposplat.mdx b/ko/tutorials/3d/triposplat.mdx index 26224ebc6..bc4d6a0b2 100644 --- a/ko/tutorials/3d/triposplat.mdx +++ b/ko/tutorials/3d/triposplat.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Model downloads": 44894623 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" **TripoSplat**는 단일 2D 이미지에서 직접 **3D 가우시안 스플래트** 표현을 생성하는 오픈소스 모델입니다. VAST-AI가 개발했으며 오픈소스 라이선스로 공개되었습니다. @@ -51,83 +50,132 @@ TripoSplat은 단일 RGB 이미지를 입력으로 받아 3D 가우시안 프리 - 입력 이미지(PNG/JPG)를 로드합니다. - 샘플 이미지: `white-hotel-on-rocky-island.png` (템플릿 라이브러리에서 사용 가능) -### TripoSplat (서브그래프) +### TripoSplat: Image to Gaussian Splat (`3d_triposplat_image_to_gaussian_splat`) + +Upload a single 2D image. Generate a high-quality 3D Gaussian splat representation with controllable density and budget for rendering. + +TripoSplat workflow preview + + + + + + Run this workflow instantly on Comfy Cloud + + + Download JSON or search "TripoSplat: Image to Gaussian Splat" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 99 · `white-hotel-on-rocky-island.png` + + + +
+ Input image +
+ +## How it works + +TripoSplat uses a **feed-forward architecture** that takes a single RGB image and directly predicts a set of 3D Gaussian primitives. The pipeline involves: + +1. **Image encoding**: the input image is processed by a vision encoder (DINOv2) +2. **Triplane generation**: features are decoded into a triplane representation +3. **Gaussian prediction**: the triplane is sampled to produce Gaussian parameters (position, scale, rotation, opacity, color) +4. **Rendering**: Gaussians are rendered from arbitrary viewpoints using differentiable splatting + + + This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. + +## Workflow node guide + +### LoadImage +- Loads your input image (PNG/JPG) +- Sample image: `white-hotel-on-rocky-island.png` (available in Template Library) + +### TripoSplat (subgraph) -주요 서브그래프 노드는 이미지를 처리해 3D 가우시안 스플래트를 생성합니다. 노출된 매개변수: +The main subgraph node processes the image and generates the 3D Gaussian splat. Exposed parameters: -| 매개변수 | 기본값 | 설명 | +| Parameter | Default | Description | |---|---|---| -| `switch` | — | 서브그래프 활성화/비활성화 | -| `num_gaussians` | — | 생성할 가우시안 프리미티브의 수 (품질/성능 제어) | -| `seed` | — | 재현성을 위한 난수 시드 | -| `unet_name` | — | TripoSplat 확산 모델 체크포인트 | -| `clip_name` | — | CLIP 비전 인코더 모델 | -| `vae_name` | — | VAE 인코딩/디코딩용 (메인 VAE와 인코더용 두 개 항목) | -| `bg_removal_name` | — | 배경 제거 모델 | +| `switch` | — | Enable/disable the subgraph | +| `num_gaussians` | — | Number of Gaussian primitives to generate (controls quality/performance) | +| `seed` | — | Random seed for reproducibility | +| `unet_name` | — | TripoSplat diffusion model checkpoint | +| `clip_name` | — | CLIP vision encoder model | +| `vae_name` | — | VAE for encoding/decoding (2 entries: one for the main VAE, one for the encoder) | +| `bg_removal_name` | — | Background removal model | ### CreateCameraInfo -- 결과를 렌더링하기 위한 카메라 궤도를 정의합니다. -- 매개변수: 궤도 유형, 각도, 거리, 시야각 등. -- 기본값: 35° 고도, 30 거리, 2.5 줌 +- Defines the camera orbit for rendering the result +- Parameters: orbit type, angle, distance, field of view, etc. +- Default: orbit at 35° elevation, 30 distance, 2.5 zoom ### RenderSplat -- 정의된 카메라 각도에서 가우시안 스플래트를 2D 이미지로 렌더링합니다. -- 매개변수: 출력 해상도(기본 1024×1024), 이미지 품질 설정 +- Renders the Gaussian splat into a 2D image from the defined camera angle +- Parameters: output resolution (default 1024×1024), image quality settings ### SplatToMesh -- 가우시안 스플래트를 메쉬로 변환합니다(선택사항). -- 매개변수: 메쉬 밀도, 평탄화, 간소화 +- Converts the Gaussian splat to a mesh (optional) +- Parameters: mesh density, smoothing, simplification ### SaveGLB -- 결과를 GLB 3D 파일로 저장합니다. +- Saves the result as a GLB 3D file ### SaveVideo -- 렌더링된 3D 장면의 동영상을 저장합니다. +- Saves a video of the rendered 3D scene ### SplatToFile3D -- 가우시안 스플래트를 SPZ 형식으로 보냅니다. +- Exports the Gaussian splat in SPZ format ### CreateVideo -- 렌더링된 프레임들로 동영상을 생성합니다. +- Creates a video from rendered frames -## 실행 단계 +## Steps to run -1. **이미지 로드**: **LoadImage** 노드를 사용해 단일 2D 이미지를 로드하세요. -2. **TripoSplat 서브그래프 실행**: 모델이 가우시안 스플래트 표현을 생성합니다. -3. **출력 형식 선택**: GLB, SPZ, 동영상으로 보내거나 메쉬로 렌더링하세요. -4. **결과 보기**: 생성된 3D 파일이나 렌더링된 미리보기를 활용하세요. +1. **Load an image**: use the **LoadImage** node to load a single 2D image +2. **Run the TripoSplat subgraph**: the model will generate a Gaussian splat representation +3. **Choose output format**: export as GLB, SPZ, video, or render to mesh +4. **View results**: use the created 3D file or rendered preview -## 출력 옵션 +## Output options -| 노드 | 형식 | 활용 사례 | +| Node | Format | Use case | |---|---|---| -| **SaveGLB** | `.glb` | 표준 3D 파일 형식, 3D 소프트웨어로 불러올 수 있음 | -| **SplatToFile3D** | `.spz` | 압축된 가우시안 스플래트 형식, 효율적 저장용 | -| **RenderSplat** | 2D 이미지 | 임의 각도에서의 결과 미리보기 | -| **SplatToMesh** | 메쉬 | 전통적인 메쉬로 변환해 추가 편집 가능 | +| **SaveGLB** | `.glb` | Standard 3D file format, importable into 3D software | +| **SplatToFile3D** | `.spz` | Compressed Gaussian splat format for efficient storage | +| **RenderSplat** | 2D image | Quick preview of the result from any angle | +| **SplatToMesh** | Mesh | Convert to traditional mesh for further editing | -## 모델 다운로드 +## Model downloads -TripoSplat 모델과 필요한 파일을 다운로드하세요. 해당 `models/` 하위 디렉토리에 배치하세요. +Download the TripoSplat model and required files. Place them in the corresponding `models/` subdirectories. - - triposplat_fp16.safetensors: TripoSplat 확산 모델 체크포인트 + + triposplat_fp16.safetensors: TripoSplat diffusion model checkpoint - - triposplat_vae_decoder_fp16.safetensors: VAE 디코더 + + triposplat_vae_decoder_fp16.safetensors: VAE decoder - flux2-vae.safetensors: Flux.2 VAE, 잠재적 인코딩용 + flux2-vae.safetensors: Flux.2 VAE for latent encoding - dino_v3_vit_h.safetensors: CLIP 비전 인코더 (DINOv2) + dino_v3_vit_h.safetensors: CLIP vision encoder (DINOv2) - - birefnet.safetensors: 전처리용 배경 제거 모델 + + birefnet.safetensors: Background removal model for preprocessing -### 모델 저장 위치 +### Model storage location ``` 📂 ComfyUI/ diff --git a/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx index 4dfcd158a..2f0a96af3 100644 --- a/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ko/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "ACE-Step 1.5 ComfyUI Related Resources": 518b9b68 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## ComfyUI의 ACE-Step 1.5 소개 @@ -29,27 +28,30 @@ ACE-Step 1.5는 오픈소스 음악 생성 모델의 주요 업데이트로, 이 -## 옵션 1: 올인원 체크포인트 (권장) +## Option 1: All-in-One Checkpoint (Recommended) -AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공하므로 다운로드와 관리가 더욱 용이합니다. +The AIO version packages all models into a single checkpoint file, making it easier to download and manage. -### AIO 워크플로우 +### ACE-Step 1.5 Music Generation AIO (`audio_ace_step_1_5_checkpoint`) - - AIO 워크플로우를 Comfy Cloud에서 바로 실행하세요. - +Input style tags and lyrics to generate a full song. The workflow uses the ACE-Step 1.5 model to produce commercial-grade music in under 10 seconds on consumer hardware. - - 올인원 체크포인트 워크플로우를 로컬에서 사용하기 위해 다운로드하세요. + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation AIO" in Template Library + -### AIO 모델 다운로드 +### AIO Model Download - - 올인원 체크포인트 파일 (대부분의 사용자에게 권장). + + All-in-one checkpoint file (recommended for most users). -**AIO 모델 저장 위치** +**AIO Model Storage Location** ``` 📂 ComfyUI/ @@ -58,39 +60,44 @@ AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공 │ └── ace_step_1.5_turbo_aio.safetensors ``` -## 옵션 2: 분할 모델 파일 +## Option 2: Split Model Files -분할 버전은 개별 모델 구성 요소를 따로 다운로드할 수 있도록 합니다. +The split version allows you to download individual model components separately. -### 분할 워크플로우 +### ACE-Step 1.5 Music Generation Workflow (`audio_ace_step_1_5_split`) - - 분할 모델 워크플로우를 Comfy Cloud에서 바로 실행하세요. - +Input a text prompt describing the music style and optional lyrics. Generate a full, high-quality audio song in under 10 seconds on consumer hardware. - - 분할 모델 워크플로우를 로컬에서 사용하기 위해 다운로드하세요. + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation Workflow" in Template Library + -### 분할 모델 다운로드 +### Split Model Downloads - - 디퓨전 모델. + + + Diffusion model. - - 텍스트 인코더 (0.6B). + + Text encoder (0.6B). - - 텍스트 인코더 (1.7B). + + Text encoder (1.7B). - - VAE 모델. + + VAE model. + -**분할 모델 저장 위치** +**Split Models Storage Location** ``` 📂 ComfyUI/ @@ -104,26 +111,26 @@ AIO 버전은 모든 모델을 하나의 체크포인트 파일에 묶어 제공 │ └── ace_1.5_vae.safetensors ``` -## ComfyUI에서 ACE-Step 1.5 주요 기능 +## ACE-Step 1.5 Key Features in ComfyUI -### 사고의 연쇄 계획 +### Chain-of-Thought Planning -ACE-Step 1.5 모델은 사고의 연쇄 추론을 통해 메타데이터, 가사, 캡션을 합성하여 디퓨전 과정을 안내하며, 더 일관성 있는 장편 곡을 만들어냅니다. +The ACE-Step 1.5 model synthesizes metadata, lyrics, and captions via Chain-of-Thought reasoning to guide the diffusion process, resulting in more coherent long-form compositions. -### 하이브리드 LM + DiT 아키텍처 +### Hybrid LM + DiT Architecture -ACE-Step 1.5는 곡 구조를 계획하는 언어 모델과 오디오 합성을 담당하는 디퓨전 트랜스포머(DiT)를 결합하며, 모두 ComfyUI에서 기본적으로 실행됩니다. +ACE-Step 1.5 combines a Language Model that plans the song structure with a Diffusion Transformer (DiT) that handles audio synthesis, all running natively in ComfyUI. -## 곧 ComfyUI에 추가될 기능 +## Coming Soon to ComfyUI -다음 기능들은 ACE-Step 1.5에 포함되어 있지만 아직 ComfyUI에서는 지원되지 않습니다: +These features are available in ACE-Step 1.5 but not yet supported in ComfyUI: -- **커버**: 모델에 어떤 곡이든 입력으로 주고 새로운 프롬프트와 가사를 제공하면, 모델이 전혀 다른 스타일로 곡을 재구성합니다. -- **재페인팅**: 특정 구간을 선택해 해당 부분만 다시 생성하고, 모델은 나머지 부분은 그대로 유지하면서 이를 다시 연결합니다. +- **Cover**: Give the model any song as input along with a new prompt and lyrics, and it will reimagine the track in a completely different style +- **Repaint**: Select a segment, regenerate just that section, and the model stitches it back in while keeping everything else untouched -## ACE-Step 1.5 ComfyUI 관련 자료 +## ACE-Step 1.5 ComfyUI Related Resources -- [프로젝트 페이지](https://ace-step.github.io/) +- [Project Page](https://ace-step.github.io/) - [Hugging Face](https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files) - [GitHub](https://github.com/ace-step/ACE-Step) -- [블로그 게시글](https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui) +- [Blog Post](https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui) diff --git a/ko/tutorials/audio/ace-step/ace-step-v1.mdx b/ko/tutorials/audio/ace-step/ace-step-v1.mdx index 9e80a4df1..0f57d8571 100644 --- a/ko/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ko/tutorials/audio/ace-step/ace-step-v1.mdx @@ -68,7 +68,6 @@ ACE-Step은 중국 팀 StepFun과 ACE Studio가 공동 개발한 오픈소스

입력용 예시 오디오 파일 다운로드

- ### 2. 워크플로우 단계별 완료하기 ![ACE-Step 단계 안내](/images/tutorial/audio/ace_step/ace_step_1_m2m_step_guide.jpg) diff --git a/ko/tutorials/audio/stable-audio/stable-audio-1.mdx b/ko/tutorials/audio/stable-audio/stable-audio-1.mdx index a732bad3d..e6531430e 100644 --- a/ko/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ko/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -15,37 +15,42 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" **관련 링크**: - [GitHub: Stability-AI/stable-audio-open-1.0](https://github.com/Stability-AI/stable-audio-open-1.0) -## 워크플로우 +## Workflow - - JSON 파일 다운로드 또는 템플릿 라이브러리에서 "Stable Audio 1.0" 검색 - +### Stable Audio 1.0: Text to Audio (`audio_stable_audio_example`) - - Comfy Cloud에서 열기 - +Generate audio from text prompts using Stable Audio. + +Stable Audio 1.0 text to audio workflow preview -![Stable Audio 1.0 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_example-1.webp) + + + Open in Comfy Cloud + + + Download JSON or search "Stable Audio 1.0: Text to Audio" in Template Library + + -이 워크플로우는 **표준 ComfyUI 노드**만 사용하며, 커스텀 노드는 필요하지 않습니다. Stable Audio 1.0 체크포인트를 로드하고, CLIP 텍스트 인코더(t5-base)를 통해 프롬프트를 인코딩한 후 KSampler로 잠재적 오디오를 노이즈 제거하고, 모델의 VAE를 통해 오디오로 디코드합니다. +The workflow uses **standard ComfyUI nodes** — no custom nodes required. It loads the Stable Audio 1.0 checkpoint, encodes your prompt via a CLIP text encoder (t5-base), denoises the latent audio with a KSampler, and decodes it to audio through the model's VAE. -**사용 방법**: -1. **체크포인트 로드하기** — `CheckpointLoaderSimple` 노드는 `stable-audio-open-1.0.safetensors`를 사용합니다. -2. **프롬프트 작성하기** — `CLIPTextEncode` 노드에 설명을 입력하세요(예: "heaven church electronic dance music"). -3. **지속시간 설정하기** — `EmptyLatentAudio` 노드의 길이 값을 조정하세요(기본값 47.6초). -4. **실행하기** (`Ctrl/Cmd + Enter`)를 클릭해 생성하세요. 오디오는 `ComfyUI/output/audio/`에 저장됩니다. +**How to use**: +1. **Load the checkpoint** — The `CheckpointLoaderSimple` node uses `stable-audio-open-1.0.safetensors` +2. **Write a prompt** — Enter your description in the `CLIPTextEncode` node (e.g. "heaven church electronic dance music") +3. **Set duration** — Adjust the `EmptyLatentAudio` node's length value (default 47.6 seconds) +4. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -## 모델 다운로드 +## Model download -워크플로우를 로드할 때 ComfyUI는 누락된 모델에 대한 다운로드 링크를 표시합니다. 수동으로 설정하려면 아래 파일을 다운로드해 올바른 폴더에 배치하세요. +When loading the workflow, ComfyUI will prompt you with download links for any missing models. To set up manually, download the files below and place them in the correct folders. -### 체크포인트 +### Checkpoint - - 2.3GB. models/checkpoints/ 폴더에 배치하세요. + + 2.3GB. Place in models/checkpoints/ -체크포인트를 다음 경로에 배치하세요: +Place the checkpoint in: ``` 📂 ComfyUI/ @@ -54,13 +59,13 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" │ └── stable-audio-open-1.0.safetensors ``` -### 텍스트 인코더 +### Text encoder - - 프롬프트 조건부 설정을 위한 텍스트 인코더. models/text_encoders/ 폴더에 배치하세요. + + Text encoder for prompt conditioning. Place in models/text_encoders/ -텍스트 인코더를 다음 경로에 배치하세요: +Place the text encoder in: ``` 📂 ComfyUI/ @@ -69,4 +74,4 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" │ └── t5-base.safetensors ``` -파일을 배치한 후 ComfyUI에서 **R**을 눌러 노드를 새로고침하고 최신 모델을 로드하세요. \ No newline at end of file +After placing the files, press **R** in ComfyUI to refresh nodes and load the latest models. diff --git a/ko/tutorials/audio/stable-audio/stable-audio-3.mdx b/ko/tutorials/audio/stable-audio/stable-audio-3.mdx index bc5d2baa4..81524f641 100644 --- a/ko/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ko/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Model download": d729490a --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -27,65 +26,73 @@ Stable Audio 3는 세 가지 변형으로 제공됩니다: - [Hugging Face (Comfy-Org/stable-audio-3)](https://huggingface.co/Comfy-Org/stable-audio-3) - [블로그: 발표](https://blog.comfy.org/p/stable-audio-3-day-0-support) -## 이용 가능한 워크플로우 +## Available workflows -### Stable Audio 3 Medium +### Stable Audio 3.0 Medium (`audio_stable_audio_3_medium`) - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Stable Audio 3 Medium" 검색 - +Input a short text idea, optional duration, seed, and category. Generate stereo audio (music, SFX, or instruments) using Stable Audio 3 with optional AI-driven text expansion. - - Comfy Cloud에서 열기 +Stable Audio 3 Medium workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Stable Audio 3.0 Medium" in Template Library + -![Stable Audio 3 Medium 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium-1.webp) +The **Stable Audio 3 Medium** workflow is a full-featured text-to-audio generation pipeline. You provide a short text idea, optional duration, seed, and category — the workflow expands your prompt using Qwen with a **category-aware reprompt template**, then generates stereo audio via the Stable Audio 3 checkpoint. -**Stable Audio 3 Medium** 워크플로우는 완벽한 기능을 갖춘 텍스트를 오디오로 변환하는 파이프라인입니다. 짧은 텍스트 아이디어, 선택적 지속 시간, 시드값, 카테고리를 제공하면, 워크플로우는 Qwen을 사용해 **카테고리 인식 리프롬프트 템플릿**으로 프롬프트를 확장한 후 Stable Audio 3 체크포인트를 통해 스테레오 오디오를 생성합니다. +**How to use**: +1. **Text idea** — Enter a short description of the sound, music, or effect you want (e.g. "upbeat electronic dance track with heavy bass") +2. **Duration** — Set the desired clip length in seconds (default varies) +3. **Seed** — Control reproducibility by adjusting the seed value +4. **Category** — Choose a reprompt preset: **Music**, **Instrument**, **SFX**, or **One-shot** +5. **Enable reprompt** — Toggle `use_reprompt` on to let Qwen expand your short idea into a detailed prompt before generation +6. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -**사용 방법**: -1. **텍스트 아이디어** — 원하는 사운드, 음악 또는 효과에 대한 간략한 설명을 입력하세요 (예: "강렬한 베이스가 있는 활기찬 일렉트로닉 댄스 트랙") -2. **지속 시간** — 원하는 클립 길이를 초 단위로 설정하세요 (기본값은 다양함) -3. **시드** — 시드 값을 조정해 재현성을 제어하세요 -4. **카테고리** — 리프롬프트 사전 설정을 선택하세요: **Music**, **Instrument**, **SFX**, 또는 **One-shot** -5. **리프롬프트 활성화** — `use_reprompt`을 켜서 Qwen이 짧은 아이디어를 자세한 프롬프트로 확장한 후 생성하도록 하세요 -6. **실행** (`Ctrl/Cmd + Enter`)을 클릭해 생성하세요. 오디오는 `ComfyUI/output/audio/`에 저장됩니다. +### Stable Audio 3.0 Medium Base (`audio_stable_audio_3_medium_base`) -### Stable Audio 3 Medium Base +Input a short text description of a sound, music, or effect. The workflow expands your prompt with Qwen and generates a stereo audio clip from Stable Audio 3. - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Stable Audio 3 Medium Base" 검색 - +Stable Audio 3 Medium Base workflow preview - - Comfy Cloud에서 열기 + + + Open in Comfy Cloud + + Download JSON or search "Stable Audio 3.0 Medium Base" in Template Library + + -![Stable Audio 3 Medium Base 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium_base-1.webp) - -Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화된 버전입니다. 완전한 텍스트 프롬프트를 기대하고 바로 모델에 전달합니다. 이미 상세한 프롬프트를 가지고 있고 더 빠른 생성을 원할 때 사용하세요. +A simplified version of Stable Audio 3 without Qwen reprompt expansion. It expects a complete text prompt and passes it directly to the model. Use this when you already have a detailed prompt and want faster generation. -**사용 방법**: -1. **텍스트 프롬프트** — 원하는 오디오에 대한 상세한 설명을 입력하세요 -2. **지속 시간** — 클립 길이를 초 단위로 설정하세요 -3. **시드** — 재현성을 제어하세요 -4. **실행** (`Ctrl/Cmd + Enter`)을 클릭해 생성하세요 +**How to use**: +1. **Text prompt** — Enter a detailed description of the audio you want +2. **Duration** — Set the clip length in seconds +3. **Seed** — Control reproducibility +4. Click **Run** (`Ctrl/Cmd + Enter`) to generate -## 모델 다운로드 +## Model download -워크플로우를 로드할 때 ComfyUI는 누락된 모델에 대한 다운로드 링크를 표시합니다. 수동으로 설정하려면 아래 파일을 다운로드해 올바른 폴더에 배치하세요. +When loading the workflow, ComfyUI will prompt you with download links for any missing models. To set up manually, download the files below and place them in the correct folders. -### 체크포인트 +### Checkpoints - - Medium 워크플로우용. models/checkpoints/에 배치하세요 + + + For the Medium workflow. Place in models/checkpoints/ - - Medium Base 워크플로우용. models/checkpoints/에 배치하세요 + + For the Medium Base workflow. Place in models/checkpoints/ + -체크포인트를 다음 위치에 배치하세요: +Place checkpoints in: ``` 📂 ComfyUI/ @@ -95,17 +102,19 @@ Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화 │ └── stable_audio_3_medium_base.safetensors ``` -### 텍스트 인코더 +### Text encoders - - 모든 Stable Audio 3 워크플로우에 필수. models/text_encoders/에 배치하세요 + + + Required for all Stable Audio 3 workflows. Place in models/text_encoders/ - - Medium 워크플로우용 (Qwen 리프롬프트). models/text_encoders/에 배치하세요 + + Required for the Medium workflow (Qwen reprompt). Place in models/text_encoders/ + -텍스트 인코더를 다음 위치에 배치하세요: +Place text encoders in: ``` 📂 ComfyUI/ @@ -115,4 +124,4 @@ Qwen 리프롬프트 확장을 포함하지 않은 Stable Audio 3의 간소화 │ └── qwen3.5_2b_bf16.safetensors ``` -파일을 배치한 후 ComfyUI에서 **R**을 눌러 노드를 새로고침하고 최신 모델을 로드하세요. +After placing the files, press **R** in ComfyUI to refresh nodes and load the latest models. diff --git a/ko/tutorials/basic/image-to-image.mdx b/ko/tutorials/basic/image-to-image.mdx index c88ba3809..9675b8bf7 100644 --- a/ko/tutorials/basic/image-to-image.mdx +++ b/ko/tutorials/basic/image-to-image.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Try It Yourself": 6b2e3528 --- - ## 이미지 기반 이미지 생성이란 무엇인가 이미지 기반 이미지 생성은 ComfyUI에서 사용자가 이미지를 입력하면 이를 기반으로 새로운 이미지를 생성할 수 있는 워크플로우입니다. diff --git a/ko/tutorials/basic/inpaint.mdx b/ko/tutorials/basic/inpaint.mdx index d542c9610..74f20d48e 100644 --- a/ko/tutorials/basic/inpaint.mdx +++ b/ko/tutorials/basic/inpaint.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "VAE Encoder (for Inpainting) Node": 222c8a2c --- - 이 기사에서는 AI 이미지 생성에서 인페인팅의 개념을 소개하고, ComfyUI에서 인페인팅 워크플로우를 만드는 방법을 안내합니다. 다음 내용을 다룹니다: - 이미지를 수정하는 데 인페인팅 워크플로우 사용하기 - ComfyUI 마스크 편집기를 사용해 마스크 그리기 diff --git a/ko/tutorials/basic/lora.mdx b/ko/tutorials/basic/lora.mdx index b7b54af8e..59e774baa 100644 --- a/ko/tutorials/basic/lora.mdx +++ b/ko/tutorials/basic/lora.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Try It Yourself": 9f276881 --- - **LoRA(저랭크 적응)**는 Stable Diffusion와 같은 대규모 생성 모델을 미세 조정하는 효율적인 기법입니다. 이는 사전 훈련된 모델에 학습 가능한 저랭크 행렬을 도입하여, 전체 모델을 다시 훈련하는 대신 일부 파라미터만 조정함으로써 더 낮은 계산 비용으로 특정 작업에 대한 최적화를 달성합니다. SD1.5와 같은 기본 모델과 비교하면 LoRA 모델은 크기가 작고 훈련하기도 쉽습니다. diff --git a/ko/tutorials/basic/multiple-loras.mdx b/ko/tutorials/basic/multiple-loras.mdx index c45fe1c61..ebfe59411 100644 --- a/ko/tutorials/basic/multiple-loras.mdx +++ b/ko/tutorials/basic/multiple-loras.mdx @@ -33,7 +33,6 @@ translationFrom: tutorials/basic/multiple-loras.mdx 아래 이미지를 다운로드한 후 **ComfyUI로 드래그**하여 워크플로우를 로드하세요: ![ComfyUI 워크플로우 - 다중 LoRA](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/multiple_loras.png) - 워크플로우 JSON이 메타데이터에 포함된 이미지는 바로 ComfyUI로 드래그하거나, 메뉴 `Workflows` -> `Open (ctrl+o)`를 통해 로드할 수 있습니다. diff --git a/ko/tutorials/basic/outpaint.mdx b/ko/tutorials/basic/outpaint.mdx index fb246ba21..fcdf021b1 100644 --- a/ko/tutorials/basic/outpaint.mdx +++ b/ko/tutorials/basic/outpaint.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "ComfyUI Outpainting Workflow Example Explanation": e06f66e2 --- - 이 가이드에서는 AI 이미지 생성에서 아웃페인팅의 개념과 ComfyUI에서 아웃페인팅 워크플로우를 만드는 방법을 소개합니다. 다음 내용을 다룹니다: - 아웃페인팅 워크플로우를 사용해 이미지를 확장하는 방법 - ComfyUI에서 아웃페인팅 관련 노드를 이해하고 사용하는 방법 diff --git a/ko/tutorials/basic/text-to-image.mdx b/ko/tutorials/basic/text-to-image.mdx index 5752556e2..fd8a714ef 100644 --- a/ko/tutorials/basic/text-to-image.mdx +++ b/ko/tutorials/basic/text-to-image.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Introduction to SD1.5 Model": cd91e138 --- - 이 가이드는 ComfyUI의 텍스트 기반 이미지 생성 워크플로우를 소개하고 다양한 ComfyUI 노드의 기능과 사용법을 이해하도록 돕습니다. 이 문서에서는 다음을 다룹니다: diff --git a/ko/tutorials/basic/upscale.mdx b/ko/tutorials/basic/upscale.mdx index 9964ca6ea..f32542b95 100644 --- a/ko/tutorials/basic/upscale.mdx +++ b/ko/tutorials/basic/upscale.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Additional Tips": 5352d8c0 --- - ## 이미지 업스케일링이란? 이미지 업스케일링은 알고리즘을 사용해 저해상도 이미지를 고해상로 변환하는 과정입니다. @@ -54,7 +53,6 @@ translationBlockHashes: - ### 워크플로우 및 자산 다음 이미지를 ComfyUI로 다운로드하고 끌어다 놓아 기본 업스케일링 워크플로우를 로드하세요: diff --git a/ko/tutorials/controlnet/controlnet.mdx b/ko/tutorials/controlnet/controlnet.mdx index ca0a1ca24..b6ee9e2c2 100644 --- a/ko/tutorials/controlnet/controlnet.mdx +++ b/ko/tutorials/controlnet/controlnet.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Start Your Exploration": c5020328 --- - AI 이미지 생성에서 이미지 생성을 정밀하게 제어하려면 단 한 번의 클릭만으로는 불가능합니다. 일반적으로 만족스러운 이미지를 얻기 위해 수많은 생성 시도를 거쳐야 합니다. 그러나 **ControlNet**의 등장으로 이러한 어려움이 효과적으로 해결되었습니다. ControlNet은 확산 모델(예: Stable Diffusion) 기반의 조건부 제어 생성 모델로, 2023년 [Lvmin Zhang](https://lllyasviel.github.io/)과 Maneesh Agrawala 등이 발표한 논문 [텍스트-투-이미지 확산 모델에 조건부 제어 추가하기](https://arxiv.org/abs/2302.05543)에서 처음 제안되었습니다. diff --git a/ko/tutorials/controlnet/depth-controlnet.mdx b/ko/tutorials/controlnet/depth-controlnet.mdx index eded209da..80a7f7989 100644 --- a/ko/tutorials/controlnet/depth-controlnet.mdx +++ b/ko/tutorials/controlnet/depth-controlnet.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Combining Depth Control with Other Techniques": 9e3bcc3d --- - ## 깊이 맵과 Depth ControlNet 소개 깊이 맵은 회색조 값을 사용해 장면 내 객체와 관찰자 또는 카메라 간의 거리를 나타내는 특수한 이미지입니다. 깊이 맵에서 회색조 값은 거리와 반비례합니다: 밝은 영역(흰색에 가까움)은 더 가까운 객체를, 어두운 영역(검정에 가까움)은 더 먼 객체를 나타냅니다. diff --git a/ko/tutorials/controlnet/depth-t2i-adapter.mdx b/ko/tutorials/controlnet/depth-t2i-adapter.mdx index 413009f36..6af569520 100644 --- a/ko/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/ko/tutorials/controlnet/depth-t2i-adapter.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Characteristics of T2I Adapter Usage": d485c810 --- - ## T2I 어댑터 소개 [T2I-Adapter](https://huggingface.co/TencentARC/T2I-Adapter)는 [Tencent ARC Lab](https://github.com/TencentARC)에서 개발한 경량 어댑터로, 텍스트 기반 이미지 생성 모델(예: Stable Diffusion)의 구조적, 색상 및 스타일 제어 능력을 향상시키기 위해 설계되었습니다. diff --git a/ko/tutorials/controlnet/mixing-controlnets.mdx b/ko/tutorials/controlnet/mixing-controlnets.mdx index 2063e94ab..ad44a50c6 100644 --- a/ko/tutorials/controlnet/mixing-controlnets.mdx +++ b/ko/tutorials/controlnet/mixing-controlnets.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Multi-dimensional Control Applications for a Single Subject": 5790d753 --- - AI 이미지 생성에서 단일 제어 조건은 종종 복잡한 장면의 요구를 충족하지 못합니다. 여러 ControlNet을 혼합하면 이미지의 서로 다른 영역이나 측면을 동시에 제어할 수 있어, 이미지 생성을 더욱 정밀하게 제어할 수 있습니다. 특정 시나리오에서는 여러 ControlNet을 혼합함으로써 서로 다른 제어 조건의 특성을 활용해 더욱 세밀한 조건 제어를 달성할 수 있습니다: diff --git a/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx b/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx index 16a13ffcf..1040c5cb9 100644 --- a/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/ko/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Advantages of 2-Pass Image Generation": fba5db79 --- - ## OpenPose 소개 [OpenPose](https://github.com/CMU-Perceptual-Computing-Lab/openpose)는 카네기 멜론 대학(CMU)에서 개발한 오픈소스 실시간 다인종 자세 추정 시스템으로, 컴퓨터 비전 분야에서 중요한 돌파구를 마련했습니다. 이 시스템은 이미지 내 여러 사람을 동시에 감지하며 다음을 캡처합니다: diff --git a/ko/tutorials/flux/flux-1-controlnet.mdx b/ko/tutorials/flux/flux-1-controlnet.mdx index e3d0d9c95..6648fce24 100644 --- a/ko/tutorials/flux/flux-1-controlnet.mdx +++ b/ko/tutorials/flux/flux-1-controlnet.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Community Versions of Flux Controlnets": 29728dcf --- - - ![Flux.1 Canny Controlnet](/images/tutorial/flux/flux-1-canny-controlnet.png) ![Flux.1 Depth Controlnet](/images/tutorial/flux/flux-1-depth-controlnet.png) @@ -48,14 +46,36 @@ Depth 버전은 깊이 맵 추출 기법을 통해 원본 이미지의 공간적 - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -## FLUX.1-Canny-dev 전체 버전 워크플로우 +### Flux.1 Canny Model (`flux_canny_model_example`) + +Generate images guided by edge detection using Flux.1 Canny. + +Flux.1 Canny 워크플로 미리보기 - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Canny" 검색 - Comfy Cloud에서 열기 + + Comfy Cloud에서 열기 + + + Download JSON or search "Flux.1 Canny" in Template Library + -### 1. 워크플로우 및 자산 +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_canny_model_example_input_image.png` + + + +
+ flux_canny_model_example_input_image.png +
+ +**## 1. 워크플로우 및 자산 아래 워크플로우 이미지를 다운로드해 ComfyUI로 드래그하여 워크플로우를 로드해주세요. @@ -116,14 +136,34 @@ ComfyUI/ - [ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet) - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -## FLUX.1-Depth-dev-lora 워크플로우 +### Flux.1 Depth Lora (`flux_depth_lora_example`) + +Generate images guided by depth information using Flux.1 LoRA. + +Flux.1 Depth LoRA 워크플로 미리보기 + + + + Comfy Cloud에서 열기 + + + Download JSON or search "Flux.1 Depth LoRA" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Depth LoRA" 검색 - Comfy Cloud에서 열기 + + `LoadImage` node 17 · `flux_depth_lora_example_input_image.png` + -LoRA 버전 워크플로우는 완전한 버전을 기반으로 LoRA 모델을 추가한 것입니다. [Flux 워크플로우의 전체 버전](/ko/tutorials/flux/flux-1-text-to-image)과 비교해, 해당 LoRA 모델을 로드하고 사용하는 노드가 추가되었습니다. +
+ flux_depth_lora_example_input_image.png +
### 1. 워크플로우 및 자산 diff --git a/ko/tutorials/flux/flux-1-fill-dev.mdx b/ko/tutorials/flux/flux-1-fill-dev.mdx index b8e20cd31..5192cad49 100644 --- a/ko/tutorials/flux/flux-1-fill-dev.mdx +++ b/ko/tutorials/flux/flux-1-fill-dev.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Flux.1 Fill dev Outpainting Workflow": dd827926 --- - - ![Flux.1 fill dev](/images/tutorial/flux/flux-fill-dev-demo.jpeg) ## Flux.1 fill dev 모델 소개 @@ -61,24 +59,41 @@ ComfyUI/ ## Flux.1 Fill dev 인페인팅 워크플로우 -### 1. 인페인팅 워크플로우 및 자산 +### Flux.1 Inpaint (`flux_fill_inpaint_example`) + +Fill missing parts of images using Flux.1 Fill Inpainting. + +Flux.1 inpaint 워크플로 미리보기 - JSON 다운로드 또는 템플릿 라이브러리에서 "flux_fill_inpaint" 검색 - Comfy Cloud에서 열기 + + Comfy Cloud에서 열기 + + + Download JSON or search "Flux.1 Inpaint" in Template Library + +**입력 자료** -아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. -![ComfyUI Flux.1 인페인팅](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) +Upload this file to the matching `LoadImage` node: -아래 이미지를 다운로드해 입력 이미지로 사용하겠습니다. -![ComfyUI Flux.1 인페인팅 입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input.png) + + + `LoadImage` node 17 · `flux_fill_inpaint_example_input_image.png` + + + +
+ flux_fill_inpaint_example_input_image.png +
+ +**출력 예시** - -해당 이미지는 이미 알파 채널을 포함하고 있으므로 별도로 마스크를 그릴 필요가 없습니다. -자신만의 마스크를 그리고 싶다면, [여기를 클릭](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input_original.png)해 마스크가 없는 이미지를 받고, [ComfyUI 레이아웃 인페인팅 예제](/ko/tutorials/basic/inpaint#using-the-mask-editor)의 MaskEditor 사용법을 참고해 `Load Image` 노드에서 마스크를 그리는 방법을 알아보세요. - +
+ 입력 이미지 + Flux.1 inpaint 출력 예시 +
### 2. 워크플로우 실행 단계 diff --git a/ko/tutorials/flux/flux-1-kontext-dev.mdx b/ko/tutorials/flux/flux-1-kontext-dev.mdx index 24962af73..158308de8 100644 --- a/ko/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ko/tutorials/flux/flux-1-kontext-dev.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Flux.1 Kontext Dev Workflow": ffae9c25 --- - - import PromptTechniques from "/snippets/ko/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' @@ -79,14 +77,43 @@ FLUX.1 Kontext는 블랙 포레스트 랩스가 개발한 획기적인 멀티모 │ └── t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn_scaled.safetensors ``` -## Flux.1 Kontext Dev 워크플로우 +### Flux Kontext Dev Image Edit (`flux_kontext_dev_basic`) + +Smart image editing that keeps characters consistent, edits specific parts without affecting others, and preserves original styles. + +Flux Kontext Dev 워크플로 미리보기 + + + + Comfy Cloud에서 열기 + + + Download JSON or search "Flux Kontext Dev" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux Kontext Dev" 검색 - Comfy Cloud에서 열기 + + `LoadImage` node 190 · `flux_kontext_dev_basic_input_image.jpg` + -이 워크플로우는 `Load Image(from output)` 노드를 사용해 편집할 이미지를 불러오므로, 여러 차례의 편집을 위해 편집된 이미지에 더욱 편리하게 접근할 수 있습니다. +
+ flux_kontext_dev_basic_input_image.jpg +
+ +**출력 예시** + +
+ 입력 이미지 + Flux Kontext Dev 출력 예시 +
+ +This workflow uses the `Load Image(from output)` node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing. ### 1. 워크플로우 및 입력 이미지 다운로드 diff --git a/ko/tutorials/flux/flux-1-text-to-image.mdx b/ko/tutorials/flux/flux-1-text-to-image.mdx index 9cb3e38c6..79f6f6420 100644 --- a/ko/tutorials/flux/flux-1-text-to-image.mdx +++ b/ko/tutorials/flux/flux-1-text-to-image.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Flux.1 FP8 Checkpoint Version Text-to-Image Example": daf5c58e --- - - ![Flux](/images/tutorial/flux/flux_example.png) Flux는 120억 파라미터를 가진 오픈소스 텍스트 기반 이미지 생성 모델 중 가장 큰 모델 중 하나로, 원본 파일 크기는 약 23GB입니다. 이 모델은 전 스테이블 디퓨전 팀원들이 설립한 [Black Forest Labs](https://blackforestlabs.ai/)에서 개발되었습니다. Flux는 뛰어난 이미지 품질과 유연성으로 고화질의 다양한 이미지를 생성할 수 있습니다. @@ -47,35 +45,41 @@ Flux는 뛰어난 이미지 품질과 유연성으로 고화질의 다양한 이 ![Flux 계약](/images/tutorial/flux/flux_agreement.jpg) -### Flux.1 Dev +### Flux.1 Dev fp8: Text to Image (`flux_dev_checkpoint_example`) -#### 1. 워크플로우 파일 +Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices with limited VRAM, requires only one model file, but image quality is slightly reduced compared to the full version. -아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. -![Flux Dev 원본 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) +Flux.1 Dev fp8 워크플로 미리보기 - 이 워크플로우를 Comfy Cloud에서 실행하세요 + Comfy Cloud에서 이 워크플로 실행 - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Dev" 검색 + + Download JSON or search "Flux.1 Dev fp8" in Template Library -#### 2. 수동 모델 설치 + +**출력 예시** + +![Flux.1 Dev fp8 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_checkpoint_example.png) + +#### 1. Workflow File + +#### 2. Manual Model Installation -- `flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) 계약에 동의해야 합니다. -- VRAM이 부족한 경우, [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true)를 사용해 `t5xxl_fp16.safetensors` 파일을 대체해보세요. +- The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) agreement before downloading via browser. +- If your VRAM is low, you can try using [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) to replace the `t5xxl_fp16.safetensors` file. -다음 모델 파일을 다운로드하세요: +Please download the following model files: - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) - [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) -저장 위치: +Storage location: ``` ComfyUI/ ├── models/ @@ -88,99 +92,73 @@ ComfyUI/ │ └── flux1-dev.safetensors ``` -#### 3. 워크플로우 실행 단계 +#### 3. Steps to Run the Workflow -아래 이미지를 참조해 모든 모델 파일이 올바르게 로드되었는지 확인하세요. +Please refer to the image below to ensure all model files are loaded correctly -![ComfyUI Flux Dev 워크플로우](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) +![ComfyUI Flux Dev Workflow](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) -1. `DualCLIPLoader` 노드에 다음 모델이 로드되었는지 확인하세요: +1. Ensure the `DualCLIPLoader` node has the following models loaded: - clip_name1: t5xxl_fp16.safetensors - clip_name2: clip_l.safetensors -2. `Load Diffusion Model` 노드에 `flux1-dev.safetensors`가 로드되었는지 확인하세요. -3. `Load VAE` 노드에 `ae.safetensors`가 로드되었는지 확인하세요. -4. `Queue` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. +2. Ensure the `Load Diffusion Model` node has `flux1-dev.safetensors` loaded +3. Make sure the `Load VAE` node has `ae.safetensors` loaded +4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -Flux의 뛰어난 프롬프트 추종 능력 덕분에 부정적인 프롬프트는 필요하지 않습니다. +Thanks to Flux's excellent prompt following capability, we don't need any negative prompts -### Flux.1 Schnell -#### 1. 워크플로우 파일 +### Flux.1 Schnell FP8 (`flux_schnell`) -아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. +Quickly generate images with Flux.1 Schnell fp8 quantized version. Ideal for low-end hardware, requires only 4 steps to generate images. -![Flux Schnell 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) +Flux.1 Schnell FP8 checkpoint 워크플로 미리보기 - 이 워크플로우를 Comfy Cloud에서 실행하세요 + Comfy Cloud에서 이 워크플로 실행 - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Schnell" 검색 + + Download JSON or search "Flux.1 Schnell FP8" in Template Library -#### 2. 수동 모델 설치 - -이 워크플로우에서는 두 모델 파일만 Flux1 Dev 버전 워크플로우와 다릅니다. t5xxl의 경우, 더 나은 결과를 위해 여전히 fp16 버전을 사용할 수 있습니다. -- **t5xxl_fp16.safetensors** -> **t5xxl_fp8.safetensors** -- **flux1-dev.safetensors** -> **flux1-schnell.safetensors** - +Please download the image below and drag it into ComfyUI to load the workflow. -완전한 모델 파일 목록: -- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) -- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) +Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. -파일 저장 위치: -``` -ComfyUI/ -├── models/ -│ ├── text_encoders/ -│ │ ├── clip_l.safetensors -│ │ └── t5xxl_fp8_e4m3fn.safetensors -│ ├── vae/ -│ │ └── ae.safetensors -│ └── diffusion_models/ -│ └── flux1-schnell.safetensors -``` - -#### 3. 워크플로우 실행 단계 - -![Flux Schnell 버전 워크플로우](/images/tutorial/flux/flow_diagram_flux_schnell_t5fp8.jpg) - -1. `DualCLIPLoader` 노드에 다음 모델이 로드되었는지 확인하세요: - - clip_name1: t5xxl_fp8_e4m3fn.safetensors - - clip_name2: clip_l.safetensors -2. `Load Diffusion Model` 노드에 `flux1-schnell.safetensors`가 로드되었는지 확인하세요. -3. `Load VAE` 노드에 `ae.safetensors`가 로드되었는지 확인하세요. -4. `Queue` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. +Ensure that the corresponding `Load Checkpoint` node loads `flux1-schnell-fp8.safetensors`, and you can try to run the workflow. ## Flux.1 FP8 체크포인트 버전 텍스트 기반 이미지 생성 예시 fp8 버전은 원본 Flux.1 fp16 버전의 양자화된 버전입니다. 어느 정도는 이 버전의 품질이 fp16 버전보다 낮아지지만, VRAM 요구량도 줄어들고, 모델 파일을 하나만 설치하면 바로 실행해볼 수 있습니다. -### Flux.1 Dev +### Flux.1 Dev: Text to Image (`flux_dev_full_text_to_image`) + +Generate high-quality images with Flux Dev full version. Requires larger VRAM and multiple model files, but provides the best prompt following capability. + +Flux.1 Dev text-to-image 워크플로 미리보기 - 이 워크플로우를 Comfy Cloud에서 실행하세요 + Comfy Cloud에서 이 워크플로 실행 - - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Dev FP8" 검색 + + Download JSON or search "Flux.1 Dev FP8" in Template Library -아래 이미지를 다운로드해 ComfyUI에 드래그하여 워크플로우를 로드하세요. +**출력 예시** + +![Flux.1 Dev FP8 checkpoint 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_full_text_to_image.png) -![Flux Dev fp8 체크포인트 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) +Please download the image below and drag it into ComfyUI to load the workflow. -[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true)를 다운로드해 `ComfyUI/models/checkpoints/` 디렉토리에 저장하세요. +Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. -해당 `Load Checkpoint` 노드가 `flux1-dev-fp8.safetensors`를 로드하도록 설정했는지 확인하고, 워크플로우를 실행해보세요. +Ensure that the corresponding `Load Checkpoint` node loads `flux1-dev-fp8.safetensors`, and you can try to run the workflow. ### Flux.1 Schnell diff --git a/ko/tutorials/flux/flux-1-uso.mdx b/ko/tutorials/flux/flux-1-uso.mdx index 06979597f..9e6e2413d 100644 --- a/ko/tutorials/flux/flux-1-uso.mdx +++ b/ko/tutorials/flux/flux-1-uso.mdx @@ -25,26 +25,43 @@ USO는 세 가지 주요 방식을 지원합니다: -### 1. 워크플로우 및 입력 +### Flux.1 Dev USO Reference Image Generation (`flux1_dev_uso_reference_image_gen`) -아래 이미지를 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. +Use reference images to control both style and subject. Keep your character's face while changing artistic style, or apply artistic styles to new scenes. -![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) +Flux.1 Dev USO reference image 워크플로 미리보기 - + + Comfy Cloud에서 이 워크플로 실행 + + Download the workflow JSON and drag it into ComfyUI - 이 워크플로우를 Comfy Cloud에서 실행하세요 +**입력 자료** -아래 이미지를 입력 이미지로 사용하세요. +Upload this file to the matching `LoadImage` node: -![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/input.png) + + + `LoadImage` node 47 · `flux1_dev_uso_reference_image_gen_input_image.png` + + -### 2. 모델 링크 +
+ flux1_dev_uso_reference_image_gen_input_image.png +
+**출력 예시** + +
+ 입력 이미지 + Flux.1 Dev USO 출력 예시 +
+ +### 2. 모델 링크 **체크포인트** @@ -77,7 +94,6 @@ USO는 세 가지 주요 방식을 지원합니다: │ │ └── sigclip_vision_patch14_384.safetensors ``` - ### 3. 워크플로우 지침 ![워크플로우 지침](/images/tutorial/flux/flux1_uso_reference_image_gen.jpg) diff --git a/ko/tutorials/flux/flux-2-dev.mdx b/ko/tutorials/flux/flux-2-dev.mdx index 1696ffdb8..0a98268e2 100644 --- a/ko/tutorials/flux/flux-2-dev.mdx +++ b/ko/tutorials/flux/flux-2-dev.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Model links": f4c1677b --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" diff --git a/ko/tutorials/flux/flux-2-klein.mdx b/ko/tutorials/flux/flux-2-klein.mdx index da2f98a4b..877c5c60a 100644 --- a/ko/tutorials/flux/flux-2-klein.mdx +++ b/ko/tutorials/flux/flux-2-klein.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Flux.2 Klein 9B Model Downloads": 0f38ad48 --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" diff --git a/ko/tutorials/flux/flux1-krea-dev.mdx b/ko/tutorials/flux/flux1-krea-dev.mdx index a005f4cf7..1bf94ad9b 100644 --- a/ko/tutorials/flux/flux1-krea-dev.mdx +++ b/ko/tutorials/flux/flux1-krea-dev.mdx @@ -22,75 +22,81 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **모델 라이선스** 이 모델은 [flux-1-dev-non-commercial-license](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/LICENSE.md)에 따라 배포됩니다. -## Flux.1 Krea Dev ComfyUI 워크플로우 +### Flux.1 Krea Dev (`flux1_krea_dev`) - +A fine-tuned FLUX model pushing photorealism to the max. -#### 1. 워크플로우 파일 - -아래 이미지 또는 JSON 파일을 다운로드한 후 ComfyUI로 드래그하여 해당 워크플로우를 로드하세요. -![Flux Krea Dev 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) +Flux.1 Krea Dev 워크플로 미리보기 - 이 워크플로우를 Comfy Cloud에서 실행하세요 - JSON 다운로드 또는 템플릿 라이브러리에서 "Flux.1 Krea Dev" 검색 + + Comfy Cloud에서 이 워크플로 실행 + + + Download JSON or search "Flux.1 Krea Dev" in Template Library + +**출력 예시** + +![Flux.1 Krea Dev 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux1_krea_dev.png) + +#### 1. Workflow Files -#### 2. 수동 모델 설치 +#### 2. Manual Model Installation -다음 모델 파일을 다운로드해 주세요: -**확산 모델** +Please download the following model files: +**Diffusion model** - [flux1-krea-dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/FLUX.1-Krea-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-krea-dev_fp8_scaled.safetensors) -더 높은 품질을 원하고 VRAM이 충분하다면 원본 모델 가중치를 사용해 보실 수 있습니다. +If you want to pursue higher quality and have enough VRAM, you can try the original model weights - [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) -`flux1-dev.safetensors` 파일은 브라우저를 통해 다운로드하기 전에 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 약정에 동의해야 합니다. +The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) agreement before downloading via browser. -이전에 Flux 관련 워크플로우를 사용한 적이 있다면, 다음 모델들은 동일하므로 다시 다운로드할 필요가 없습니다. +If you have used Flux related workflows before, the following models are the same and don't need to be downloaded again -**텍스트 인코더** +**Text encoders** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM이 32GB 이상인 경우 권장합니다. -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) 낮은 VRAM용 +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM **VAE** - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -파일 저장 위치: +File save location: ``` ComfyUI/ ├── models/ │ ├── diffusion_models/ -│ │ └── flux1-krea-dev_fp8_scaled.safetensors 또는 flux1-krea-dev.safetensors +│ │ └── flux1-krea-dev_fp8_scaled.safetensors or flux1-krea-dev.safetensors │ ├── text_encoders/ │ │ ├── clip_l.safetensors -│ │ └── t5xxl_fp16.safetensors 또는 t5xxl_fp8_e4m3fn.safetensors +│ │ └── t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors │ ├── vae/ │ │ └── ae.safetensors ``` -#### 3. 워크플로우가 올바르게 실행되는지 단계별 확인 방법 +#### 3. Step-by-step Verification to Ensure Workflow Runs Properly -낮은 VRAM 사용자의 경우 이 모델이 기기에서 원활하게 작동하지 않을 수 있으며, 커뮤니티에서 FP8 또는 GGUF 버전을 제공할 때까지 기다릴 수 있습니다. + For low VRAM users, this model may not run smoothly on your device, you can wait for the community to provide FP8 or GGUF version. -모든 모델 파일이 올바르게 로드되었는지 아래 이미지를 참고해 주세요. +Please refer to the image below to ensure all model files have been loaded correctly -![ComfyUI Flux Krea Dev 워크플로우](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) +![ComfyUI Flux Krea Dev Workflow](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) -1. `Load Diffusion Model` 노드에 `flux1-krea-dev_fp8_scaled.safetensors` 또는 `flux1-krea-dev.safetensors`가 로드되었는지 확인하세요. - - 낮은 VRAM 사용자는 `flux1-krea-dev_fp8_scaled.safetensors`를 권장합니다. - - `flux1-krea-dev.safetensors`는 원본 가중치로, 24GB 등 충분한 VRAM을 보유한 경우 더 높은 품질을 위해 사용할 수 있습니다. -2. `DualCLIPLoader` 노드에 다음 모델들이 로드되었는지 확인하세요: - - clip_name1: t5xxl_fp16.safetensors 또는 t5xxl_fp8_e4m3fn.safetensors +1. Ensure that `flux1-krea-dev_fp8_scaled.safetensors` or `flux1-krea-dev.safetensors` is loaded in the `Load Diffusion Model` node + - `flux1-krea-dev_fp8_scaled.safetensors` is recommended for low VRAM users + - `flux1-krea-dev.safetensors` is the original weights, if you have enough VRAM like 24GB you can use it for better quality +2. Ensure the following models are loaded in the `DualCLIPLoader` node: + - clip_name1: t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors - clip_name2: clip_l.safetensors -3. `Load VAE` 노드에 `ae.safetensors`가 로드되었는지 확인하세요. -4. `Queue` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. \ No newline at end of file +3. Ensure that `ae.safetensors` is loaded in the `Load VAE` node +4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow diff --git a/ko/tutorials/image/anima/anima.mdx b/ko/tutorials/image/anima/anima.mdx index 038e74d36..dfa575030 100644 --- a/ko/tutorials/image/anima/anima.mdx +++ b/ko/tutorials/image/anima/anima.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Limitations": 0adbf360 --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' @@ -38,47 +37,57 @@ Anima는 두 가지 워크플로를 제공합니다. 일반 사용을 위한 베 이 워크플로는 모듈식 처리를 위해 Subgraph 노드를 사용합니다. Subgraph 문서를 확인해 워크플로를 사용자 정의하고 확장하는 방법을 알아보세요. -### Anima Base v1: 텍스트 기반 이미지 생성 +### Anima Base v1: Text to Image (`image_anima_base_v1`) - - JSON 파일 다운로드 또는 템플릿 라이브러리에서 "Anima Base v1" 검색 - +Input a text prompt describing an anime or artistic illustration. Generate a non-photorealistic image focused on anime concepts, characters, or styles. - +Anima Base v1 text-to-image 워크플로 미리보기 + + + Comfy Cloud에서 열기 + + Download JSON or search "Anima Base v1" in Template Library + + -#### 시작하기 +**출력 예시** -1. ComfyUI를 최신 버전으로 업데이트하거나 [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima)를 사용하세요 -2. **템플릿**으로 이동해 **Anima Base v1**을 검색하세요 -3. **Anima Base v1: 텍스트 기반 이미지 생성** 워크플로를 선택하세요 -4. 누락된 모델을 다운로드하고([모델 다운로드](#anima-model-downloads) 참조), 프롬프트를 입력한 후 **큐**를 클릭하세요 +![Anima Base v1 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_base_v1.png) -#### 예시 출력 +#### Get started -Anima Base v1 예시 출력 +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) +2. Go to **Template** and search for **Anima Base v1** +3. Select the **Anima Base v1: Text to Image** workflow +4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** -### Anima 미리보기: 애니메이션 텍스트 기반 이미지 생성 +### Anima Preview: Anime Text-to-Image Generation (`image_anima_preview`) - - JSON 파일 다운로드 또는 템플릿 라이브러리에서 "Anima Preview" 검색 - +Input a text prompt to generate an anime-style image using the Anima model. Configure settings like steps and CFG scale to control the output. + +Anima Preview text-to-image 워크플로 미리보기 - + + Comfy Cloud에서 열기 + + Download JSON or search "Anima Preview" in Template Library + + -#### 시작하기 +**출력 예시** -1. ComfyUI를 최신 버전으로 업데이트하거나 [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima)를 사용하세요 -2. **템플릿**으로 이동해 **Anima Preview**를 검색하세요 -3. **Anima 애니메이션 텍스트 기반 이미지 생성** 워크플로를 선택하세요 -4. 누락된 모델을 다운로드하고([모델 다운로드](#anima-model-downloads) 참조), 프롬프트를 입력한 후 **큐**를 클릭하세요 +![Anima Preview 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_preview.png) -#### 예시 출력 +#### Get started -Anima Preview 예시 출력 +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) +2. Go to **Template** and search for **Anima Preview** +3. Select the **Anima Anime Text-to-Image Generation** workflow +4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** ## Anima 모델 다운로드 diff --git a/ko/tutorials/image/boogu/boogu-image-0.1.mdx b/ko/tutorials/image/boogu/boogu-image-0.1.mdx index ae1e7aa94..fe13835d4 100644 --- a/ko/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ko/tutorials/image/boogu/boogu-image-0.1.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Boogu-Image-0.1-Edit image editing workflow": 3c1752bb --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" **Boogu-Image-0.1**은 Apache-2.0 오픈소스 통합 이미지 생성 및 편집 모델 제품군입니다. 이해와 생성을 통합한 시스템을 사용하여 사진, 텍스트 렌더링, 스타일화 및 이미지 편집 작업 전반에서 경쟁력 있는 성능을 제공합니다. diff --git a/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index e72a01eb8..d931ca71f 100644 --- a/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ko/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -16,7 +16,6 @@ Cosmos-Predict2는 텍스트 기반 이미지 생성(Text2Image)과 비디오 GitHub: [Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) - 이 가이드에서는 ComfyUI에서 **텍스트 기반 이미지 생성** 워크플로우를 완료하는 과정을 안내합니다. 비디오 생성 부분에 대해서는 다음 섹션을 참고해 주세요: @@ -37,7 +36,6 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- ![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/cosmos/predict2/cosmos_predict2_2B_t2i.png) - ### 2. 수동 모델 설치 모델 다운로드가 성공하지 못했다면, 이 섹션에서 직접 수동으로 다운로드해 보세요. @@ -56,7 +54,6 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - 파일 저장 위치 ``` 📂 ComfyUI/ diff --git a/ko/tutorials/image/ernie-image/ernie-image.mdx b/ko/tutorials/image/ernie-image/ernie-image.mdx index 8c72a09e3..14844da0b 100644 --- a/ko/tutorials/image/ernie-image/ernie-image.mdx +++ b/ko/tutorials/image/ernie-image/ernie-image.mdx @@ -31,15 +31,26 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ## ERNIE-Image 텍스트 기반 이미지 생성 워크플로 - - ERNIE-Image 텍스트 기반 이미지 생성 워크플로 JSON 파일을 다운로드하세요. - + + +### Ernie Image: Text to Image (`image_ernie_image`) + +Generate images from text prompts using the ERNIE-Image model. Input a text description to produce detailed, structured visuals with a broad stylistic range. + +ERNIE-Image text-to-image 워크플로 미리보기 - - 이 워크플로를 Comfy Cloud에서 바로 실행하세요. + + + Comfy Cloud에서 이 워크플로를 직접 실행 + + Download the ERNIE-Image text-to-image workflow JSON file + + - +**출력 예시** + +![ERNIE-Image 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image.png) ### 시작하기 @@ -52,21 +63,20 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" 모든 재포장된 모델 파일은 Hugging Face의 [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image)에서 확인할 수 있습니다. - + + ERNIE-Image용 디퓨전 모델입니다. - - + ERNIE-Image용 텍스트 인코더입니다. - - + ERNIE-Image용 프롬프트 향상기 텍스트 인코더입니다. - - + ERNIE-Image용 VAE입니다. + **모델 저장 위치** @@ -84,33 +94,43 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" ## ERNIE-Image-Turbo -[ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo)는 DMD와 RL로 최적화된 더 빠른 변형으로, 표준 모델이 약 50단계를 필요로 하는 것과 달리 단 **8단계**만으로 이미지를 생성합니다. +[ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo)는 DMD 및 RL로 최적화된 더 빠른 변형으로, 표준 모델의 약 50단계 대비 **8단계**만으로 이미지를 생성합니다. - - ERNIE-Image-Turbo 텍스트 기반 이미지 생성 워크플로 JSON 파일을 다운로드하세요. - +### Ernie Image Turbo: Text To Image (`image_ernie_image_turbo`) + +Generate images from text prompts using the ERNIE-Image turbo model. Input a text description and receive a high-quality image with precise text rendering. - - 이 워크플로를 Comfy Cloud에서 바로 실행하세요. +ERNIE-Image-Turbo text-to-image 워크플로 미리보기 + + + + Comfy Cloud에서 이 워크플로를 직접 실행 + + + Download the ERNIE-Image-Turbo text-to-image workflow JSON file + + +**출력 예시** + +![ERNIE-Image-Turbo 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image_turbo.png) ### ERNIE-Image-Turbo 모델 다운로드 - - ERNIE-Image-Turbo용 diffusion 모델입니다. + + + ERNIE-Image-Turbo용 디퓨전 모델입니다. - - + ERNIE-Image-Turbo용 텍스트 인코더입니다. - - + ERNIE-Image-Turbo용 프롬프트 향상기 텍스트 인코더입니다. - - + ERNIE-Image-Turbo용 VAE입니다. + **모델 저장 위치** diff --git a/ko/tutorials/image/hidream/hidream-e1.mdx b/ko/tutorials/image/hidream/hidream-e1.mdx index 7aa3f3868..14938bc4d 100644 --- a/ko/tutorials/image/hidream/hidream-e1.mdx +++ b/ko/tutorials/image/hidream/hidream-e1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ![HiDream-E1 데모](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -74,19 +72,36 @@ HiDream-E1은 HiDream-ai가 공식적으로 오픈소스로 배포한 대규모 │ └── hidream_e1_full_bf16.safetensors ``` - ## HiDream E1.1 ComfyUI 네이티브 워크플로우 예시 -E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전은 동적 1메가픽셀 해상도를 지원하며, 워크플로우에서는 `Scale Image to Total Pixels` 노드를 사용해 입력 이미지를 동적으로 100만 픽셀로 조정합니다. +### HiDream E1.1 Image Editing (`hidream_e1_1`) - -테스트 중 VRAM 사용량 참고: -1. A100 40GB (VRAM 사용률 95%): 1세대: 211초, 2세대: 73초 +Edit images with HiDream E1.1. Superior image quality and editing accuracy compared to HiDream-E1-Full. -2. 4090D 24GB (VRAM 사용률 98%) -- 풀버전: 메모리 부족 -- FP8_e4m3fn_fast (VRAM 98%) 1세대: 120초, 2세대: 91초 - +HiDream E1.1 image editing 워크플로 미리보기 + + + + Comfy Cloud에서 열기 + + + Download JSON or search "HiDream E1.1" in Template Library + + + +**입력 자료** + +이 파일을 일치하는 `LoadImage` 노드에 업로드하세요： + + + + `LoadImage` node 13 · `hidream_e1_1_input_image.jpg` + + + +
+ hidream_e1_1_input_image.jpg +
### 1. HiDream E1.1 워크플로우 및 관련 자료 @@ -117,14 +132,46 @@ E1.1은 2025년 7월 16일에 출시된 업데이트 버전입니다. 이 버전 - HiDream E1.1은 총 100만 픽셀의 동적 입력을 지원하기 때문에, 워크플로우에서는 `Scale Image to Total Pixels`를 사용해 모든 입력 이미지를 처리하고 변환하며, 이로 인해 종횡비가 원본 입력 이미지와 다를 수 있습니다. - 모델의 fp16 버전을 사용할 경우, 실제 테스트에서 풀버전은 A100 40GB와 4090D 24GB 모두에서 메모리 부족 현상이 발생했으므로, 기본적으로 워크플로우는 `fp8_e4m3fn_fast`를 사용하도록 설정되었습니다. - ## HiDream E1 ComfyUI 네이티브 워크플로우 예시 +### HiDream E1 Image Edit (`hidream_e1_full`) + +Edit images with HiDream E1. Professional natural language image editing model. + +HiDream E1 image editing 워크플로 미리보기 + - Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "HiDream E1 Full" 검색 + + Comfy Cloud에서 열기 + + + Download JSON or search "HiDream E1 Full" in Template Library + +**입력 자료** + +이 파일을 일치하는 `LoadImage` 노드에 업로드하세요： + + + + `LoadImage` node 13 · `hidream_e1_full_input_image.jpg` + + + +
+ hidream_e1_full_input_image.jpg +
+ +**출력 예시** + +
+ 입력 이미지 + HiDream E1 출력 예시 +
+ +E1 is a model released on April 28, 2025. + E1은 2025년 4월 28일에 출시된 모델입니다. 이 모델은 768*768 해상도만 지원합니다. diff --git a/ko/tutorials/image/hidream/hidream-i1.mdx b/ko/tutorials/image/hidream/hidream-i1.mdx index 30484cb72..900729384 100644 --- a/ko/tutorials/image/hidream/hidream-i1.mdx +++ b/ko/tutorials/image/hidream/hidream-i1.mdx @@ -97,125 +97,155 @@ HiDream-ai는 서로 다른 요구를 충족하기 위해 HiDream-I1 모델의 │ └── ... # 해당 버전 워크플로우에서 설치하도록 안내하겠습니다 ``` -### HiDream-I1 풀버전 워크플로우 +### HiDream I1 Full (`hidream_i1_full`) + +Generate images with HiDream I1 Full. Complete version with 50 inference steps for highest quality output. + +HiDream I1 Full 워크플로 미리보기 - Run this workflow on Comfy Cloud with zero setup + Comfy Cloud에서 설정 없이 실행 + + + Download the workflow JSON file - 워크플로우 JSON 파일 다운로드 -#### 1. 모델 파일 다운로드 +**출력 예시** + +![HiDream I1 Full 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_full.png) -하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. +#### 1. Model File Download -- FP8 버전: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +Please select the appropriate version based on your hardware. Click the link and download the corresponding model file to save it to the `ComfyUI/models/diffusion_models/` folder. -#### 2. 워크플로우 파일 다운로드 +- FP8 version: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) requires more than 27GB of VRAM -아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. -![HiDream-I1 풀버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) +#### 2. Workflow File Download -#### 3. 워크플로우 단계별 완료 +Please download the image below and drag it into ComfyUI to load the corresponding workflow +![HiDream-I1 Full Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) -![HiDream-I1 풀버전 흐름 다이어그램](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) +#### 3. Complete the Workflow Step by Step -워크플로우를 단계별로 완료하세요: -1. `Load Diffusion Model` 노드가 `hidream_i1_full_fp8.safetensors` 파일을 사용하고 있는지 확인하세요. -2. `QuadrupleCLIPLoader`에서 네 개의 해당 텍스트 인코더가 올바르게 로드되었는지 확인하세요. - - clip_l_hidream.safetensors - - clip_g_hidream.safetensors - - t5xxl_fp8_e4m3fn_scaled.safetensors - - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` 노드가 `ae.safetensors` 파일을 사용하고 있는지 확인하세요. -4. **풀버전**의 경우, `ModelSamplingSD3`의 `shift` 파라미터를 `3.0`으로 설정해야 합니다. -5. `Ksampler` 노드에서 다음과 같은 설정을 해야 합니다: - - `steps`를 `50`로 설정하세요. - - `cfg`를 `5.0`으로 설정하세요. - - (선택) `sampler`를 `lcm`로 설정하세요. - - (선택) `scheduler`를 `normal`으로 설정하세요. -6. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 이미지 생성을 실행하세요. +![HiDream-I1 Full Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) -### HiDream-I1 Dev 버전 워크플로우 +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_full_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly + - clip_l_hidream.safetensors + - clip_g_hidream.safetensors + - t5xxl_fp8_e4m3fn_scaled.safetensors + - llama_3.1_8b_instruct_fp8_scaled.safetensors +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **full** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `50` + - Set `cfg` to `5.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation + +### HiDream I1 Dev (`hidream_i1_dev`) + +Generate images with HiDream I1 Dev. Balanced version with 28 inference steps, suitable for medium-range hardware. + +HiDream I1 Dev 워크플로 미리보기 - Run this workflow on Comfy Cloud with zero setup + Comfy Cloud에서 설정 없이 실행 + + + Download the workflow JSON file - 워크플로우 JSON 파일 다운로드 -#### 1. 모델 파일 다운로드 -하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. +**출력 예시** -- FP8 버전: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +![HiDream I1 Dev 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_dev.png) -#### 2. 워크플로우 파일 다운로드 -아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. +#### 1. Model File Download +Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -![HiDream-I1 Dev 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) +- FP8 version: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) requires more than 27GB of VRAM -#### 3. 워크플로우 단계별 완료 +#### 2. Workflow File Download +Please download the image below and drag it into ComfyUI to load the corresponding workflow -![HiDream-I1 Dev 버전 흐름 다이어그램](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) -워크플로우를 단계별로 완료하세요: -1. `Load Diffusion Model` 노드가 `hidream_i1_dev_fp8.safetensors` 파일을 사용하고 있는지 확인하세요. -2. `QuadrupleCLIPLoader`에서 네 개의 해당 텍스트 인코더가 올바르게 로드되었는지 확인하세요. - - clip_l_hidream.safetensors - - clip_g_hidream.safetensors - - t5xxl_fp8_e4m3fn_scaled.safetensors - - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` 노드가 `ae.safetensors` 파일을 사용하고 있는지 확인하세요. -4. **dev** 버전의 경우, `ModelSamplingSD3`의 `shift` 파라미터를 `6.0`으로 설정해야 합니다. -5. `Ksampler` 노드에서 다음과 같은 설정을 해야 합니다: - - `steps`를 `28`로 설정하세요. - - (중요) `cfg`를 `1.0`으로 설정하세요. - - (선택) `sampler`를 `lcm`로 설정하세요. - - (선택) `scheduler`를 `normal`으로 설정하세요. -6. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 이미지 생성을 실행하세요. +![HiDream-I1 Dev Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) -### HiDream-I1 Fast 버전 워크플로우 +#### 3. Complete the Workflow Step by Step + +![HiDream-I1 Dev Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_dev_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly + - clip_l_hidream.safetensors + - clip_g_hidream.safetensors + - t5xxl_fp8_e4m3fn_scaled.safetensors + - llama_3.1_8b_instruct_fp8_scaled.safetensors +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **dev** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `6.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `28` + - (Important) Set `cfg` to `1.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation + +### HiDream I1 Fast (`hidream_i1_fast`) + +Generate images quickly with HiDream I1 Fast. Lightweight version with 16 inference steps, ideal for rapid previews on lower-end hardware. + +HiDream I1 Fast 워크플로 미리보기 - Run this workflow on Comfy Cloud with zero setup + Comfy Cloud에서 설정 없이 실행 + + + Download the workflow JSON file - 워크플로우 JSON 파일 다운로드 -#### 1. 모델 파일 다운로드 -하드웨어에 맞는 적절한 버전을 선택해 링크를 클릭하고, 해당 모델 파일을 다운로드해 `ComfyUI/models/diffusion_models/` 폴더에 저장해 주세요. +**출력 예시** + +![HiDream I1 Fast 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_fast.png) + +#### 1. Model File Download +Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -- FP8 버전: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)는 16GB 이상의 VRAM을 요구합니다. -- 풀버전: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)는 27GB 이상의 VRAM을 요구합니다. +- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM -#### 2. 워크플로우 파일 다운로드 -아래 이미지를 다운로드해 ComfyUI로 드래그하여 해당 워크플로우를 로드해 주세요. +#### 2. Workflow File Download +Please download the image below and drag it into ComfyUI to load the corresponding workflow -![HiDream-I1 Fast 버전 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) +![HiDream-I1 Fast Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) -#### 3. 워크플로우 단계별 완료 +#### 3. Complete the Workflow Step by Step -![HiDream-I1 Fast 버전 흐름 다이어그램](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) +![HiDream-I1 Fast Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) -워크플로우를 단계별로 완료하세요: -1. `Load Diffusion Model` 노드가 `hidream_i1_fast_fp8.safetensors` 파일을 사용하고 있는지 확인하세요. -2. `QuadrupleCLIPLoader`에서 네 개의 해당 텍스트 인코더가 올바르게 로드되었는지 확인하세요. - - clip_l_hidream.safetensors - - clip_g_hidream.safetensors - - t5xxl_fp8_e4m3fn_scaled.safetensors - - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` 노드가 `ae.safetensors` 파일을 사용하고 있는지 확인하세요. -4. **fast** 버전의 경우, `ModelSamplingSD3`의 `shift` 파라미터를 `3.0`으로 설정해야 합니다. -5. `Ksampler` 노드에서 다음과 같은 설정을 해야 합니다: - - `steps`를 `16`로 설정하세요. - - (중요) `cfg`를 `1.0`으로 설정하세요. - - (선택) `sampler`를 `lcm`로 설정하세요. - - (선택) `scheduler`를 `normal`으로 설정하세요. -6. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 이미지 생성을 실행하세요. +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_fast_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly + - clip_l_hidream.safetensors + - clip_g_hidream.safetensors + - t5xxl_fp8_e4m3fn_scaled.safetensors + - llama_3.1_8b_instruct_fp8_scaled.safetensors +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **fast** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `16` + - (Important) Set `cfg` to `1.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation ## 기타 관련 자료 diff --git a/ko/tutorials/image/hidream/hidream-o1.mdx b/ko/tutorials/image/hidream/hidream-o1.mdx index 4741dcbbe..1ecdeb824 100644 --- a/ko/tutorials/image/hidream/hidream-o1.mdx +++ b/ko/tutorials/image/hidream/hidream-o1.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Additional Notes": 762758eb --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -33,49 +32,24 @@ HiDream-O1-Image는 [MIT 라이선스](https://github.com/HiDream-ai/HiDream-O1- ## HiDream-O1-Image 풀 워크플로우 -### 1. 워크플로우 다운로드 +### HiDream O1 Full: Image generation (`image_hidream_o1`) -ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플릿 탐색` -> `이미지`로 이동해 "HiDream O1 풀: 이미지 생성"을 찾으세요. +Input a text prompt and optionally upload reference images. Generate a high-resolution image up to 2048x2048 with text-to-image, editing, or subject-driven personalization. -![HiDream-O1-Image 풀 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) +HiDream O1 Full 워크플로 미리보기 - - 워크플로우 다운로드 + + + Comfy Cloud에서 열기 - - 클라우드에서 열기 + + Download JSON or search "HiDream O1 Full" in Template Library + -### 2. 모델 다운로드 - -**체크포인트** — 재패키징되고 양자화되었습니다. 모든 모델은 최악의 이상치에 대해 bf16을 사용하며, 사용되지 않는 딥스택 레이어는 제거되었습니다: - -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 양자화된 변형 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) - -**텍스트 인코더**(프롬프트 강화) — 모든 버전에서 공유됩니다: +**출력 예시** -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) - -**LoRA(선택사항)** — Dev 디스틸레이션은 풀 모델에도 LoRA로 적용할 수 있으며, 디스틸레이션 강도를 조정할 수 있습니다([Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 제공): - -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 전체 랭크 -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 프룬드 변형 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 대안적인 체크포인트 기반 디스틸레이션 - -``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 checkpoints/ -│ │ ├── hidream_o1_image_fp8_scaled.safetensors -│ │ ├── hidream_o1_image_mxfp8.safetensors -│ │ └── hidream_o1_image_bf16.safetensors -│ ├── 📂 loras/ -│ │ └── hidream_o1_dev_lora_rank_64_bf16.safetensors -│ └── 📂 text_encoders/ -│ └── gemma4_e4b_it_fp8_scaled.safetensors -``` +![HiDream O1 Full 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) ### 3. 워크플로우 사용하기 @@ -88,41 +62,41 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 ## HiDream-O1-Image Dev 워크플로우 -### 1. 워크플로우 다운로드 +### HiDream O1 Dev (`image_hidream_o1_dev`) -`워크플로우` -> `템플릿 탐색` -> `이미지`로 이동해 "HiDream O1 Dev"를 찾으세요. +Input a text prompt and optional reference images. Generate a high-resolution image (up to 2048x2048) with support for text-to-image, image editing, and subject-driven personalization. -![HiDream-O1-Image Dev 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1_dev.png) +HiDream O1 Dev 워크플로 미리보기 - - 워크플로우 다운로드 + + + Comfy Cloud에서 열기 - - 클라우드에서 열기 + + Download JSON or search "HiDream O1 Dev" in Template Library + -### 2. 모델 다운로드 +**입력 자료** -**체크포인트(Dev)** — 재패키징되고 양자화되었습니다. 모든 모델은 최악의 이상치에 대해 bf16을 사용하며, 사용되지 않는 딥스택 레이어는 제거되었습니다: +Upload this file to the matching `LoadImage` node: -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, 지원되는 하드웨어에서 속도 향상을 위해 안전한 MLP 레이어에서 fp8/mxfp8 매트릭스 곱셈을 사용합니다. -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 양자화된 변형 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — 전체 bf16 정밀도(가장 큰) + + + `LoadImage` node 213 · `noir_portrait.png` + + -**텍스트 인코더**(프롬프트 강화) — 모든 버전에서 공유됩니다: +
+ noir_portrait.png +
-- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +**출력 예시** -``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 checkpoints/ -│ │ ├── hidream_o1_image_dev_fp8_scaled.safetensors -│ │ ├── hidream_o1_image_dev_mxfp8.safetensors -│ │ └── hidream_o1_image_dev_bf16.safetensors -│ └── 📂 text_encoders/ -│ └── gemma4_e4b_it_fp8_scaled.safetensors -``` +
+ 입력 이미지 + HiDream O1 Dev 출력 예시 +
### 3. 워크플로우 사용하기 diff --git a/ko/tutorials/image/ideogram/ideogram-v4.mdx b/ko/tutorials/image/ideogram/ideogram-v4.mdx index 9ff8cc452..871d08039 100644 --- a/ko/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ko/tutorials/image/ideogram/ideogram-v4.mdx @@ -10,25 +10,30 @@ translationBlockHashes: "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; Ideogram 4.0은 Ideogram에서 출시한 최신 텍스트 기반 이미지 생성 모델로, 완전히 자체 하드웨어에서 실행되는 오픈소스 모델입니다. 뛰어난 포토리얼리즘 품질, 정확한 텍스트 렌더링, 정밀한 스타일 제어를 제공합니다. 일반 자연어 또는 **구조화된 JSON 프롬프트**를 사용해 레이아웃, 색상, 이미지 내 텍스트까지 세밀하게 제어할 수 있습니다. -## Ideogram 4.0 텍스트 기반 이미지 생성 워크플로 +### Ideogram v4: Text to Image (`image_ideogram4_t2i`) + +Input a text prompt or structured JSON description. Generate an image with precise layout, color, and style control using Ideogram 4.0. - +Ideogram 4.0 text-to-image 워크플로 미리보기 + + + Comfy Cloud에서 열기 - - - JSON 다운로드하거나 템플릿 라이브러리에서 "Ideogram v4: Text to Image" 검색하기 + + Download JSON or search "Ideogram v4: Text to Image" in Template Library + + +**출력 예시** -![Ideogram 4.0 예제 출력](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) -*Ideogram 4.0 모델의 예제 출력* +![Ideogram 4.0 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) ### 프롬프트 형식 @@ -46,25 +51,23 @@ Ideogram 4.0은 Ideogram에서 출시한 최신 텍스트 기반 이미지 생 Hugging Face의 [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4)에서 모든 재포장된 모델 파일을 확인할 수 있습니다. - + + Ideogram 4.0용 디퓨전 모델 (~13.8 GB). models/diffusion_models/에 저장하세요 - - + Ideogram 4.0용 비조건 디퓨전 모델 (~13.8 GB). models/diffusion_models/에 저장하세요 - - + Ideogram 4.0용 텍스트 인코더 (~8 GB). models/text_encoders/에 저장하세요 - - + Ideogram 4.0용 텍스트 인코더 (~2 GB). models/text_encoders/에 저장하세요 - - + Ideogram 4.0용 VAE (~335 MB). models/vae/에 저장하세요 + **모델 저장 위치** diff --git a/ko/tutorials/image/krea/krea-2.mdx b/ko/tutorials/image/krea/krea-2.mdx index 0bd78e24e..44b0127e2 100644 --- a/ko/tutorials/image/krea/krea-2.mdx +++ b/ko/tutorials/image/krea/krea-2.mdx @@ -12,9 +12,6 @@ translationBlockHashes: "Krea-2 Turbo style reference workflow": 5e966a16 --- - - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' @@ -46,30 +43,32 @@ Krea 2는 두 가지 변형으로 제공됩니다: - [공식 GitHub 저장소](https://github.com/krea-ai/krea-2) - [기술 보고서](https://www.krea.ai/blog/krea-2-technical-report) -## Krea-2 Turbo 텍스트 기반 이미지 생성 워크플로 +### Krea-2: Text to Image (`image_krea2_turbo_t2i`) + +Generate images from text prompts using Krea 2, a foundation model built for aesthetic quality and creative control. It focuses on rendering expressive, stylistically diverse images. -Krea-2 Turbo 텍스트 기반 이미지 생성 워크플로 +Krea-2 Turbo text-to-image 워크플로 미리보기 Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "Krea-2" 검색 + Download JSON or search "Krea-2" in Template Library -워크플로는 다음과 같은 부분으로 구성됩니다: +The workflow is organized into a few parts: -1. **Text to Image (Krea-2 Turbo) 서브그래프**: 모델 로딩, 프롬프트 처리, 샘플링, VAE 디코드를 포함한 핵심 생성 파이프라인 -2. **ResolutionSelector**: 원하는 출력 해상도를 선택하세요. Krea 2는 1K에서 2K까지 출력을 지원합니다. 메가픽셀 값을 2.0으로 설정하면 2K 해상도를 얻을 수 있습니다. -3. **CustomCombo (LoRA 선택기)**: 사용 가능한 스타일 LoRA에 대해 사전 구성된 트리거 단어 선택기입니다. 추가 LoRA를 다운로드한 경우 이 선택기를 사용자 정의하고 해당 LoRA 파일과 페어링할 수 있습니다. -4. **SaveImage**: 생성된 이미지 저장 +1. **Text to Image (Krea-2 Turbo) subgraph**: the core generation pipeline, containing model loading, prompt handling, sampling, and VAE decode +2. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K — set the megapixels value to 2.0 to get 2K resolution. +3. **CustomCombo (LoRA selector)**: a pre-built trigger word selector for the available style LoRAs. If you download additional LoRAs, you can customize this selector and pair them with the corresponding LoRA files accordingly. +4. **SaveImage**: saves the generated image - - 이 워크플로는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인하여 워크플로를 사용자 정의하고 확장하는 방법을 알아보세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ### 원클릭 생성 @@ -145,27 +144,47 @@ Krea는 또한 Krea 2용 스타일 LoRA 컬렉션을 출시했습니다. **Custo │ └── krea2_softwatercolor.safetensors (and other style LoRAs) ``` -## Krea-2 Turbo 스타일 참조 워크플로 +### Krea-2 Int8: Image Style Reference (`image_krea2_turbo_int8_image_style_reference`) -Krea-2 Turbo 스타일 참조 워크플로 +Generate images with the Krea-2 Turbo model while referencing the style of 1–2 uploaded images, using the high-performance Int8 Convrot format for fast inference. + +Krea-2 Turbo style reference 워크플로 미리보기 Comfy Cloud에서 열기 - JSON을 다운로드하거나 템플릿 라이브러리에서 "Krea-2 Style Reference"를 검색하세요. + Download JSON or search "Krea-2 Style Reference" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 69 · `krea2_reference_image.png` -스타일 참조 워크플로는 Krea-2 Turbo 파이프라인에 참조 이미지 조건화를 추가하여 구축됩니다. 하나 이상의 참조 이미지를 업로드하여 생성된 출력물의 미적 스타일, 분위기 및 시각적 방향에 영향을 줍니다. +
+ krea2_reference_image.png +
+ +**출력 예시** + +![Krea-2 style reference 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_krea2_turbo_int8_image_style_reference.png) + +The style reference workflow builds on the Krea-2 Turbo pipeline by adding reference image conditioning. Upload one or more reference images to influence the aesthetic style, mood, and visual direction of the generated output. -워크플로는 몇 가지 부분으로 구성됩니다: +The workflow is organized into a few parts: -1. **이미지 스타일 참조 (Krea-2 Turbo) 서브그래프**: 스타일 참조 지원이 포함된 핵심 생성 파이프라인으로, 모델 로딩, 참조 이미지 조건화, 프롬프트 처리 및 샘플링을 포함합니다. -2. **LoadImage**: 스타일 참조 이미지를 업로드합니다. -3. **ResolutionSelector**: 원하는 출력 해상도를 선택합니다. Krea 2는 1K에서 2K까지의 출력을 지원합니다. -4. **SaveImage**: 생성된 이미지를 저장합니다. +1. **Image Style Reference (Krea-2 Turbo) subgraph**: the core generation pipeline with style reference support, containing model loading, reference image conditioning, prompt handling, and sampling +2. **LoadImage**: upload your style reference images +3. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K. +4. **SaveImage**: saves the generated image ### 스타일 참조 사용 diff --git a/ko/tutorials/image/lens/lens.mdx b/ko/tutorials/image/lens/lens.mdx index 8ceae12be..fe9acec99 100644 --- a/ko/tutorials/image/lens/lens.mdx +++ b/ko/tutorials/image/lens/lens.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Available models": 5876b860 --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Lens**는 **마이크로소프트**가 개발한 오픈 소스 텍스트 기반 이미지 생성 모델로, MIT 라이선스를 적용했습니다. **38억 개**의 파라미터를 가진 이 모델은 **듀얼 스트림 MMDiT** 아키텍처를 사용하며, **GPT-OSS-20B** 텍스트 인코더 특징과 **FLUX.2 세맨틱 VAE**를 통해 더 큰 T2I 모델들에 비해 훨씬 적은 학습 계산량으로 경쟁력 있는 이미지 품질을 제공합니다. @@ -37,57 +36,63 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 이 워크플로우는 모듈식 처리를 위해 Subgraph 노드를 사용합니다. Subgraph 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 배워보세요. -### Lens +### Lens: Text to Image (`image_lens_t2i`) + +Input a text prompt and select resolution and aspect ratio. Generate a high-quality image using the efficient Lens text-to-image model. + +Lens text-to-image 워크플로 미리보기 - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Lens" 검색하세요 + + Download JSON or search "Lens" in Template Library - {/* TODO: Lens가 Comfy Cloud에서 이용 가능해지면 클라우드 템플릿 활성화 */} - {/**/} + {/* TODO: Enable Cloud template when Lens is available on Comfy Cloud */} + {/**/} {/* Comfy Cloud에서 열기*/} {/**/} - +**출력 예시** + +![Lens text-to-image 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_t2i.png) -#### 시작하기 + -1. ComfyUI를 최신 버전으로 업데이트하세요 - {/* TODO: 템플릿 이용 가능 시 클라우드 옵션 추가 */} -2. **템플릿**으로 이동해 "Lens"를 검색하세요 -3. **Lens** 워크플로우를 선택하세요 -4. 누락된 모델을 다운로드하고([모델 다운로드 참조](#lens-model-downloads)), 프롬프트를 입력한 후 **큐**를 클릭하세요 +#### Get started -#### 예시 출력 +1. Update ComfyUI to the latest version + {/* TODO: Add Cloud option when template is available */} +2. Go to **Template** and search for **Lens** +3. Select the **Lens** workflow +4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** -Lens 텍스트 기반 이미지 생성 예시 출력 +### Lens Turbo: Text to Image (`image_lens_turbo_t2i`) -### Lens Turbo +Input a text prompt and select resolution, aspect ratio, and inference steps. Generate a high-quality image using the Lens text-to-image model. -Lens Turbo는 추출된 변형으로, 더 적은 샘플링 단계로 이미지를 생성해 더 빠른 추론을 가능하게 합니다. +Lens Turbo text-to-image 워크플로 미리보기 - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Lens Turbo" 검색하세요 + + Download JSON or search "Lens Turbo" in Template Library - {/* TODO: Lens Turbo가 Comfy Cloud에서 이용 가능해지면 클라우드 템플릿 활성화 */} - {/**/} + {/* TODO: Enable Cloud template when Lens Turbo is available on Comfy Cloud */} + {/**/} {/* Comfy Cloud에서 열기*/} {/**/} -#### 시작하기 +**출력 예시** -1. ComfyUI를 최신 버전으로 업데이트하세요 - {/* TODO: 템플릿 이용 가능 시 클라우드 옵션 추가 */} -2. **템플릿**으로 이동해 "Lens Turbo"를 검색하세요 -3. **Lens Turbo** 워크플로우를 선택하세요 -4. 누락된 모델을 다운로드하고([모델 다운로드 참조](#lens-model-downloads)), 프롬프트를 입력한 후 **큐**를 클릭하세요 +![Lens Turbo text-to-image 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_turbo_t2i.png) -#### 예시 출력 +#### Get started -Lens Turbo 텍스트 기반 이미지 생성 예시 출력 +1. Update ComfyUI to the latest version + {/* TODO: Add Cloud option when template is available */} +2. Go to **Template** and search for **Lens Turbo** +3. Select the **Lens Turbo** workflow +4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** ## Lens 모델 다운로드 diff --git a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index d42173bc1..c57012e07 100644 --- a/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ko/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Prompt format": 5b819c6c --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **NewBie-image-Exp0.1**은 NewBieAI Lab에서 개발한 3.5B 파라미터의 DiT 모델로, 애니메이션 스타일 텍스트 기반 이미지 생성을 위해 설계되었습니다. Next-DiT 아키텍처를 기반으로 하며, 뛰어난 디테일과 시각적으로 강렬한 애니메이션 이미지를 제공합니다. @@ -29,13 +27,24 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - [Hugging Face](https://huggingface.co/NewBie-AI/NewBie-image-Exp0.1) - [시작하기 가이드](https://ai.feishu.cn/wiki/NZl9wm7V1iuNzmkRKCUcb1USnsh) -## NewBie-image 텍스트 기반 이미지 생성 워크플로 +### NewBie Exp0.1: Anime Generation (`image_newbieimage_exp0_1-t2i`) + +Generate detailed anime-style images with NewBie Exp0.1's Next-DiT architecture. Supports XML structured prompts for better multi-character scenes and attribute binding. + +NewBie-image text-to-image 워크플로 미리보기 - JSON 다운로드 또는 템플릿 라이브러리에서 "NewBie-image" 검색 - 클라우드에서 열기 + + Comfy Cloud에서 열기 + + + Download JSON or search "NewBie-image" in Template Library + +**출력 예시** + +![NewBie-image 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_newbieimage_exp0_1-t2i.png) diff --git a/ko/tutorials/image/omnigen/omnigen2.mdx b/ko/tutorials/image/omnigen/omnigen2.mdx index 1c5473e0b..dfbd9e074 100644 --- a/ko/tutorials/image/omnigen/omnigen2.mdx +++ b/ko/tutorials/image/omnigen/omnigen2.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "ComfyUI OmniGen2 Image Editing Workflow": 1e06072e --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## OmniGen2 소개 @@ -65,15 +63,24 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 ## ComfyUI OmniGen2 텍스트 기반 이미지 생성 워크플로우 -### 1. 워크플로우 파일 다운로드 +### OmniGen2: Text to Image (`image_omnigen2_t2i`) + +Generate high-quality images from text prompts using OmniGen2's unified 7B multimodal model with dual-path architecture. + +OmniGen2 text-to-image 워크플로 미리보기 - + - Open and run this workflow directly in Comfy Cloud. + Comfy Cloud에서 직접 열어 실행 + + + Download JSON or search "OmniGen2" in Template Library -![텍스트 기반 이미지 생성 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) +**출력 예시** + +![OmniGen2 text-to-image 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_omnigen2_t2i.png) ### 2. 워크플로우 단계별 완료 @@ -95,18 +102,41 @@ OmniGen2는 약 **7B**의 총 파라미터(텍스트 모델 3B + 이미지 생 OmniGen2는 풍부한 이미지 편집 기능을 갖추고 있으며, 이미지에 텍스트 추가도 지원합니다. -### 1. 워크플로우 파일 다운로드 +### OmniGen2 Image Edit (`image_omnigen2_image_edit`) + +Edit images with natural language instructions using OmniGen2's advanced image editing capabilities and text rendering support. + +OmniGen2 image edit 워크플로 미리보기 - + - Open and run this workflow directly in Comfy Cloud. + Comfy Cloud에서 직접 열어 실행 + + + Download JSON or search "OmniGen2 Image Edit" in Template Library -![이미지 편집 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 16 · `image_omnigen2_image_edit_input_image.png` + + + +
+ image_omnigen2_image_edit_input_image.png +
+ +**출력 예시** -아래 이미지를 다운로드하세요. 이 이미지를 입력 이미지로 사용하겠습니다. -![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/input_fairy.png) +
+ 입력 이미지 + OmniGen2 image edit 출력 예시 +
### 2. 워크플로우 단계별 완료 diff --git a/ko/tutorials/image/ovis/ovis-image.mdx b/ko/tutorials/image/ovis/ovis-image.mdx index dc0ede6d9..752dd21a3 100644 --- a/ko/tutorials/image/ovis/ovis-image.mdx +++ b/ko/tutorials/image/ovis/ovis-image.mdx @@ -20,14 +20,21 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - [GitHub](https://github.com/AIDC-AI/Ovis-Image) - [Hugging Face](https://huggingface.co/AIDC-AI/Ovis-Image-7B) -## Ovis-Image 텍스트 기반 이미지 생성 워크플로우 +### Ovis-Image Text to Image (`image_ovis_text_to_image`) + +Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering in generated images. Designed to operate efficiently under stringent computational constraints. + +Ovis-Image text-to-image 워크플로 미리보기 - Comfy Cloud에서 열기 - JSON 다운로드 또는 템플릿 라이브러리에서 "Ovis image" 검색 + + Comfy Cloud에서 열기 + + + Download JSON or search "Ovis image" in Template Library + - ## 모델 링크 diff --git a/ko/tutorials/image/pixeldit/pixeldit.mdx b/ko/tutorials/image/pixeldit/pixeldit.mdx index 0a371da8f..baa09a71d 100644 --- a/ko/tutorials/image/pixeldit/pixeldit.mdx +++ b/ko/tutorials/image/pixeldit/pixeldit.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Model downloads": e4bafb0a --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" **PixelDiT**는 1024px 텍스트 기반 이미지 생성을 위한 NVIDIA의 픽셀 공간 확산 변환기입니다. 잠재 공간에서 작동하는 전통적인 확산 모델과 달리 PixelDiT는 패치 수준 DiT와 픽셀 수준 DiT를 결합한 이중 레벨 DiT 아키텍처를 사용해 직접 픽셀 공간에서 이미지를 생성하며, MM-DiT 융합을 통해 텍스트와 이미지 토큰 간의 공동 주의를 구현합니다. diff --git a/ko/tutorials/image/qwen/qwen-image-2512.mdx b/ko/tutorials/image/qwen/qwen-image-2512.mdx index ffca6338b..ba39e2f43 100644 --- a/ko/tutorials/image/qwen/qwen-image-2512.mdx +++ b/ko/tutorials/image/qwen/qwen-image-2512.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Qwen-Image-2512 ComfyUI Native Workflow Example": 50fbe75c --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image-2512**는 Qwen-Image의 텍스트 기반 이미지 생성 모델의 12월 업데이트입니다. 지난 8월에 출시된 기본 Qwen-Image 모델과 비교해 Qwen-Image-2512는 이미지 품질과 사실성이 크게 향상되었습니다. diff --git a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx index 08acf49bd..045586a9e 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -40,7 +40,6 @@ ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거 - ### 2. 모델 다운로드 **텍스트 인코더** diff --git a/ko/tutorials/image/qwen/qwen-image-edit.mdx b/ko/tutorials/image/qwen/qwen-image-edit.mdx index 96e5e01b7..161f36023 100644 --- a/ko/tutorials/image/qwen/qwen-image-edit.mdx +++ b/ko/tutorials/image/qwen/qwen-image-edit.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Qwen-Image-Edit ComfyUI Native Workflow Example": 6703460f --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image-Edit**는 Qwen-Image의 이미지 편집 버전입니다. 20B Qwen-Image 모델을 기반으로 추가로 학습되었으며, Qwen-Image만의 독특한 텍스트 렌더링 기능을 편집 작업에 성공적으로 확장해 정밀한 텍스트 편집이 가능합니다. 또한 Qwen-Image-Edit는 입력 이미지를 Qwen2.5-VL(시각적 세미틱 제어용)과 VAE 인코더(시각적 외관 제어용)에 동시에 입력하여 이중 세미틱 및 외관 편집 기능을 구현합니다. @@ -43,21 +41,41 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' -### 1. 워크플로우 파일 +### Qwen Image Edit (`image_qwen_image_edit`) + +Edit images with precise bilingual text editing and dual semantic/appearance editing capabilities using Qwen-Image-Edit's 20B MMDiT model. -ComfyUI를 업데이트한 후, 템플릿에서 워크플로우 파일을 찾거나 아래 워크플로우를 ComfyUI로 드래그해 불러올 수 있습니다. -![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) +Qwen-Image-Edit 워크플로 미리보기 - JSON 다운로드 또는 템플릿 라이브러리에서 "image_qwen_image_edit" 검색 - Run this workflow on Cloud GPUs with zero setup + Cloud GPU에서 설정 없이 실행 + + + Download JSON or search "Qwen Image Edit" in Template Library +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 78 · `image_qwen_image_edit_input_image.png` + + + +
+ image_qwen_image_edit_input_image.png +
+ +**출력 예시** -아래 이미지를 입력으로 다운로드하세요 -![Qwen-image 텍스트-이미지 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) +
+ 입력 이미지 + Qwen-Image-Edit 출력 예시 +
### 2. 모델 다운로드 diff --git a/ko/tutorials/image/qwen/qwen-image-layered.mdx b/ko/tutorials/image/qwen/qwen-image-layered.mdx index 513808568..e5c2c81ad 100644 --- a/ko/tutorials/image/qwen/qwen-image-layered.mdx +++ b/ko/tutorials/image/qwen/qwen-image-layered.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Workflow settings": 098636f1 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image-Layered**는 알리바바의 Qwen팀에서 개발한 모델로, 이미지를 여러 개의 RGBA 레이어로 분해할 수 있습니다. 이 계층적 표현은 각 레이어가 독립적으로 조작 가능하도록 하여 다른 콘텐츠에 영향을 주지 않고도 각 레이어를 개별적으로 수정할 수 있게 합니다. @@ -41,7 +39,6 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' | | - ## 모델 링크 diff --git a/ko/tutorials/image/qwen/qwen-image.mdx b/ko/tutorials/image/qwen/qwen-image.mdx index b5a2b3b9b..de5d8e693 100644 --- a/ko/tutorials/image/qwen/qwen-image.mdx +++ b/ko/tutorials/image/qwen/qwen-image.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Qwen Image Union ControlNet LoRA Workflow": a08d8e37 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Qwen-Image**는 알리바바의 Qwen 팀이 발표한 최초의 이미지 생성 기반 모델입니다. 이 모델은 Apache 2.0 라이선스로 오픈소스화된 20B 파라미터 MMDiT(멀티모달 디퓨전 트랜스포머) 모델입니다. 이 모델은 **복잡한 텍스트 렌더링**과 **정밀한 이미지 편집**에서 큰 진전을 이루었으며, 영어와 중국어를 포함한 여러 언어에 대해 고화질 출력을 달성했습니다. @@ -54,32 +52,38 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' allowFullScreen > +### Qwen-Image: Text to Image (`image_qwen_image`) -## Qwen-Image 네이티브 워크플로우 예시 +Generate images with exceptional multilingual text rendering and editing capabilities using Qwen-Image's 20B MMDiT model. - +Qwen-Image text-to-image 워크플로 미리보기 + Comfy Cloud에서 열기 - + + Download JSON or search "Qwen-Image" in Template Library -이 문서에 첨부된 워크플로우에는 세 가지 다른 모델이 사용됩니다: -1. Qwen-Image 원본 모델 fp8_e4m3fn -2. 8단계 가속 버전: Qwen-Image 원본 모델 fp8_e4m3fn에 lightx2v 8단계 LoRA 적용 -3. 증류 버전: Qwen-Image 증류 모델 fp8_e4m3fn +**출력 예시** -**VRAM 사용량 참고** -GPU: RTX4090D 24GB +![Qwen-Image 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_qwen_image.png) + +There are three different models used in the workflow attached to this document: +1. Qwen-Image original model fp8_e4m3fn +2. 8-step accelerated version: Qwen-Image original model fp8_e4m3fn with lightx2v 8-step LoRA +3. Distilled version: Qwen-Image distilled model fp8_e4m3fn -| 모델 사용 | VRAM 사용량 | 1세대 | 2세대 | -| ---------- | ----------- | ----- | ----- | -| fp8_e4m3fn | 86% | ≈ 94s | ≈ 71s | -| fp8_e4m3fn에 lightx2v 8단계 LoRA 적용 | 86% | ≈ 55s | ≈ 34s | -| 증류된 fp8_e4m3fn | 86% | ≈ 69s | ≈ 36s | +**VRAM Usage Reference** +GPU: RTX4090D 24GB +| Model Used | VRAM Usage | First Generation | Second Generation | +| --------------------------------------- | ---------- | --------------- | ---------------- | +| fp8_e4m3fn | 86% | ≈ 94s | ≈ 71s | +| fp8_e4m3fn with lightx2v 8-step LoRA | 86% | ≈ 55s | ≈ 34s | +| Distilled fp8_e4m3fn | 86% | ≈ 69s | ≈ 36s | ### 1. 워크플로우 파일 @@ -158,18 +162,36 @@ Qwen_image_distill 증류 모델과 lightx2v의 8단계 가속 LoRA는 동시에 사용되지 않는 것으로 보입니다. 서로 다른 조합을 실험해 함께 사용할 수 있는지 확인해보세요. -## Qwen Image InstantX ControlNet 워크플로우 +### Qwen-Image InstantX Union ControlNet (`image_qwen_image_instantx_controlnet`) + +Generate images with Qwen-Image InstantX ControlNet, supporting canny, soft edge, depth, and pose. -이것은 ControlNet 모델이므로 일반 ControlNet처럼 사용할 수 있습니다. +Qwen-Image InstantX ControlNet 워크플로 미리보기 + Comfy Cloud에서 열기 - + + Download JSON or search "Qwen-Image InstantX ControlNet" in Template Library -### 1. 워크플로우 및 입력 이미지 +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 71 · `image_qwen_image_instantx_controlnet_input_image.jpg` + + + +
+ image_qwen_image_instantx_controlnet_input_image.jpg +
+ +**출# 1. 워크플로우 및 입력 이미지 아래 이미지를 다운로드해 ComfyUI로 드래그해 워크플로우를 불러오세요. ![워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) @@ -215,20 +237,34 @@ ComfyUI/ 3. 이 서브그래프는 Lotus Depth 모델을 사용합니다. 템플릿에서 찾거나 서브그래프를 편집해 더 자세히 알아볼 수 있으며, 모든 모델이 정확히 로드되었는지 확인하세요. 4. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. -## Qwen Image ControlNet DiffSynth-ControlNets 모델 패치 워크플로우 +### Qwen-Image ControlNet Model Patch (`image_qwen_image_controlnet_patch`) + +Control image generation using Qwen-Image ControlNet models. Supports canny, depth, and inpainting controls through model patching. + +Qwen-Image ControlNet model patch 워크플로 미리보기 + Comfy Cloud에서 열기 - + + Download JSON or search "Qwen-Image ControlNet Patch" in Template Library -이 모델은 실제로 ControlNet이 아니라, 캐니, 딥스, 인페인트 등 세 가지 다른 제어 모드를 지원하는 모델 패치입니다. +**입력 자료** -원본 모델 주소: [DiffSynth-Studio/Qwen-Image ControlNet](https://www.modelscope.cn/collections/Qwen-Image-ControlNet-6157b44e89d444) -Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/tree/main/split_files/model_patches) +Upload this file to the matching `LoadImage` node: + + + `LoadImage` node 71 · `image_qwen_image_controlnet_patch_input_image.png` + + + +
+ image_qwen_image_controlnet_patch_input_image.png +
### 1. 워크플로우 및 입력 이미지 @@ -247,7 +283,6 @@ Comfy Org 재호스팅 주소: [Qwen-Image-DiffSynth-ControlNets/model_patches]( - [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) - [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) - ### 3. 워크플로우 사용 지침 현재 diffsynth는 캐니, 딥스, 인페인트 세 가지 패치 모델을 제공합니다. @@ -286,17 +321,44 @@ ControlNet 관련 워크플로우를 처음 사용한다면, 제어 이미지는 5. 필요하다면 `QwenImageDiffsynthControlnet` 노드의 `strength`를 수정해 해당 제어 강도를 조정할 수 있습니다. 6. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 워크플로우를 실행하세요. -## Qwen Image Union ControlNet LoRA 워크플로우 +### Qwen-Image Union Control (`image_qwen_image_union_control_lora`) + +Generate images with precise structural control using Qwen-Image's unified ControlNet LoRA. Supports multiple control types including canny, depth, lineart, softedge, normal, and openpose. + +Qwen-Image Union Control 워크플로 미리보기 + Comfy Cloud에서 열기 + + + Download JSON or search "Qwen-Image Union Control" in Template Library - + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 73 · `image_qwen_image_union_control_lora_input_image.png` -원본 모델 주소: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) -Comfy Org 재호스팅 주소: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 캐니, 딥스, 포즈, 라인아트, 소프트엣지, 노말, 오픈포즈 지원 이미지 구조 제어 LoRA +
+ image_qwen_image_union_control_lora_input_image.png +
+ +**출력 예시** + +
+ 입력 이미지 + Qwen-Image Union Control 출력 예시 +
+ +Original model address: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) +Comfy Org rehost address: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): Image structure control LoRA supporting canny, depth, pose, lineart, softedge, normal, openpose ### 1. 워크플로우 및 입력 이미지 diff --git a/ko/tutorials/image/z-image/z-image-turbo.mdx b/ko/tutorials/image/z-image/z-image-turbo.mdx index 05e9548bd..238ff7a31 100644 --- a/ko/tutorials/image/z-image/z-image-turbo.mdx +++ b/ko/tutorials/image/z-image/z-image-turbo.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Z-Image-Turbo Fun Union ControlNet workflow": 34191a15 --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -33,17 +31,24 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" - [GitHub](https://github.com/Tongyi-MAI/Z-Image) - [Hugging Face](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo) -## Z-Image-Turbo 텍스트-이미지 변환 워크플로우 +### Z-Image-Turbo: Text to Image (`image_z_image_turbo`) + +An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer, supports English & Chinese. + +Z-Image-Turbo text-to-image 워크플로 미리보기 - - Z-Image-Turbo 텍스트-이미지 변환 워크플로우 JSON 파일을 다운로드하세요. + + Comfy Cloud에서 이 워크플로를 직접 실행 - - 이 워크플로우를 ComfyUI Cloud에서 바로 실행하세요. + + Download the Z-Image-Turbo text-to-image workflow JSON file - + +**출력 예시** + +![Z-Image-Turbo 출력 예시](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_z_image_turbo.png) ### Z-Image-Turbo 모델 다운로드 @@ -71,28 +76,34 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" │ └── ae.safetensors ``` -## Z-Image-Turbo Fun Union ControlNet 워크플로우 +## Z-Image-Turbo Fun Union ControlNet 워크플로 -이 워크플로우는 Z-Image-Turbo Fun Union ControlNet 모델을 사용하여 ControlNet 가이드를 기반으로 이미지를 생성합니다. 참조 이미지에 Canny 에지 감지를 적용하고 ControlNet을 통해 생성 과정을 제어합니다. +### Z-Image-Turbo Fun Union ControlNet (`image_z_image_turbo_fun_union_controlnet`) - - Z-Image-Turbo Fun Union ControlNet 워크플로우 JSON 파일을 다운로드하세요. +Multi-control ControlNet supporting Canny, HED, Depth, Pose, and MLSD for Z-Image-Turbo. + +Z-Image-Turbo Fun Union ControlNet 워크플로 미리보기 + + + + Comfy Cloud에서 이 워크플로를 직접 실행 -### ControlNet용 추가 모델 + + Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file + + + +**입력 자료** - - Z-Image-Turbo용 ControlNet 모델 패치입니다. +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 58 · `image_z_image_turbo_fun_union_controlnet_input_image.png` -**모델 저장 위치** + + +
+ image_z_image_turbo_fun_union_controlnet_input_image.png +
-``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 text_encoders/ -│ │ └── qwen_3_4b.safetensors -│ ├── 📂 diffusion_models/ -│ │ └── z_image_turbo_bf16.safetensors -│ ├── 📂 vae/ -│ │ └── ae.safetensors -│ └── 📂 model_patches/ -│ └── Z-Image-Turbo-Fun-Controlnet-Union.safetensors diff --git a/ko/tutorials/llm/gemma4/gemma4.mdx b/ko/tutorials/llm/gemma4/gemma4.mdx index 81eff1335..33b513a3e 100644 --- a/ko/tutorials/llm/gemma4/gemma4.mdx +++ b/ko/tutorials/llm/gemma4/gemma4.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Model Download": 9f2919ea --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -35,61 +34,83 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" - [Google AI for Developers](https://ai.google.dev/gemma) - [ComfyUI 소스 코드 (nodes_textgen.py)](https://github.com/Comfy-Org/ComfyUI/blob/master/comfy_extras/nodes_textgen.py) -## 이용 가능한 워크플로 +## Available workflow -### Gemma 4: 텍스트 생성 +### Gemma4: Text Generation (`llm_gemma4_text_gen`) - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Gemma 4 텍스트 생성"을 검색하세요. - +Input your text prompt and optionally an image, audio, or video. Generate text output with configurable reasoning, coding, and multilingual support. + +Gemma 4 text generation workflow preview - - Comfy Cloud에서 열기 + + + Open in Comfy Cloud + + Download JSON or search "Gemma4: Text Generation" in Template Library + + + +**입력 자료** -![Gemma 4 텍스트 생성 워크플로](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_gemma4_text_gen-1.webp) +Upload these optional files to the matching nodes: -이 워크플로는 Gemma 4의 핵심 **텍스트 생성** 기능을 보여줍니다. 텍스트 프롬프트와 함께 추가적인 컨텍스트로 이미지, 오디오 파일 또는 비디오를 선택할 수 있으며, 자연어 출력을 생성합니다. 추론, 코딩 및 다국어 프롬프트까지 지원합니다. + + + `LoadImage` node 2 · `the_lily_veil.png` + + + `LoadAudio` node 5 · `voice_demo.mp3` + + + `LoadVideo` node 6 · `video_wan_vace_inpainting_input_video.mp4` + + -**입력**: -- **텍스트 프롬프트** — 질문이나 지침 -- **이미지** (선택사항) — 시각적 이해 작업용 (OCR, 객체 감지, 차트 읽기 등) -- **오디오** (선택사항) — 음성 인식 또는 오디오 전사용 -- **비디오** (선택사항) — 프레임별 비디오 이해용 (내부적으로 1 FPS로 하위 샘플링됨) +This workflow demonstrates the core **text generation** capabilities of Gemma 4. It accepts an optional image, audio file, or video as additional context alongside your text prompt, and generates natural language output — with support for reasoning, coding, and multilingual prompts. -**주요 제어 옵션**: -- **최대 길이** — 생성할 최대 토큰 수 (기본값 256) -- **샘플링 모드** — 샘플링 켜기/끄기 및 온도, top-k, top-p, 반복 페널티, 시드 조정 가능 -- **사고 모드** — 최종 답변 전 단계별 추론 활성화 -- **기본 템플릿 사용** — 모델에 맞는 내장 시스템 프롬프트 적용 +**Inputs**: +- **Text prompt** — your question or instruction +- **Image** (optional) — for visual understanding tasks (OCR, object detection, chart reading, etc.) +- **Audio** (optional) — for speech recognition or transcription +- **Video** (optional) — for video understanding across frames (subsampled to 1 FPS internally) -**출력**: -- **생성된 텍스트** — 모델의 응답을 일반 텍스트 문자열로 반환 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Thinking mode** — enable step-by-step reasoning before the final answer +- **Use default template** — apply the built-in system prompt for the model - - 이 워크플로는 서브그래프 노드를 사용해 모듈식 처리를 수행합니다. 서브그래프 문서를 확인하여 워크플로를 사용자 정의하고 확장하는 방법을 배워보세요. +**Output**: +- **Generated text** — the model's response as a plain text string + + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 모델 다운로드 +## Model Download -Gemma 4 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 관련 모델 파일을 다운로드하여 올바른 디렉토리에 저장하세요: +Gemma 4 models are loaded as text encoders in ComfyUI. Download the relevant model file and place it in the correct directory: - - 빠르고 가벼움. 소비자용 GPU에 권장됩니다. + + + Fast, lightweight. Recommended for consumer GPUs. - - 균형 잡힌 성능. 워크플로의 기본 모델입니다. + + Balanced performance. The default model in the workflow. - - 모든 Gemma 4 모델 가중치 찾아보기. + + Browse all Gemma 4 model weights. + -다운로드한 `.safetensors` 파일을 다음 위치에 저장하세요: +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ │ └── gemma4_e4b_it_fp8_scaled.safetensors +``` diff --git a/ko/tutorials/llm/qwen/qwen3.mdx b/ko/tutorials/llm/qwen/qwen3.mdx index f80db86ee..f260d024e 100644 --- a/ko/tutorials/llm/qwen/qwen3.mdx +++ b/ko/tutorials/llm/qwen/qwen3.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Model Download": d1d0cfd2 --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' @@ -25,7 +24,6 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - **ComfyUI 기본 지원** — 내장된 `TextGenerate` 노드와 함께 작동하며, 커스텀 노드가 필요하지 않습니다. - **경량성** — Qwen3.5와 동일한 텍스트 인코더 형식을 공유하며, 2B, 4B 및 9B 변형으로 제공되어 다양한 하드웨어에 맞게 사용할 수 있습니다. - ## 활용 사례 Qwen 3.0은 ComfyUI 워크플로우 내에서 구조화된 텍스트 생성과 지능형 추론이 필요한 작업에 적합합니다: @@ -35,58 +33,64 @@ Qwen 3.0은 ComfyUI 워크플로우 내에서 구조화된 텍스트 생성과 - **텍스트 기반 분석** — 텍스트 입력에서 정보를 추출하고, 콘텐츠를 분류하거나 구조화된 보고서를 생성할 수 있습니다. - **사고 체인 추론** — 복잡한 다단계 작업에서 중간 과정의 추론을 거친 후 최종 결과를 생성하는 데 유리한 사고 모드를 활성화하세요. -## 이용 가능한 워크플로우 +## Available workflow -### Qwen 3.0: 텍스트 생성 +### Qwen3.0: Text Generation (`llm_qwen3_text_gen`) - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Qwen 3.0 텍스트 생성"을 검색하세요. - +Input a text prompt to generate detailed, reasoned responses using the Qwen3-4B-Thinking model. - - Comfy Cloud에서 열기 - +Qwen 3.0 text generation workflow preview -![Qwen 3.0 텍스트 생성 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_text_gen-1.webp) + + + Open in Comfy Cloud + + + Download JSON or search "Qwen3.0: Text Generation" in Template Library + + -이 워크플로우는 Qwen 3.0의 핵심 **텍스트 생성** 기능을 보여줍니다. 텍스트 프롬프트를 받아 모델의 내장된 추론 기능을 사용해 상세하고 구조화된 응답을 생성합니다. +This workflow demonstrates the core **text generation** capabilities of Qwen 3.0. It accepts a text prompt and generates detailed, structured responses using the model's built-in reasoning capabilities. -**입력**: -- **텍스트 프롬프트** — 질문, 지시사항 또는 작업 설명 +**Inputs**: +- **Text prompt** — your question, instruction, or task description -**주요 컨트롤**: -- **최대 길이** — 생성할 토큰의 최대 수 (기본값 256) -- **샘플링 모드** — 샘플링 켜기/끄기 및 온도, top-k, top-p, 반복 페널티, 시드 조정 -- **사고 모드** — 최종 답변 전에 단계별 추론을 활성화 -- **기본 템플릿 사용** — 모델용 내장 시스템 프롬프트 적용 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Thinking mode** — enable step-by-step reasoning before the final answer +- **Use default template** — apply the built-in system prompt for the model -**출력**: -- **생성된 텍스트** — 모델의 응답을 일반 텍스트 문자열로 반환 +**Output**: +- **Generated text** — the model's response as a plain text string - - 이 워크플로우는 서브그래프 노드를 사용해 모듈식 처리를 수행합니다. 서브그래프 문서를 확인하여 워크플로우를 맞춤화하고 확장하는 방법을 배워보세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 모델 다운로드 +## Model Download -Qwen 3.0 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 모델 파일은 Qwen3.5와 공유되며, 하드웨어에 가장 적합한 변형을 다운로드하세요: +Qwen 3.0 models are loaded as text encoders in ComfyUI. The model files are shared with Qwen3.5 — download the variant that best fits your hardware: - - 경량형, 약 4.5GB. 낮은 VRAM 환경과 빠른 다운로드에 최적입니다. + + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - - 균형 잡힌 크기와 품질. 대부분의 소비자용 GPU에 권장됩니다. + + Balanced size and quality. Recommended for most consumer GPUs. - - 가장 큰 변형, 약 19GB. 더 높은 품질의 출력을 제공하며, 더 많은 VRAM이 필요합니다. + + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + -다운로드한 `.safetensors` 파일을 다음 위치에 저장하세요: +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ -│ └── qwen3.5_4b_bf16.safetensors # 또는 2B / 9B 변형 +│ └── qwen3.5_4b_bf16.safetensors # or 2b / 9b variant +``` diff --git a/ko/tutorials/llm/qwen/qwen3_5.mdx b/ko/tutorials/llm/qwen/qwen3_5.mdx index 098918a6c..802cb0d03 100644 --- a/ko/tutorials/llm/qwen/qwen3_5.mdx +++ b/ko/tutorials/llm/qwen/qwen3_5.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Model Download": 84a7c321 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -26,7 +25,6 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" - **ComfyUI 기본 제공** — 내장된 `TextGenerate` 노드와 함께 작동하며, 커스텀 노드가 필요하지 않습니다. - **경량성** — 4B 파라미터 모델로, 소비자용 GPU에 적합합니다. - ## 사용 사례 Qwen3.5는 시각적 이해와 텍스트 생성을 결합해 ComfyUI 워크플로우에 가치를 더하는 시나리오에서 뛰어난 성능을 발휘합니다: @@ -37,58 +35,78 @@ Qwen3.5는 시각적 이해와 텍스트 생성을 결합해 ComfyUI 워크플 - **시각적 질문 응답** — 이미지 내용에 대한 질문("이 장면에는 어떤 물체가 있나요?", "배경 색깔은 무엇인가요?")을 던지고 구조화된 텍스트 답변을 받을 수 있습니다. - **텍스트 읽기** — 적절한 프롬프트를 사용하면 이미지 속 보이는 텍스트나 라벨을 읽으려고 할 수 있지만, 신뢰성은 렌더링된 텍스트의 품질과 명확성에 따라 달라집니다. -## 이용 가능한 워크플로우 +## Available workflow + +### Qwen3.5: Text Generation (`llm_qwen3_5_text_gen`) + +Use the Qwen3.5 model to analyze an input image and generate descriptive text prompts. This workflow performs image captioning and reverse prompt engineering. -### Qwen3.5: 텍스트 생성 +Qwen3.5 text generation workflow preview - - JSON 파일을 다운로드하거나 템플릿 라이브러리에서 "Qwen3.5 텍스트 생성"을 검색하세요. + + + Open in Comfy Cloud + + Download JSON or search "Qwen3.5: Text Generation" in Template Library + + + +**입력 자료** - - Comfy Cloud에서 열기 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 2 · `man_with_red_hat.png` + -![Qwen3.5 텍스트 생성 워크플로우](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_5_text_gen-1.webp) +
+ Input image +
-이 워크플로우는 Qwen3.5의 **텍스트 생성 및 이미지 이해** 기능을 보여줍니다. 텍스트 프롬프트와 선택적 이미지를 받아들여 입력에 기반한 묘사 텍스트나 구조화된 분석을 생성합니다. +This workflow demonstrates the **text generation and image understanding** capabilities of Qwen3.5. It accepts a text prompt and an optional image, and generates descriptive text or structured analysis based on the input. -**입력**: -- **텍스트 프롬프트** — 질문, 지침 또는 작업 설명 -- **이미지** (선택적) — 시각적 이해 작업용 (이미지 캡션, 리버스 프롬프트 엔지니어링, 프롬프트 최적화 등) +**Inputs**: +- **Text prompt** — your question, instruction, or task description +- **Image** (optional) — for visual understanding tasks (image captioning, reverse prompt engineering, prompt optimization, etc.) -**주요 컨트롤**: -- **최대 길이** — 생성할 토큰의 최대 수 (기본값 256) -- **샘플링 모드** — 샘플링 켜기/끄기 및 온도, top-k, top-p, 반복 페널티, 시드 조정 -- **기본 템플릿 사용** — 모델용 내장 시스템 프롬프트 적용 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Use default template** — apply the built-in system prompt for the model -**출력**: -- **생성된 텍스트** — 모델의 응답을 일반 텍스트 문자열로 반환합니다. +**Output**: +- **Generated text** — the model's response as a plain text string - - 이 워크플로우는 서브그래프 노드를 사용해 모듈식 처리를 수행합니다. 서브그래프 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 배워보세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 모델 다운로드 +## Model Download -Qwen3.5 모델은 ComfyUI에서 텍스트 인코더로 로드됩니다. 하드웨어에 가장 적합한 변형을 선택하세요: +Qwen3.5 models are loaded as text encoders in ComfyUI. Choose the variant that best suits your hardware: - - 경량형, 약 4.5GB. 낮은 VRAM 환경과 빠른 다운로드에 최적입니다. + + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - - 균형 잡힌 크기와 품질. 대부분의 소비자용 GPU에 권장됩니다. + + Balanced size and quality. Recommended for most consumer GPUs. - - 가장 큰 변형, 약 19GB. 더 높은 품질의 출력을 제공하며, 더 많은 VRAM을 요구합니다. + + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + -다운로드한 `.safetensors` 파일을 다음 위치에 저장하세요: +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ -│ └── qwen3.5_4b_bf16.safetensors # 또는 2b / 9b 변형 +│ └── qwen3.5_4b_bf16.safetensors # or 2b / 9b variant +``` diff --git a/ko/tutorials/partner-nodes/anthropic/claude.mdx b/ko/tutorials/partner-nodes/anthropic/claude.mdx index 7f4670052..6dfd58a18 100644 --- a/ko/tutorials/partner-nodes/anthropic/claude.mdx +++ b/ko/tutorials/partner-nodes/anthropic/claude.mdx @@ -37,7 +37,6 @@ Anthropic Claude는 강력한 추론 능력, 안전성 및 긴 컨텍스트 처 - 해당 템플릿에서는 역할 프롬프트를 분석하고 생성하는 프롬프트를 구축했습니다. 이를 통해 사용자의 이미지를 해당 드로잉 프롬프트로 해석합니다. diff --git a/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 0961835a0..7cff413eb 100644 --- a/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/ko/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Beeble SwitchX: Video Edit": 166c2e59 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx b/ko/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx index d2266b184..5a12769ee 100644 --- a/ko/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx +++ b/ko/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Flux 1.1[pro] Image-to-Image Tutorial": 6b3cfbea --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx b/ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx index 072fe8b11..d3a0e54b7 100644 --- a/ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx +++ b/ko/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Flux.1 Kontext Max Image Partner Nodes Workflow": df399d73 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import PromptTechniques from "/snippets/ko/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/bria/background-removal.mdx b/ko/tutorials/partner-nodes/bria/background-removal.mdx index 0a9179494..798d791d7 100644 --- a/ko/tutorials/partner-nodes/bria/background-removal.mdx +++ b/ko/tutorials/partner-nodes/bria/background-removal.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Video Background Processing": aab02df5 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index b69566f3a..d43a6f2ed 100644 --- a/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 1ae5cf45 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index bdc045829..ef9824150 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Available workflows": 16803a18 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index 53fd63a09..2a9e90f56 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Using real-person and AI-generated portraits in ComfyUI for Seedance 2.0": d9966fae --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index e8fc92f6e..333dfcf73 100644 --- a/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/ko/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Get started": d14874f4 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx index 10d5dae94..cf061dc92 100644 --- a/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/ko/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Get started": 64517938 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/google/nano-banana-2-lite.mdx b/ko/tutorials/partner-nodes/google/nano-banana-2-lite.mdx index a75cb7760..5ca142f34 100644 --- a/ko/tutorials/partner-nodes/google/nano-banana-2-lite.mdx +++ b/ko/tutorials/partner-nodes/google/nano-banana-2-lite.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 5a827068 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/google/nano-banana-2.mdx b/ko/tutorials/partner-nodes/google/nano-banana-2.mdx index 1bad8129b..4800f7aa7 100644 --- a/ko/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/ko/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Get started": f6189d9e --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/google/nano-banana-pro.mdx b/ko/tutorials/partner-nodes/google/nano-banana-pro.mdx index 7516b6ec4..b058136ab 100644 --- a/ko/tutorials/partner-nodes/google/nano-banana-pro.mdx +++ b/ko/tutorials/partner-nodes/google/nano-banana-pro.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 38fdf056 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index 61ca9f379..069e97943 100644 --- a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "HappyHorse 1.0 video edit": 5b531078 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index ea0ec41a9..267533319 100644 --- a/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/ko/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Getting started": 27bbb438 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 51d95b1d0..06ab5fd82 100644 --- a/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/ko/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Advanced features": 6b37a964 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index af4d3202f..a657242c2 100644 --- a/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/ko/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Multi-view-to-3D workflow": ff251560 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 88c0bb015..a0b341db6 100644 --- a/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/ko/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx index ce4f0354d..cc195fb6d 100644 --- a/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/ko/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Tips for better results": b6fdceb5 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/krea2/krea2-t2i.mdx b/ko/tutorials/partner-nodes/krea2/krea2-t2i.mdx index 0709ea2bc..0d008276e 100644 --- a/ko/tutorials/partner-nodes/krea2/krea2-t2i.mdx +++ b/ko/tutorials/partner-nodes/krea2/krea2-t2i.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Additional Notes": 7099b5ae --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx b/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx index a8afd2bdc..9ded9c7b5 100644 --- a/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/ko/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -22,8 +22,6 @@ translationBlockHashes: "Key takeaway": 77d51b1d --- - - **ComfyUI**에서 **Luma Uni-1** 은 통합 이미지 작업을 위한 **파트너 노드**로 제공됩니다: **Create** 그래프는 프롬프트(선택적 레퍼런스 포함)로 새 이미지를 생성하고, **Modify** 그래프는 기존 이미지를 입력으로 받아 편집합니다. 두 모드 모두 평소 워크플로처럼 **Load Image** / **Save Image** 노드를 연결하고, Luma 노드에서 프롬프트, 시드, 종횡비, 레퍼런스 슬롯을 설정한 후 로컬에서 그래프를 실행하거나 **Comfy Cloud**에서 템플릿을 열면 됩니다. Luma는 Uni-1을 비확산, 디코더 전용 자기회귀 모델로 설명하며, 그리기 전에 프롬프트를 추론합니다. 캔버스에서 가장 중요한 것은 **Create vs Modify** 모드를 선택하고, 레퍼런스를 명확히 라벨링하고, 시드로 반복 작업하는 것입니다. diff --git a/ko/tutorials/partner-nodes/meshy/meshy-6.mdx b/ko/tutorials/partner-nodes/meshy/meshy-6.mdx index 1cd382114..8dcecf775 100644 --- a/ko/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/ko/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Multi-view to Model Workflow": 0cf6bb73 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 36c342eb2..7db972d3d 100644 --- a/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/ko/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -12,12 +12,10 @@ translationBlockHashes: "Moonvalley Video-to-Video Workflow": 5de07f68 --- - **서비스 제공 불가**: Moonvalley API 서비스가 더 이상 제공되지 않습니다. 이러한 노드는 지원 중단되었으며, 예상대로 작동하지 않을 수 있습니다. - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -58,7 +56,6 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로

워크플로우 파일(JSON 형식) 다운로드하기

- ### 2. 워크플로우 실행 단계 따라하기 ![텍스트 기반 비디오 생성 워크플로우](/images/tutorial/api_nodes/moonvalley/api_moonvalley_text_to_video.jpg) @@ -69,7 +66,6 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 4. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 비디오 생성 시작 5. API가 결과를 반환한 후, `Save Video` 노드에서 생성된 비디오를 확인할 수 있습니다. 비디오는 또한 `ComfyUI/output/` 디렉토리에 저장됩니다 - ## Moonvalley 이미지 기반 비디오 생성 워크플로우 ### 1. 워크플로우 파일 다운로드하기 @@ -99,7 +95,6 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 5. `Run` 버튼을 클릭하거나, 단축키 `Ctrl(cmd) + Enter`를 사용해 비디오 생성 시작 6. API가 결과를 반환한 후, `Save Video` 노드에서 생성된 비디오를 확인할 수 있습니다. 비디오는 또한 `ComfyUI/output/` 디렉토리에 저장됩니다 - ## Moonvalley 비디오 기반 비디오 생성 워크플로우 `Moonvalley Marey Video to Video` 노드를 통해 참조 비디오를 입력해 비디오를 다시 그리는 작업을 할 수 있습니다. 참조 비디오의 모션이나 캐릭터 포즈를 활용해 비디오를 생성할 수 있습니다. @@ -124,7 +119,6 @@ Moonvalley Marey 리얼리즘 v1.5는 시네마틱 수준의 창작을 목표로 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video_input.mp4" > - ### 2. 워크플로우 실행 단계 따라하기 ![비디오 기반 비디오 생성 워크플로우](/images/tutorial/api_nodes/moonvalley/api_moonvalley_video_to_video.jpg) diff --git a/ko/tutorials/partner-nodes/openai/dall-e-2.mdx b/ko/tutorials/partner-nodes/openai/dall-e-2.mdx index 394c2fdfb..0204f30df 100644 --- a/ko/tutorials/partner-nodes/openai/dall-e-2.mdx +++ b/ko/tutorials/partner-nodes/openai/dall-e-2.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/ko/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/openai/dall-e-3.mdx b/ko/tutorials/partner-nodes/openai/dall-e-3.mdx index 46c6a83dd..53cacf0d5 100644 --- a/ko/tutorials/partner-nodes/openai/dall-e-3.mdx +++ b/ko/tutorials/partner-nodes/openai/dall-e-3.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/ko/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -57,7 +56,6 @@ DALL·E 3는 OpenAI의 최신 이미지 생성 모델로, 텍스트 프롬프트 ![ComfyUI openai-dall-e-3 워크플로우](/images/tutorial/api_nodes/openai/openai-dall-e-3/text2image.jpg) - 1. ComfyUI에서 **OpenAI DALL·E 3** 노드를 추가하세요. 2. 프롬프트 텍스트 박스에 생성하고자 하는 이미지의 설명을 입력하세요. 3. 필요에 따라 선택적 매개변수(품질, 스타일, 크기 등)를 조정하세요. diff --git a/ko/tutorials/partner-nodes/openai/gpt-image-1.mdx b/ko/tutorials/partner-nodes/openai/gpt-image-1.mdx index 8644349c1..ca6c4a82c 100644 --- a/ko/tutorials/partner-nodes/openai/gpt-image-1.mdx +++ b/ko/tutorials/partner-nodes/openai/gpt-image-1.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/ko/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx b/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx index 4d31d2521..03cc1406c 100644 --- a/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/ko/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Hybrid Pipelines": 9c6479f0 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/openrouter/llm.mdx b/ko/tutorials/partner-nodes/openrouter/llm.mdx index 7fbc1f3c1..f2109e54b 100644 --- a/ko/tutorials/partner-nodes/openrouter/llm.mdx +++ b/ko/tutorials/partner-nodes/openrouter/llm.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Example workflow (`api_openrouter_llm`)": f1dbfe6a --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/overview.mdx b/ko/tutorials/partner-nodes/overview.mdx index bf55ec225..017ea3ecb 100644 --- a/ko/tutorials/partner-nodes/overview.mdx +++ b/ko/tutorials/partner-nodes/overview.mdx @@ -17,7 +17,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import 요구사항 from "/snippets/ko/tutorials/partner-nodes/requirements.mdx"; import FAQ from "/snippets/ko/tutorials/partner-nodes/faq.mdx"; @@ -39,7 +38,6 @@ ComfyUI 계정 API 키로 로그인하는 방법 알아보기 ![Comfy API 키 로그인 선택](/images/interface/setting/user/user-login-api-1.jpg) - ## ComfyUI 계정 API 키 통합을 이용해 유료 모델 파트너 노드 호출하기 현재 우리는 ComfyUI API 키 통합을 통해 유료 모델 파트너 노드를 호출하는 서비스에 접근할 수 있도록 지원하고 있습니다. API 키 통합 섹션을 참고해 유료 모델 파트너 노드를 호출하는 방법을 알아보세요. @@ -52,8 +50,6 @@ ComfyUI 계정 API 키로 로그인하는 방법 알아보기 API 키 통합 섹션을 참고해 유료 모델 파트너 노드를 호출하는 방법 알아보기 - - ## 파트너 노드의 장점 파트너 노드는 ComfyUI 사용자들에게 다음과 같은 중요한 장점을 제공합니다: @@ -63,7 +59,6 @@ API 키 통합 섹션을 참고해 유료 모델 파트너 노드를 호출하 - **간편한 경험**: API 키를 관리하거나 복잡한 API 요청을 처리할 필요가 없습니다 - **비용 통제**: 선불 시스템 덕분에 예상치 못한 비용 없이 지출을 완벽히 관리할 수 있습니다 - ## 가격 정책 diff --git a/ko/tutorials/partner-nodes/pricing.mdx b/ko/tutorials/partner-nodes/pricing.mdx index 5fc97d7e2..76612a737 100644 --- a/ko/tutorials/partner-nodes/pricing.mdx +++ b/ko/tutorials/partner-nodes/pricing.mdx @@ -45,7 +45,6 @@ translationBlockHashes: "Cloud GPU": 103c55e6 --- - 다음 표에는 현재 파트너 노드의 가격 정책이 나와 있습니다. 모든 가격은 크레딧 단위로 표시됩니다. ## Anthropic diff --git a/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx b/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx index 9c03e84ac..ba503bb67 100644 --- a/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/ko/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Additional notes": d5f3f109 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/reve/reve-image.mdx b/ko/tutorials/partner-nodes/reve/reve-image.mdx index 4b8a140c2..111bf5c9e 100644 --- a/ko/tutorials/partner-nodes/reve/reve-image.mdx +++ b/ko/tutorials/partner-nodes/reve/reve-image.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Available nodes": 5927bf70 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/rodin/model-generation.mdx b/ko/tutorials/partner-nodes/rodin/model-generation.mdx index a7388b30a..fce19b7c7 100644 --- a/ko/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/ko/tutorials/partner-nodes/rodin/model-generation.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Other Related Nodes": c885ce31 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/runway/image-generation.mdx b/ko/tutorials/partner-nodes/runway/image-generation.mdx index 3fe2720d7..cc6e049a9 100644 --- a/ko/tutorials/partner-nodes/runway/image-generation.mdx +++ b/ko/tutorials/partner-nodes/runway/image-generation.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Runway Image Reference-to-Image Workflow": cc84e030 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/runway/video-generation.mdx b/ko/tutorials/partner-nodes/runway/video-generation.mdx index aab91e761..a57789da3 100644 --- a/ko/tutorials/partner-nodes/runway/video-generation.mdx +++ b/ko/tutorials/partner-nodes/runway/video-generation.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "First-Last Frame Video Generation Workflow": b7f1102b --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -88,7 +87,6 @@ Runway는 생성형 AI에 중점을 둔 기업으로, 강력한 동영상 생성 ## 첫 번째 및 마지막 프레임 동영상 생성 워크플로우 - ### 1. 워크플로우 파일 다운로드 아래 동영상에는 `메타데이터`에 워크플로우 정보가 포함되어 있습니다. 해당 워크플로우를 다운로드해 ComfyUI로 드래그하여 로드해 주세요. diff --git a/ko/tutorials/partner-nodes/sonilo/video-to-music.mdx b/ko/tutorials/partner-nodes/sonilo/video-to-music.mdx index a7758159d..c262ddc4e 100644 --- a/ko/tutorials/partner-nodes/sonilo/video-to-music.mdx +++ b/ko/tutorials/partner-nodes/sonilo/video-to-music.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Use cases": 90983ea2 --- - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/tripo/model-generation.mdx b/ko/tutorials/partner-nodes/tripo/model-generation.mdx index 8d336dfc3..2d294579d 100644 --- a/ko/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/ko/tutorials/partner-nodes/tripo/model-generation.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Subsequent Task Processing for the Same Task": d24582e4 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; @@ -73,7 +71,6 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 - 아래 이미지를 입력 이미지로 다운로드하세요. ![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/image_to_model/panda.jpg) @@ -90,7 +87,6 @@ ComfyUI는 현재 해당 Tripo API를 기본적으로 통합하여 ComfyUI에서 3. `Run` 버튼을 클릭하거나 단축키 `Ctrl(cmd) + Enter`를 사용해 모델 생성을 실행하세요. 워크플로우가 완료된 후 해당 모델은 자동으로 `ComfyUI/output/` 디렉토리에 저장됩니다. 4. 모델 다운로드는 텍스트 기반 3D 모델 생성 섹션의 지침을 참조하세요. - ## 멀티뷰 모델 생성 워크플로우 ### 1. 워크플로우 파일 다운로드 @@ -110,7 +106,6 @@ Generate a 3D model from multiple view images for enhanced accuracy. - 아래 이미지를 입력 이미지로 다운로드하세요. ![전면 보기](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/multiview_to_image/front.jpg) diff --git a/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx index 6152aaaac..8a8523661 100644 --- a/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/ko/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Available Workflows": 5ec1d9c5 --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/partner-nodes/wan/wan2-7.mdx b/ko/tutorials/partner-nodes/wan/wan2-7.mdx index 39ded22ed..1a1d00326 100644 --- a/ko/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/ko/tutorials/partner-nodes/wan/wan2-7.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Wan2.7 video edit": 0d2511bb --- - - import ReqHint from "/snippets/ko/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; diff --git a/ko/tutorials/utility/depth-anything-3.mdx b/ko/tutorials/utility/depth-anything-3.mdx index 6dddf7005..3a084a4d6 100644 --- a/ko/tutorials/utility/depth-anything-3.mdx +++ b/ko/tutorials/utility/depth-anything-3.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Community Resources": 8c5d05ab --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' # ComfyUI Depth Anything 3 소개 @@ -54,80 +53,103 @@ ComfyUI/ │ │ └── depth_anything_3_metric_large.safetensors ``` -## 예제 워크플로우 +## Example Workflows ---- +### Depth Anything 3: Image Depth Estimation (`utility_depth_anything3_image_depth_estimation`) -## 1. 이미지 깊이 추정 +Upload one image and generate a depth map using Depth Anything 3. View a side-by-side comparison of the original image and depth output. -**기능 설명:** 이미지 1장을 업로드하고 **Image Depth Estimation (Depth Anything 3)**을 실행하여 깊이 맵을 생성합니다. **Depth Preview**에 원본 이미지와 깊이 출력의 나란히 비교 결과가 표시됩니다. +Depth Anything 3 image depth estimation workflow preview - - JSON 다운로드 또는 템플릿 라이브러리에서 "Depth Anything 3" 검색 + + Open in Comfy Cloud - - 이 워크플로우의 예제 입력 이미지 가져오기 + + Download JSON or search "Depth Anything 3: Image Depth Estimation" in Template Library -
- 이미지 깊이 추정 출력 - 이미지 깊이 추정 비교 +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 85 · `retro_futuristic_home.png` + + + +
+ Input image
-### 실행 단계 +### Steps to Run -1. **LoadImage** — 입력 이미지 로드 -2. **LoadDA3Model** — Depth Anything 3 변형 선택 -3. **실행** — Queue 클릭 또는 `Cmd+Enter` -4. 워크플로우가 깊이 맵과 나란히 비교 결과 출력 +1. **LoadImage** — load your input image +2. **LoadDA3Model** — select a Depth Anything 3 variant +3. **Run** — click Queue or use `Cmd+Enter` +4. The workflow outputs a depth map and side-by-side comparison - - 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 사용자 지정 및 확장에 대한 자세한 내용은 서브그래프 문서를 확인하세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ---- +### Depth Anything 3: Video Depth Estimation (`utility_depth_anything3_video_depth_estimation`) -## 2. 비디오 깊이 추정 +Upload a video to generate a per-frame depth sequence. Inside the subgraph, **GetVideoComponents** splits the input video into frames, **LoadDA3Model** loads the model, and **SetVideoComponents** reassembles the depth frames back into a video output. -**기능 설명:** 비디오를 업로드하고 **Video Depth Estimation (Depth Anything 3)**을 실행하여 프레임별 깊이 시퀀스를 생성합니다. 서브그래프 내에서 **GetVideoComponents**가 입력 비디오를 프레임으로 분할하고, **LoadDA3Model**이 모델을 로드하며, **SetVideoComponents**가 깊이 프레임을 비디오로 재구성합니다. +Depth Anything 3 video depth estimation workflow preview - - JSON 다운로드 또는 템플릿 라이브러리에서 "Depth Anything 3" 검색 - - Comfy Cloud에서 열기 + Open in Comfy Cloud + + + Download JSON or search "Depth Anything 3: Video Depth Estimation" in Template Library -![비디오 깊이 추정 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_depth_anything3_video_depth_estimation-1.webp) +**입력 자료** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 87 · `empty_room_assembly.mp4` + + + +
+ +
-### 실행 단계 +### Steps to Run -1. **LoadVideo** — 입력 비디오 로드 -2. **모델 선택** — **Small**, **Base**, **Mono-Large** 또는 **Metric-Large** 중 선택 -3. **실행** — Queue 클릭 또는 `Cmd+Enter` -4. 워크플로우가 프레임별 깊이 맵 비디오 출력 +1. **LoadVideo** — load your input video +2. **Select Model** — choose between **Small**, **Base**, **Mono-Large**, or **Metric-Large** +3. **Run** — click Queue or use `Cmd+Enter` +4. The workflow outputs a video with per-frame depth maps - - 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 사용자 지정 및 확장에 대한 자세한 내용은 서브그래프 문서를 확인하세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 모델 변형 +## Model Variants -| 변형 | head_type | 하늘 감지 | 신뢰도 | 카메라 디코더 | 최적 용도 | -|------|-----------|:---------:|:------:|:-------------:|-----------| -| **Small** | dualdpt | ❌ | ✅ | ✅ | 빠른 추론, 모바일/엣지 | -| **Base** | dualdpt | ❌ | ✅ | ✅ | 균형 잡힌 성능 | -| **Mono-Large** | dpt | ✅ | ❌ | ❌ | 하늘 감지 지원 단일 뷰 깊이 | -| **Metric-Large** | dpt | ✅ | ❌ | ❌ | 미터 단위 물리적 깊이 | +| Variant | head_type | has_sky | has_confidence | camera_decoder | Best for | +|---------|-----------|:-------:|:--------------:|:--------------:|----------| +| **Small** | dualdpt | ❌ | ✅ | ✅ | Fast inference, mobile/edge | +| **Base** | dualdpt | ❌ | ✅ | ✅ | Balanced performance | +| **Mono-Large** | dpt | ✅ | ❌ | ❌ | Monocular depth with sky detection | +| **Metric-Large** | dpt | ✅ | ❌ | ❌ | Physical metric depth in metres | -- **Small**과 **Base**는 `dualdpt` 헤드 유형을 사용하며, 신뢰도 추정 및 카메라 디코더를 지원합니다(다중 뷰 애플리케이션용). -- **Mono-Large**와 **Metric-Large**는 `dpt` 헤드 유형을 사용하며, 하늘 감지를 지원합니다. Metric-Large는 미터 단위의 원시 깊이를 출력합니다。 +- **Small** and **Base** use the `dualdpt` head type with confidence estimation and camera decoder support for multi-view applications. +- **Mono-Large** and **Metric-Large** use the `dpt` head type with sky detection. Metric-Large outputs raw depth in metres. -## 커뮤니티 리소스 +## Community Resources -- [Depth Anything 3 GitHub (ByteDance-Seed)](https://github.com/ByteDance-Seed/Depth-Anything-3) — 연구 논문 및 코드 -- [Comfy-Org/Depth-Anything-3](https://huggingface.co/Comfy-Org/Depth-Anything-3) — 공식 ComfyUI 모델 가중치 +- [Depth Anything 3 GitHub (ByteDance-Seed)](https://github.com/ByteDance-Seed/Depth-Anything-3) — Research paper and code +- [Comfy-Org/Depth-Anything-3](https://huggingface.co/Comfy-Org/Depth-Anything-3) — Official ComfyUI model weights diff --git a/ko/tutorials/utility/face-detection/mediapipe.mdx b/ko/tutorials/utility/face-detection/mediapipe.mdx index 539561b3c..89c0c4c83 100644 --- a/ko/tutorials/utility/face-detection/mediapipe.mdx +++ b/ko/tutorials/utility/face-detection/mediapipe.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Community Resources": ec7aa0f5 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -31,33 +30,48 @@ MediaPipe 얼굴 감지는 ComfyUI에서 기본적으로 지원되며([PR #14009 > **범위:** 얼굴 감지만 제공 — BlazeFace + FaceMesh v2 + ARKit 블렌드셰이프. 이는 손, 자세 또는 전신 감지를 포함하지 않습니다. -## MediaPipe 얼굴 감지 워크플로우 +## MediaPipe Face Detection Workflow -### 1. 워크플로우 다운로드하기 +### Mediapipe: Image Face Detection (`utility_face_detection_mediapipe`) -ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` → `템플릿 둘러보기`로 이동해 유틸리티 카테고리에서 "Mediapipe: 이미지 얼굴 감지"를 찾으세요. +Input an image and detect up to 6 facial landmarks per face, enabling ultrafast multi-face detection. - - 워크플로우 다운로드 - +Mediapipe image face detection workflow preview - - 클라우드에서 열기 - + + + Open in Comfy Cloud + + + Download JSON or search "Mediapipe: Image Face Detection" in Template Library + + - - 이 워크플로우의 예시 입력 이미지 받기 - +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 5 · `soft_neon_girl.png` + + + +
+ Input image +
+ +### 1. Download the Workflow -![MediaPipe 얼굴 감지 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_face_detection_mediapipe-1.webp) +Update your ComfyUI to the latest version, then go to `Workflow` → `Browse Templates` and find "Mediapipe: Image Face Detection" under the Utility category. -### 2. 모델 다운로드하기 +### 2. Download the Model -MediaPipe 얼굴 감지 모델은 [Comfy-Org MediaPipe 모델 저장소](https://huggingface.co/Comfy-Org/mediapipe)에 호스팅되어 있습니다. +The MediaPipe Face Detection model is hosted on the [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe). -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) -다음과 같은 디렉토리 구조에 배치하세요: +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -66,54 +80,54 @@ MediaPipe 얼굴 감지 모델은 [Comfy-Org MediaPipe 모델 저장소](https:/ └── mediapipe_face_fp32.safetensors ``` -### 3. 워크플로우 사용하기 +### 3. Using the Workflow -이 워크플로우는 얼굴 감지, 시각화 및 마스크 생성을 조율하는 **서브그래프** 노드를 사용합니다. 서브그래프는 다음과 같은 컨트롤을 노출합니다: +This workflow uses a **subgraph** node that orchestrates face detection, visualization, and mask generation. The subgraph exposes the following controls: - - 이 워크플로우는 서브그래프 노드를 사용해 모듈식 처리를 수행합니다. 서브그래프 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 알아보세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -**서브그래프 입력:** +**Subgraph inputs:** -| 입력 | 설명 | +| Input | Description | |-------|-------------| -| **image** | 분석할 입력 이미지 배치 | -| **face_landmarker** | 선택적. 내장 모델 로더를 사용하려면 비워두세요. 외부 `FACE_DETECTION_MODEL` 출력을 연결해 덮어쓰세요 | +| **image** | Input image batch to analyze | +| **face_landmarker** | Optional. Leave empty to use the built-in model loader. Connect an external `FACE_DETECTION_MODEL` output to override | -**서브그래프 매개변수:** +**Subgraph parameters:** -| 매개변수 | 기본값 | 설명 | +| Parameter | Default | Description | |-----------|:-------:|-------------| -| **model_name** | `mediapipe_face_fp32.safetensors` | `ComfyUI/models/detection/`에 있는 체크포인트. 없으면 위에서 모델을 다운로드하세요 | -| **detector_variant** | `short` | **short** — 가까운/큰 얼굴용으로 조정됨(~2m 범위). **full** — 더 작고 먼 얼굴까지 포괄하며 느림. **both** — 두 감지기를 모두 실행하고 프레임당 더 많은 얼굴을 찾은 것을 유지(~2배 비용) | -| **num_faces** | `1` | 프레임당 반환할 최대 얼굴 수. `0` = 제한 없음(감지된 모든 얼굴 반환). 범위: 0–16 | -| **custom_face_oval** | `false` | 마스크 출력에 얼굴 윤곽 영역 포함 | -| **custom_lips** | `false` | 마스크에 입술 포함(활성화된 다른 영역과 합집합) | -| **custom_left_eye** | `false` | 마스크에 왼쪽 눈 영역 포함 | -| **custom_right_eye** | `false` | 마스크에 오른쪽 눈 영역 포함 | -| **custom_irises** | `false` | 마스크에 홍채 영역 포함 | +| **model_name** | `mediapipe_face_fp32.safetensors` | Checkpoint in `ComfyUI/models/detection/`. If missing, download the model above | +| **detector_variant** | `short` | **short** — tuned for close/large faces (~2 m range). **full** — covers smaller/farther faces (~5 m), slower. **both** — runs both detectors and keeps whichever found more faces per frame (~2× cost) | +| **num_faces** | `1` | Maximum faces to return per frame. `0` = no cap (return all detected). Range: 0–16 | +| **custom_face_oval** | `false` | Include face-outline region in the mask output | +| **custom_lips** | `false` | Include lips in the mask (union with other enabled regions) | +| **custom_left_eye** | `false` | Include left eye region in the mask | +| **custom_right_eye** | `false` | Include right eye region in the mask | +| **custom_irises** | `false` | Include iris regions in the mask | -마스크 토글은 내부적으로 커스텀 모드를 사용합니다: 체크된 영역만 채워지고, 여러 ON 영역은 프레임당 하나의 마스크로 **합집합**됩니다. +Mask toggles use custom mode internally: only checked regions are filled; multiple ON regions are **unioned** into one mask per frame. -**서브그래프 출력:** +**Subgraph outputs:** -| 출력 | 유형 | 설명 | +| Output | Type | Description | |--------|------|-------------| -| **face_landmarks** | `FACE_LANDMARKS` | 프레임별 얼굴과 478개의 2D/3D 랜드마크, ARKit-52 블렌드셰이프, 메쉬 토폴로지 데이터 — 시각화 및 마스크 노드에 공급 | -| **bboxes** | `BOUNDING_BOX` | 얼굴 경계 상자 — `DrawBBoxes` 노드와 호환 | -| **mask** | `MASK` | 활성화된 영역 토글에 따른 이진 마스크 | +| **face_landmarks** | `FACE_LANDMARKS` | Per-frame faces with 478 2D/3D landmarks, ARKit-52 blendshapes, mesh topology data — feeds into visualization and mask nodes | +| **bboxes** | `BOUNDING_BOX` | Face bounding boxes — compatible with `DrawBBoxes` node | +| **mask** | `MASK` | Binary mask from the enabled region toggles | -### 4. 워크플로우 실행하기 +### 4. Run the Workflow -1. 모델 파일이 `ComfyUI/models/detection/`에 배치되었는지 확인하세요. -2. `이미지 로드` 노드에서 이미지를 불러오세요. -3. 필요에 따라 감지 매개변수를 조정하세요. -4. `큐`를 클릭하거나 `Ctrl(Cmd) + Enter`를 사용해 실행하세요. -5. 워크플로우는 메쉬 오버레이, 경계 상자 및 마스크 미리보기를 출력합니다. +1. Ensure the model file is placed in `ComfyUI/models/detection/` +2. Load an image in the `Load Image` node +3. Adjust detection parameters as needed +4. Click `Queue` or use `Ctrl(Cmd) + Enter` to run +5. The workflow outputs the mesh overlay, bounding boxes, and mask preview -## 커뮤니티 리소스 +## Community Resources -- [MediaPipe GitHub](https://github.com/google-ai-edge/mediapipe) — 상위 MediaPipe 프레임워크 -- [Comfy-Org/mediapipe](https://huggingface.co/Comfy-Org/mediapipe) — 공식 ComfyUI 모델 가중치 -- [ComfyUI 서브그래프 가이드](https://docs.comfy.org/interface/features/subgraph) — 서브그래프 작동 방식 알아보기 +- [MediaPipe GitHub](https://github.com/google-ai-edge/mediapipe) — Upstream MediaPipe framework +- [Comfy-Org/mediapipe](https://huggingface.co/Comfy-Org/mediapipe) — Official ComfyUI model weights +- [ComfyUI Subgraph Guide](https://docs.comfy.org/interface/features/subgraph) — Learn how subgraphs work diff --git a/ko/tutorials/utility/image-upscale.mdx b/ko/tutorials/utility/image-upscale.mdx index b29f61203..b4666ced3 100644 --- a/ko/tutorials/utility/image-upscale.mdx +++ b/ko/tutorials/utility/image-upscale.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Tips": a966ce44 --- - 이 가이드는 ComfyUI에서의 이미지 업스케일링 워크플로우를 다룹니다. 여기에는 다양한 사용 사례에 맞는 로컬 모델과 파트너 노드 옵션이 포함됩니다. diff --git a/ko/tutorials/utility/moge.mdx b/ko/tutorials/utility/moge.mdx index 4876243f0..82fbf5410 100644 --- a/ko/tutorials/utility/moge.mdx +++ b/ko/tutorials/utility/moge.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Community Resources": d9fcd8cd --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" # ComfyUI MoGe 소개 @@ -62,84 +60,118 @@ ComfyUI/ │ │ └── moge_1_vitl_fp16.safetensors ``` -## 예제 워크플로우 - ---- +## Example Workflows -## 1. 깊이 추정 +### MoGe: Depth Estimation (`utility_moge_depth_estimation`) -**작동 방식**: 단일 이미지를 입력받아 계량 깊이 지도, 컬러화된 깊이 미리보기 및 마스크를 추정합니다. MoGe가 한 번의 순방향 전파로 추정한 동일한 계량 스케일의 깊이를 출력합니다. 합성, 깊이 기반 효과 또는 메시 생성 전 처리용 장면 깊이 참조로 유용합니다. +Upload a single RGB image and generate a colored depth preview and raw depth map. -MoGe는 이미지로부터 카메라의 시야각(FOV)도 추정하며, 이를 정확성을 더욱 높이기 위해 실제값으로 덮어쓸 수도 있습니다. +MoGe depth estimation workflow preview - - JSON 다운로드하거나 템플릿 라이브러리에서 "MoGe 깊이 추정" 검색 + + Open in Comfy Cloud - - 이 워크플로우의 예제 입력 이미지 받기 + + Download JSON or search "MoGe: Depth Estimation" in Template Library -
- 깊이 추정 컬러 미리보기 - 깊이 추정 원본 미리보기 -
-### 1.1 실행 단계 +**입력 자료** -1. `LoadMoGeModel` 노드가 MoGe 체크포인트를 로드했는지 확인 -2. `Load Image` 노드에서 이미지를 로드 -3. `Queue`를 클릭하거나 `Ctrl(cmd) + Enter`를 사용해 실행 -4. 워크플로우는 컬러화된 깊이 미리보기, 원본 깊이 미리보기 및 마스크를 출력합니다. +Upload this file to the matching `LoadImage` node: ---- + + + `LoadImage` node 9 · `alien_world.png` + + + +### 1.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded a MoGe checkpoint +2. Load an image in the `Load Image` node +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run +4. The workflow outputs colored depth preview, raw depth preview, and a mask + +### MoGe: Perspective Geometry Estimation (`3d_moge_perspective_to_mesh`) -## 2. 투시도에서 메시 생성 +Upload an image to estimate its perspective geometry. Generate a 3D depth map and surface normals from the input, then convert to a textured GLB mesh. -**작동 방식**: 단일 투시 사진을 질감이 적용된 GLB 메시로 변환하며, 법선 및 깊이 미리보기를 함께 제공합니다. MoGe는 보이는 장면에서 점 지도, 깊이 및 법선을 추정한 후 이를 메시로 변환합니다. 이는 **단안 기하학 추정**으로, 가려진 부분과 객체 뒷면은 누락되거나 조각나게 됩니다. 빠른 장면 프로토타입, 참조 기하학 또는 깊이와 법선을 메시로 시각화하는 데 유용하지만, 다중뷰 3D 재구성의 대체물은 아닙니다. +MoGe perspective to mesh workflow preview - - JSON 다운로드하거나 템플릿 라이브러리에서 "3D MoGe 투시도에서 메시 생성" 검색 + + Open in Comfy Cloud - - 이 워크플로우의 예제 입력 이미지 받기 + + Download JSON or search "MoGe: Perspective Geometry Estimation" in Template Library -![투시도에서 메시 생성 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_perspective_to_mesh-1.webp) -### 2.1 실행 단계 +**입력 자료** -1. `LoadMoGeModel` 노드가 MoGe 체크포인트를 로드했는지 확인 -2. `Load Image` 노드에서 투시 사진을 로드 -3. (선택사항) OpenGL 및 DirectX 법선 미리보기를 확인 -4. `Queue`를 클릭하거나 `Ctrl(cmd) + Enter`를 사용해 실행 +Upload this file to the matching `LoadImage` node: ---- + + + `LoadImage` node 9 · `modern_living_room.png` + + -## 3. 파노라마에서 메시 생성 +
+ Input image +
+ +This is **monocular geometry estimation**: occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. + +### 2.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded a MoGe checkpoint +2. Load a perspective photo in the `Load Image` node +3. (Optional) View the OpenGL and DirectX normal previews +4. Click `Queue` or use `Ctrl(cmd) + Enter` to run -**작동 방식**: 등거리(360°) 파노라마를 질감이 적용된 GLB 메시로 변환합니다. 워크플로우는 `MoGePanoramaInference`를 사용해 파노라마를 12개의 투시도로 분할한 후 각각의 투시도에서 단안 기하학 추정을 독립적으로 수행한 후 이를 하나의 메시로 합칩니다. 각 세그먼트는 여전히 단일뷰 추정이므로 결과는 대략적인 장면 재구성입니다. 360° 장면의 공간적 개요를 얻는 데 유용하지만, 가려진 부분과 표면 뒤쪽의 기하학은 누락되거나 조각나게 됩니다. +### Moge: Panorama to Mesh (`3d_moge_panorama_to_mesh`) + +Upload an equirectangular 360° panorama image and generate a textured GLB mesh with vertex colors. + +MoGe panorama to mesh workflow preview - - JSON 다운로드하거나 템플릿 라이브러리에서 "3D MoGe 파노라마에서 메시 생성" 검색 + + Open in Comfy Cloud - - 이 워크플로우의 예제 입력 이미지 받기 + + Download JSON or search "Moge: Panorama to Mesh" in Template Library -![파노라마에서 메시 생성 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_panorama_to_mesh-1.webp) -### 3.1 실행 단계 +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 28 · `lego_street_panorama.png` + + + +
+ Input panorama +
+ +The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, run monocular geometry estimation on each view independently, then merge them into a single mesh. + +### 3.1 Steps to Run -1. `LoadMoGeModel` 노드가 MoGe 체크포인트 중 하나를 로드했는지 확인 -2. `Load Image` 노드에서 등거리 파노라마 이미지를 로드 -3. `Queue`를 클릭하거나 `Ctrl(cmd) + Enter`를 사용해 실행 +1. Ensure the `LoadMoGeModel` node has loaded one of the MoGe checkpoints +2. Load an equirectangular panorama image in the `Load Image` node +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run --- -## 커뮤니티 자료 +## Community Resources -- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe) — 연구 논문 및 코드 -- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe) — 공식 ComfyUI 모델 가중치 +- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe): Research paper and code +- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe): Official ComfyUI model weights diff --git a/ko/tutorials/utility/pose-detection-sdpose.mdx b/ko/tutorials/utility/pose-detection-sdpose.mdx index e4f0b2dbb..731f2b34f 100644 --- a/ko/tutorials/utility/pose-detection-sdpose.mdx +++ b/ko/tutorials/utility/pose-detection-sdpose.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": 409c3058 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -31,69 +30,161 @@ SDPose + RT-DETRv4는 ComfyUI에 기본 지원됩니다 (PR [#12748](https://git > **제한 사항:** 감지 정확도는 이미지 해상도와 피사체 가시성에 따라 달라집니다. 매우 가려져 있거나 아주 작은 피사체는 키포인트가 적게 생성될 수 있습니다. -## SDPose 워크플로 +## SDPose Workflows -사용 사례에 따라 네 가지 워크플로가 제공됩니다: +Four workflows are available depending on your use case: -| 워크플로 | 입력 | 출력 | 사용 사례 | +| Workflow | Input | Output | Use Case | |----------|-------|--------|----------| -| 다중 인물 (이미지) | 단일 이미지 | 포즈 맵 + 바운딩 박스 | 여러 사람이 있는 사진 | -| 다중 인물 (비디오) | 비디오 | 프레임별 포즈 맵 + 바운딩 박스 | 비디오 포즈 추적 | -| OOD 이미지 → 포즈 | 단일 이미지 | 포즈 맵 | 스타일 전이 / 이미지→포즈 | -| OOD 비디오 → 포즈 맵 | 비디오 | 프레임별 포즈 맵 | 비디오→포즈 애니메이션 | +| Multi-Person (Image) | Single image | Pose map + BBoxes | Photos with multiple people | +| Multi-Person (Video) | Video | Per-frame pose map + BBoxes | Video pose tracking | +| OOD Image to Pose | Single image | Pose map | Style transfer / image-to-pose | +| OOD Video to Pose Map | Video | Per-frame pose map | Video-to-pose animation | + +### 1. Download Workflows + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find SDPose workflows under the Utility category. -### 1. 워크플로 다운로드 +### SDPose: Image Multi-Person Detection (`utility_sdpose_multi_person`) -ComfyUI를 최신 버전으로 업데이트한 다음, `Workflow` -> `Browse Templates`로 이동하여 Utility 카테고리에서 SDPose 워크플로를 찾으세요. +Upload an image to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose image multi-person detection workflow preview - - Comfy Cloud에서 실행 + + Open in Comfy Cloud - - JSON 다운로드 + + Download JSON or search "SDPose: Image Multi-Person Detection" in Template Library +**입력 자료** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 679 · `group_photo.png` + + + +
+ Input image +
+ +### SDPose: Video Multi-Person Detection (`utility_sdpose_multi_person_video`) + +Upload a video to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose video multi-person detection workflow preview + - - Comfy Cloud에서 실행 + + Open in Comfy Cloud - - JSON 다운로드 + + Download JSON or search "SDPose: Video Multi-Person Detection" in Template Library +**입력 자료** + +Upload this file to the matching `LoadVideo` node: + - - Comfy Cloud에서 실행 + + `LoadVideo` node 694 · `man_playing_violin.mp4` - - JSON 다운로드 + + +
+ +
+ +### SDPose-OOD: Image to Pose Map (`utility_sdpose_ood_image_to_pose`) + +Upload an image to extract pose keypoints and generate a corresponding pose map using the SDPose-OOD model. + +SDPose-OOD image to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Image to Pose Map" in Template Library +**입력 자료** + +Upload this file to the matching `LoadImage` node: + - - Comfy Cloud에서 실행 + + `LoadImage` node 667 · `dancer.png` - - JSON 다운로드 + + +**출력 예시** + +
+ Input image + SDPose-OOD image to pose map example output +
+ +### SDPose-OOD: Video to Pose Map (`utility_sdpose_ood_video_to_pose_map`) + +Upload a video to extract pose keypoints and generate a pose map. The workflow supports multiple person detection and uses an enhanced SDPose model for accurate whole-body feature extraction. + +SDPose-OOD video to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Video to Pose Map" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 679 · `man_in_the_rain.mp4` -### 2. 모델 다운로드 +**출력 예시** + +
+ + +
+ +### 2. Download Models -SDPose 및 RT-DETRv4 모델 체크포인트는 [Comfy-Org SDPose 모델 저장소](https://huggingface.co/Comfy-Org/SDPose)에 호스팅되어 있습니다. +The SDPose and RT-DETRv4 model checkpoints are hosted on the [Comfy-Org SDPose model repository](https://huggingface.co/Comfy-Org/SDPose). -**체크포인트** (SDPose 모델): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +**checkpoints** (SDPose model): +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) -**diffusion_models** (RT-DETRv4 감지기): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (권장) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (전체 정밀도, 용량이 큼) +**diffusion_models** (RT-DETRv4 detector): +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (recommended) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (full precision, larger) -다음과 같은 디렉토리 구조에 배치하세요: +Place them in the following directory structure: ``` 📂 ComfyUI/ @@ -105,44 +196,44 @@ SDPose 및 RT-DETRv4 모델 체크포인트는 [Comfy-Org SDPose 모델 저장 └── rt_detr_v4-x-hgnet_fp32.safetensors ``` -### 3. 워크플로 사용하기 +### 3. Using the Workflows -#### 다중 인물 (이미지) +#### Multi-Person (Image) -- **입력** — `Load Image` 노드를 통해 이미지를 불러옵니다. 한 명 이상의 사람이 있는 이미지(예: `group_photo.png`)를 사용하세요. -- **감지** — `Image to Pose Map (SDPose Multi-Person)` 서브그래프가 이미지를 처리하고 다음을 출력합니다: - - **IMAGE** — 이미지 위에 오버레이된 포즈 스켈레톤 시각화 - - **keypoints** — 원시 전신 키포인트 데이터 - - **bboxes** — 바운딩 박스 좌표 -- **드로잉 옵션** — 그릴 신체 부위를 구성합니다: - - `draw_body`, `draw_hands`, `draw_face`, `draw_feet` — 가시성 토글 - - `stick_width`, `face_point_size` — 시각적 스타일 조정 - - `score_threshold` — 키포인트를 표시하기 위한 최소 신뢰도 -- **감지 옵션**: - - `resize_type.longer_size` — 감지 전에 더 긴 차원을 조정 - - `max_detections` — 감지할 최대 인물 수 - - `detect_threshold` — 감지 신뢰도 임계값 - - `detect_class` — 감지할 객체 클래스(기본값: person) +- **Input** — Load an image via the `Load Image` node. Use an image with one or more people (example: `group_photo.png`). +- **Detection** — The `Image to Pose Map (SDPose Multi-Person)` subgraph processes the image and outputs: + - **IMAGE** — pose skeleton visualization overlaid on the image + - **keypoints** — raw whole-body keypoint data + - **bboxes** — bounding box coordinates +- **Drawing Options** — Configure which body parts to draw: + - `draw_body`, `draw_hands`, `draw_face`, `draw_feet` — toggle visibility + - `stick_width`, `face_point_size` — adjust visual style + - `score_threshold` — minimum confidence for displaying keypoints +- **Detection Options**: + - `resize_type.longer_size` — scale the longer dimension before detection + - `max_detections` — maximum number of people to detect + - `detect_threshold` — detection confidence threshold + - `detect_class` — object class to detect (default: person) -#### 다중 인물 (비디오) +#### Multi-Person (Video) -이미지 워크플로와 동일하지만 비디오 프레임을 순차적으로 처리합니다. 비디오 파일을 입력하려면 `Load Video`를, 결과를 내보내려면 `Save Video`를 사용하세요. +Same as the image workflow but processes video frames sequentially. Use `Load Video` to input a video file and `Save Video` to export the result. -#### OOD 이미지 → 포즈 +#### OOD Image to Pose -SDPose 모델을 사용하여 바운딩 박스 시각화 없이 이미지에서 깨끗한 포즈 맵을 생성합니다. 이는 한 이미지에서 스켈레톤 포즈를 추출하여 다른 이미지에 적용하는 스타일 전이에 유용합니다. +Uses the SDPose model to generate a clean pose map from an image, without bounding box visualization. This is useful for style transfer where you want to extract the skeleton pose from one image and apply it to another. -#### OOD 비디오 → 포즈 맵 +#### OOD Video to Pose Map -비디오에서 프레임별 포즈 맵을 생성합니다. 출력은 각 프레임에 추출된 포즈 스켈레톤이 포함된 비디오 파일로, 다운스트림 애니메이션이나 컨트롤넷 워크플로에 적합합니다. +Generates per-frame pose maps from a video. The output is a video file where each frame contains the extracted pose skeleton, suitable for downstream animation or ControlNet workflows. - - 이 워크플로는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인하여 워크플로를 사용자 정의하고 확장하는 방법을 알아보세요. + + These workflows use Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflows. -## 추가 참고 사항 +## Additional Notes -- **모델 디렉토리** — SDPose 체크포인트는 `models/checkpoints/`에, RT-DETRv4 감지기는 `models/diffusion_models/`에 넣습니다. -- **입력 이미지 예시** — `group_photo.png` 파일은 워크플로 템플릿의 `input/` 디렉토리에서 테스트용으로 제공됩니다. -- **키포인트 출력** — POSE_KEYPOINT 유형은 조건부 생성을 위해 포즈 데이터를 허용하는 다운스트림 노드에 연결할 수 있습니다. -- **업데이트 필수** — SDPose + RT-DETRv4 지원은 최근 ComfyUI 버전에서 사용 가능합니다. ComfyUI가 최신 상태인지 확인하세요. +- **Model directory** — the SDPose checkpoint goes in `models/checkpoints/`, and the RT-DETRv4 detector goes in `models/diffusion_models/` +- **Input image example** — the `group_photo.png` file is available in the workflow template's `input/` directory for testing +- **Keypoint output** — the POSE_KEYPOINT type can be connected to downstream nodes that accept pose data for conditional generation +- **Update required** — SDPose + RT-DETRv4 support is available in recent ComfyUI versions. Make sure your ComfyUI is up to date. diff --git a/ko/tutorials/utility/preprocessors.mdx b/ko/tutorials/utility/preprocessors.mdx index 810194baa..5c0ed28f2 100644 --- a/ko/tutorials/utility/preprocessors.mdx +++ b/ko/tutorials/utility/preprocessors.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Normals extraction": 65c791bb --- - ## 전처리기란 무엇인가요? @@ -27,82 +26,175 @@ translationBlockHashes: - 더 쉬운 디버깅 및 튜닝 - 보다 예측 가능한 이미지 및 비디오 결과 -## 깊이 추정 +## Depth estimation + +Depth estimation converts a flat image into a depth map representing relative distance within a scene. This structural signal is foundational for controlled generation, spatially aware edits, and relighting workflows. + +This workflow emphasizes: +- Clean, stable depth extraction +- Consistent normalization for downstream use +- Easy integration with ControlNet and image-edit pipelines + +Depth outputs can be reused across multiple passes, making it easier to iterate without re-running expensive upstream steps. + +### Video to Depth Map (`utility-depthAnything-v2-relative-video`) + +Convert a video to a temporally stable depth map. + +Video to Depth Map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Depth Map" in Template Library + + + +**입력 자료** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 6 · `roller_coaster.mp4` + + + +
+ +
+ +## Lineart conversion + +Lineart preprocessors distill an image down to its essential edges and contours, removing texture and color while preserving structure. + +This workflow is designed to: +- Produce clean, high-contrast lineart +- Minimize broken or noisy edges +- Provide reliable structural guidance for stylization and redraw workflows + +Lineart pairs especially well with depth and pose, offering strong structural constraints without overconstraining style. + +### Video to Lineart / Canny (`utility-lineart-video`) + +Convert a video to lineart or Canny edges for control processors. + +Video to Lineart workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Lineart / Canny" in Template Library + + + +**입력 자료** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 2 · `utility-lineart-video-input.mp4` + + + +
+ +
+ +## Pose detection + +Pose detection extracts body keypoints and skeletal structure from images, enabling precise control over human posture and movement. -깊이 추정은 평평한 이미지를 장면 내 상대적 거리를 나타내는 깊이 맵으로 변환합니다. 이 구조적 신호는 제어된 생성, 공간 인식 편집 및 재조명 워크플로우의 기초가 됩니다. +This workflow focuses on: +- Clear, readable pose outputs +- Stable keypoint detection suitable for reuse across frames +- Compatibility with pose-based ControlNet and animation pipelines -이 워크플로우는 다음을 중점적으로 다룹니다: -- 깔끔하고 안정적인 깊이 추출 -- 하위 파이프라인에서 사용할 수 있도록 일관된 정규화 -- ControlNet 및 이미지 편집 파이프라인과의 손쉬운 통합 +By isolating pose extraction into a dedicated workflow, pose data becomes easier to inspect, refine, and reuse. -깊이 출력은 여러 번의 패스에서 재사용할 수 있어 비용이 많이 드는 상위 단계를 다시 실행하지 않고도 쉽게 반복 작업을 수행할 수 있습니다. +### Video to Pose Map - OpenPose (`utility-openpose-video`) - - Comfy Cloud에서 실행 - +Convert a video to a temporally stable pose control map. - - JSON 다운로드 - +Video to Pose Map workflow preview -## 라인아트 변환 + + + Open in Comfy Cloud + + + Download JSON or search "Video to Pose Map - OpenPose" in Template Library + + -라인아트 전처리기는 이미지를 본질적인 가장자리와 윤곽으로 압축하여 질감과 색상을 제거하고 구조만 유지합니다. +**입력 자료** -이 워크플로우는 다음을 목표로 설계되었습니다: -- 깔끔하고 고대비 라인아트 생성 -- 끊어지거나 잡음이 많은 가장자리를 최소화 -- 스타일화 및 다시 그리기 워크플로우에 안정적인 구조적 가이드 제공 +Upload this file to the matching `VHS_LoadVideo` node: -라인아트는 깊이 및 자세와 특히 잘 어울리며, 스타일을 지나치게 제약하지 않으면서 강력한 구조적 제약을 제공합니다. + + + `VHS_LoadVideo` node 2 · `pose_input.mp4` + + - - Comfy Cloud에서 실행 - +
+ +
- - JSON 다운로드 - +## Normals extraction -## 자세 감지 +Normals estimation converts a flat image into a surface normal map—a per-pixel direction field that describes how each part of a surface is oriented (typically encoded as RGB). This signal is useful for relighting, material-aware stylization, and highly structured edits. -자세 감지는 이미지에서 몸의 관절점과 골격 구조를 추출하여 인간의 자세와 움직임을 정밀하게 제어할 수 있게 합니다. +This workflow emphasizes: +- Clean, stable normal extraction with minimal speckling +- Consistent orientation and normalization for reliable downstream use +- ControlNet-ready outputs for relighting, refinement, and structure-preserving edits +- Reuse across passes so you can iterate without re-running earlier steps -이 워크플로우는 다음에 중점을 둡니다: -- 명확하고 읽기 쉬운 자세 출력 -- 프레임 간 재사용이 가능한 안정적인 관절점 감지 -- 자세 기반 ControlNet 및 애니메이션 파이프라인과의 호환성 +Normal outputs can be used to: +- Drive relight/shading changes while preserving geometry +- Add a stronger 3D-like structure to stylization and redraw pipelines +- Improve consistency across frames when paired with pose/depth for animation work -자세 추출을 전용 워크플로우로 분리함으로써 자세 데이터를 더 쉽게 검사하고 개선하며 재사용할 수 있습니다. +### Video to Normal Map (`utility-normal_crafter-video`) - - Comfy Cloud에서 실행 - +Convert a video to a temporally stable normal map. - - JSON 다운로드 - +Video to Normal Map workflow preview -## 노멀 추출 + + + Open in Comfy Cloud + + + Download JSON or search "Video to Normal Map" in Template Library + + -노멀 추정은 평평한 이미지를 표면 법선 맵으로 변환합니다—각 픽셀마다 표면의 방향을 나타내는 벡터 필드로, 일반적으로 RGB로 인코딩됩니다. 이 신호는 재조명, 재료 인식 스타일화 및 고도로 구조화된 편집에 유용합니다. +**입력 자료** -이 워크플로우는 다음을 중점적으로 다룹니다: -- 깔끔하고 안정적인 노멀 추출, 잡티 최소화 -- 하위 파이프라인에서 사용할 수 있도록 일관된 방향 및 정규화 -- 재조명, 개선 및 구조 보존 편집을 위한 ControlNet 준비 출력 -- 패스 간 재사용으로 이전 단계를 다시 실행하지 않고도 반복 작업 가능 +Upload this file to the matching `VHS_LoadVideo` node: -노멀 출력은 다음에 사용될 수 있습니다: -- 지오메트리를 유지하면서 재조명/음영 변경을 유도 -- 스타일화 및 다시 그리기 파이프라인에 더욱 강력한 3D-like 구조 추가 -- 자세/깊이와 결합해 애니메이션 작업 시 프레임 간 일관성 향상 + + + `VHS_LoadVideo` node 3 · `normals_input.mp4` + + - - Comfy Cloud에서 실행 - +
+ +
- - JSON 다운로드 - diff --git a/ko/tutorials/utility/remove-background-birefnet.mdx b/ko/tutorials/utility/remove-background-birefnet.mdx index b23e07d8b..25cf5c516 100644 --- a/ko/tutorials/utility/remove-background-birefnet.mdx +++ b/ko/tutorials/utility/remove-background-birefnet.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": d07d5c84 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -30,27 +29,44 @@ BiRefNet은 ComfyUI에서 기본적으로 지원됩니다(PR [#12747](https://gi > **한계점:** 매우 복잡한 배경이나 배경과 동화되는 피사체는 다소 부정확한 마스크를 생성할 수 있습니다. 모델은 한 번에 하나의 이미지만 처리합니다. -## BiRefNet 배경 제거 워크플로우 +## BiRefNet Background Removal Workflow + +### 1. Download Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "BiRefNet: Remove Background" under the Utility category. + +### BiRefNet: Remove Background (`utility_birefnet_remove_background`) -### 1. 워크플로우 다운로드 +Upload an image with any background. Generate a version with the background removed and a precision segmentation mask. -ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플릿 둘러보기`로 이동해 유틸리티 카테고리 아래에서 "BiRefNet: 배경 제거"를 찾으세요. +BiRefNet remove background workflow preview - - 워크플로우 다운로드 + + + Open in Comfy Cloud + + Download JSON or search "BiRefNet: Remove Background" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: - - 클라우드에서 열기 + + + `LoadImage` node 17 · `the_lily_veil.png` + -### 2. 모델 다운로드 +### 2. Download Models -BiRefNet 모델은 [Comfy-Org BiRefNet 모델 저장소](https://huggingface.co/Comfy-Org/BiRefNet)에 호스팅되어 있습니다. +The BiRefNet model is hosted on the [Comfy-Org BiRefNet model repository](https://huggingface.co/Comfy-Org/BiRefNet). -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) -다음과 같은 디렉토리 구조에 배치하세요: +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -59,21 +75,21 @@ BiRefNet 모델은 [Comfy-Org BiRefNet 모델 저장소](https://huggingface.co/ └── birefnet.safetensors ``` -### 3. 워크플로우 사용법 +### 3. Using the Workflow -- **이미지** — `이미지 로드` 노드를 통해 이미지를 불러오세요(ComfyUI의 `input/` 폴더에 배치) -- `배경 제거(BiRefNet)` 서브그래프가 이미지를 처리하고 다음과 같은 출력을 제공합니다: - - **IMAGE** — 투명한 배경을 갖춘 결과 이미지(RGBA) - - **mask** — 추출된 전경 마스크 +- **Image** — Load an image via the `Load Image` node (place it in the ComfyUI `input/` folder) +- The `Remove Background (BiRefNet)` subgraph processes the image and outputs: + - **IMAGE** — the result with a transparent background (RGBA) + - **mask** — the extracted foreground mask -출력은 미리보기 및 다른 노드의 입력으로 사용해 합성, 추가 편집 또는 저장할 수 있습니다. +Outputs can be previewed and used as inputs to other nodes for compositing, further editing, or saving. - - 이 워크플로우는 모듈식 처리를 위해 서브그래프 노드를 사용합니다. 서브그래프 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 알아보세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 추가 참고사항 +## Additional Notes -- **모델 디렉토리** — 모델은 `checkpoints` 폴더가 아닌 `ComfyUI/models/background_removal/`에 배치해야 합니다. -- **업데이트 필요** — BiRefNet 지원은 최근 ComfyUI 버전에서 가능합니다. ComfyUI가 최신 상태인지 확인하세요. -- **RGBA 출력** — 투명한 배경 결과는 새로운 배경에 바로 합성하거나 하위 워크플로우에서 사용할 수 있습니다. +- **Model directory** — the model must be placed in `ComfyUI/models/background_removal/`, not the `checkpoints` folder +- **Update required** — BiRefNet support is available in recent ComfyUI versions. Make sure your ComfyUI is up to date. +- **RGBA output** — the transparent background result can be directly composited onto new backgrounds or used in downstream workflows diff --git a/ko/tutorials/utility/seedvr2.mdx b/ko/tutorials/utility/seedvr2.mdx index 39a239b88..87c3ec277 100644 --- a/ko/tutorials/utility/seedvr2.mdx +++ b/ko/tutorials/utility/seedvr2.mdx @@ -57,79 +57,137 @@ ComfyUI/ │ └── seedvr2_ema_vae_fp16.safetensors ``` -## 1. 이미지 업스케일 (3B INT8) +## Example Workflows -**기능:** 단일 이미지를 SeedVR2 3B INT8 모델을 사용하여 업스케일합니다. INT8 양자화 변형은 품질과 VRAM 사용량 간의 좋은 균형을 제공합니다. +### SeedVR2 3B Int8: Upscale Image (`utility_seedvr2_3b_int8_upscale_image`) - - JSON 다운로드 또는 템플릿 라이브러리에서 SeedVR2 3B Int8: Upscale Image 검색 +Upscale images using SeedVR2 3B Int8, a one-step diffusion-based video restoration model that produces high-quality results with improved temporal consistency. + +SeedVR2 3B Int8 upscale image workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Image" in Template Library + + +**입력 자료** + +Upload this file to the matching `LoadImage` node: - - 이 워크플로의 예제 입력 이미지 가져오기 + + + `LoadImage` node 1 · `watch_macro_shot.png` + -![SeedVR2 3B INT8 업스케일 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_3b_int8_upscale_image.png) +**출력 예시** -### 1.1 실행 단계 +
+ Input image + SeedVR2 3B Int8 upscale example output +
-1. 이미지를 `ComfyUI/input/` 폴더에 배치하고 `Load Image` 노드에서 선택합니다 -2. SeedVR2 모델 로더에서 `seedvr2_3b_int8_convrot.safetensors` 체크포인트를 선택합니다 -3. `Queue`를 클릭하거나 `Ctrl(cmd) + Enter`를 눌러 실행합니다 +### 1.1 Steps to Run ---- +1. Place your image in the `ComfyUI/input/` folder and select it in the `Load Image` node +2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run + +### SeedVR2 7B Int8: Upscale Image (`utility_seedvr2_7b_int8_upscale_image`) -## 2. 이미지 업스케일 (7B INT8) +Upscale images using SeedVR2 7B Int8, a one-step diffusion model that enhances resolution through adversarial training and adaptive window attention. -**기능:** SeedVR2 7B INT8 모델을 사용하여 단일 이미지를 업스케일합니다. 더 큰 7B 모델은 INT8 양자화를 통해 효율적인 VRAM 사용으로 더 높은 품질의 결과를 제공합니다. +SeedVR2 7B Int8 upscale image workflow preview - - JSON 다운로드 또는 템플릿 라이브러리에서 SeedVR2 7B Int8: Upscale Image 검색 + + + Open in Comfy Cloud + + Download JSON or search "SeedVR2 7B Int8: Upscale Image" in Template Library + + + +**입력 자료** - - 이 워크플로의 예제 입력 이미지 가져오기 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 1 · `indoor_portrait.png` + -![SeedVR2 7B INT8 업스케일 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_7b_int8_upscale_image.png) +**출력 예시** -### 2.1 실행 단계 +
+ Input image + SeedVR2 7B Int8 upscale example output +
-1. 이미지를 `ComfyUI/input/` 폴더에 배치하고 `Load Image` 노드에서 선택합니다 -2. SeedVR2 모델 로더에서 `seedvr2_7b_int8_convrot.safetensors` 체크포인트를 선택합니다 -3. `Queue`를 클릭하거나 `Ctrl(cmd) + Enter`를 눌러 실행합니다 +### 2.1 Steps to Run ---- +1. Place your image in the `ComfyUI/input/` folder and select it in the `Load Image` node +2. Select the `seedvr2_7b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run -## 3. 비디오 업스케일 (3B INT8) +### SeedVR2 3B Int8: Upscale Video (`utility_seedvr2_3b_int8_upscale_video`) -**기능:** SeedVR2 3B INT8 모델을 사용하여 비디오를 업스케일합니다. 워크플로는 프레임 간 시간적 일관성을 유지하면서 해상도를 향상시킵니다. 오래된 영상 복원 및 저해상도 비디오 업스케일에 적합합니다. +Upscale and restore video footage using SeedVR2 3B Int8, a one-step diffusion model that enhances resolution while maintaining temporal consistency across frames. - - JSON 다운로드 또는 템플릿 라이브러리에서 SeedVR2 3B Int8: Upscale Video 검색 +SeedVR2 3B Int8 upscale video workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Video" in Template Library + - - 이 워크플로의 예제 입력 비디오 가져오기 +**입력 자료** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 73 · `grainy_perfume_shot_crf32.mp4` + + +**출력 예시** + +
+ + +
-### 3.1 실행 단계 +### 3.1 Steps to Run -1. 비디오를 `ComfyUI/input/` 폴더에 넣고 `Load Video` 노드에서 선택하세요 -2. SeedVR2 모델 로더에서 `seedvr2_3b_int8_convrot.safetensors` 체크포인트를 선택하세요 -3. `Queue`를 클릭하거나 `Ctrl(cmd) + Enter`를 눌러 실행하세요 +1. Place your video in the `ComfyUI/input/` folder and select it in the `Load Video` node +2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run -### 성능 +### Performance -더 높은 대상 해상도는 더 많은 처리 시간이 필요합니다. INT8 변형은 FP16에 비해 VRAM 사용량을 줄이고 효율적인 추론을 제공합니다. +Higher target resolutions require more processing time. The INT8 variant provides efficient inference with reduced VRAM usage compared to FP16. --- -## 커뮤니티 리소스 +## Community Resources -- [SeedVR2 프로젝트 페이지](https://iceclear.github.io/projects/seedvr2/): 공식 웹사이트 -- [ByteDance SeedVR 코드베이스 (GitHub)](https://github.com/ByteDance-Seed/SeedVR): 원본 연구 코드 및 논문 -- [Comfy-Org/SeedVR2 (HuggingFace)](https://huggingface.co/Comfy-Org/SeedVR2): ComfyUI 모델 가중치 -- [ByteDance-Seed/SeedVR2-3B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-3B): 원본 3B 모델 가중치 -- [ByteDance-Seed/SeedVR2-7B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-7B): 원본 7B 모델 가중치 -- [논문 (arXiv)](https://arxiv.org/abs/2506.05301) +- [SeedVR2 Project Page](https://iceclear.github.io/projects/seedvr2/): Official project website +- [ByteDance SeedVR Codebase (GitHub)](https://github.com/ByteDance-Seed/SeedVR): Original research code and paper +- [Comfy-Org/SeedVR2 (HuggingFace)](https://huggingface.co/Comfy-Org/SeedVR2): Official ComfyUI model weights +- [ByteDance-Seed/SeedVR2-3B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-3B): Original 3B model weights +- [ByteDance-Seed/SeedVR2-7B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-7B): Original 7B model weights +- [Paper (arXiv)](https://arxiv.org/abs/2506.05301) diff --git a/ko/tutorials/utility/video-segment-sam3.mdx b/ko/tutorials/utility/video-segment-sam3.mdx index 8dfa7da68..cf597bcd8 100644 --- a/ko/tutorials/utility/video-segment-sam3.mdx +++ b/ko/tutorials/utility/video-segment-sam3.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": c38c57a0 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -38,39 +37,79 @@ SAM 3.1은 텍스트 프롬프트를 기반으로 동영상 프레임 전체에 > **제한사항:** 모델의 텍스트 프롬프트 최대 토큰 수는 32개입니다. 최상의 결과를 얻으려면 프롬프트를 짧고 대상 객체에 맞게 구체적으로 작성하세요. -## SAM 3.1 분할 워크플로우 +## SAM 3.1 Segment Workflows + +### 1. Download Workflow -### 1. 워크플로우 다운로드 +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find the SAM 3.1 workflows under the Utility category. -ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플릿 찾아보기`로 이동해 유틸리티 카테고리에서 SAM 3.1 워크플로우를 찾으세요. +### SAM3: Video Segmentation (`utility_video_segment_sam3`) -**동영상 분할:** +Use the SAM3 model to segment the main subject or content from a video, isolating specific objects or regions. - - 동영상 워크플로우 다운로드 +SAM3 video segmentation workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SAM3: Video Segmentation" in Template Library + - - 클라우드에서 열기 +**입력 자료** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 115 · `drinking_unicorn.mp4` + + +**출력 예시** + + + +### SAM3: Image Segmentation (`utility_image_segment_sam3`) + +Use the SAM3 model to segment the main subject or content from a photo or image, isolating specific objects or regions. -**이미지 분할:** +SAM3 image segmentation workflow preview - - 이미지 워크플로우 다운로드 + + + Open in Comfy Cloud + + Download JSON or search "SAM3: Image Segmentation" in Template Library + + + +**입력 자료** - - 클라우드에서 열기 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 79 · `neon_guitarist.png` + + +
+ Input image +
-### 2. 모델 다운로드 +### 2. Download Models -SAM 3.1 모델은 [Comfy-Org SAM 3.1 모델 저장소](https://huggingface.co/Comfy-Org/sam3.1)에 호스팅되어 있습니다. +The SAM 3.1 model is hosted on the [Comfy-Org SAM 3.1 model repository](https://huggingface.co/Comfy-Org/sam3.1). -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) -다음과 같은 디렉터리 구조에 배치하세요: +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -79,36 +118,36 @@ SAM 3.1 모델은 [Comfy-Org SAM 3.1 모델 저장소](https://huggingface.co/Co └── sam3.1_multiplex_fp16.safetensors ``` -### 3. 워크플로우 사용하기 +### 3. Using the Workflows -**이미지 분할:** +**Image Segmentation:** -- **이미지** — `이미지 로드` 노드를 통해 이미지를 불러오세요(ComfyUI의 `input/` 폴더에 넣으세요) -- **객체 프롬프트** — 분할할 객체에 대한 짧은 텍스트 설명, 예: `person`, `car`, `cat` -- 출력은 이미지에 적용된 마스크이며, RGBA 미리보기로 분할 결과를 확인할 수 있습니다. +- **Image** — Load an image via the `Load Image` node (place it in the ComfyUI `input/` folder) +- **Object Prompt** — A short text description of the object(s) to segment, e.g. `person`, `car`, `cat` +- The output is a mask applied to the image, with an RGBA preview showing the segmentation result -**동영상 분할:** +**Video Segmentation:** -- **동영상** — `동영상 로드` 노드를 통해 동영상을 불러오세요 -- **객체 프롬프트** — 이미지와 동일하게, 프레임 간 추적 및 분할할 내용을 설명하는 짧은 텍스트 프롬프트입니다. -- 출력은 각 프레임별 마스크와 경계 상자를 제공합니다. +- **Video** — Load a video via the `Load Video` node +- **Object Prompt** — Same as image, a short text prompt describing what to track and segment across frames +- The output provides masks and bounding boxes for each frame -**프롬프트 형식:** +**Prompt format:** -| 프롬프트 | 역할 | +| Prompt | Role | |--------|------| -| SAM3 객체 프롬프트 | 분할할 **무엇**인지에 대한 짧은 설명. 최대 32토큰. | +| SAM3 object prompt | Short description of **what** to segment. Max 32 tokens. | -여러 개체를 별도로 프롬프트하려면 쉼표로 구분하고, `:N`을 사용해 각 프롬프트당 감지 가능한 객체 수를 지정하세요: +To prompt multiple subjects separately, separate with commas and use `:N` to specify the max amount of objects detected per prompt: `eye:2, window panels:4` - - 이 워크플로우는 서브그래프 노드를 사용해 모듈식 처리를 수행합니다. 서브그래프 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 알아보세요. + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 추가 참고사항 +## Additional Notes -- **프롬프트는 짧고 구체적으로 유지하세요** — 모델의 프롬프트 한 개당 토큰 제한은 32개입니다. -- **다중 객체 탐지** — 쉼표로 서로 다른 객체 유형을 구분하고, `:N`을 사용해 각 유형별 탐지 수를 제한하세요. -- **분할 마스크** — 출력 마스크는 다른 워크플로우(예: 인페인팅, 배경 제거)의 입력으로 사용할 수 있습니다. -- **업데이트 필요** — SAM 3.1 지원을 받으려면 ComfyUI를 최신 버전으로 업데이트하세요. +- **Keep prompts short and specific** — the model has a 32-token limit per prompt +- **Multi-object detection** — use commas to separate different object types, and `:N` to cap detections per type +- **Segmentation masks** — the output mask can be used as input to other workflows (e.g., inpainting, background removal) +- **Update required** — make sure ComfyUI is updated to the latest version to access SAM 3.1 support diff --git a/ko/tutorials/utility/video-upscale.mdx b/ko/tutorials/utility/video-upscale.mdx index a88aa8147..090ee420e 100644 --- a/ko/tutorials/utility/video-upscale.mdx +++ b/ko/tutorials/utility/video-upscale.mdx @@ -15,7 +15,6 @@ translationBlockHashes: "Tips": a5dcf8f3 --- - 이 가이드에서는 ComfyUI에서의 비디오 업스케일링 워크플로우를 다룹니다. 여기에는 다양한 사용 사례에 맞는 로컬 모델과 파트너 노드 옵션이 포함됩니다. diff --git a/ko/tutorials/utility/void-video-inpainting.mdx b/ko/tutorials/utility/void-video-inpainting.mdx index e8b63bf01..b11364cf0 100644 --- a/ko/tutorials/utility/void-video-inpainting.mdx +++ b/ko/tutorials/utility/void-video-inpainting.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": a3d04a46 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx" @@ -45,44 +44,61 @@ VOID는 ComfyUI에서 기본적으로 지원되며([PR #13403](https://github.co > **한계:** 불분명한 마스크, 혼란스러운 움직임, 또는 프레임을 지배하는 대상은 여전히 최적의 결과를 내지 못할 수 있습니다—세그먼테이션 자체가 잘못된 경우 이를 고칠 수는 없습니다. -## VOID 비디오 인페인팅 워크플로우 +## VOID Video Inpainting Workflow + +### 1. Download Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "VOID: Video Inpainting" under the Utility category. + +### VOID: Video Inpainting (`utility_void_video_inpainting`) -### 1. 워크플로우 다운로드 +Upload a video and mask the object you want to remove. Generate a clean video with the object and its physical interactions deleted. -ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플릿 둘러보기`로 이동해 유틸리티 카테고리에서 "VOID: 비디오 인페인팅"을 찾으세요. +VOID video inpainting workflow preview - - 워크플로우 다운로드 + + + Open in Comfy Cloud + + Download JSON or search "VOID: Video Inpainting" in Template Library + + + +**입력 자료** + +Upload this file to the matching `LoadVideo` node: - - 클라우드에서 열기 + + + `LoadVideo` node 4 · `snowboarder.mp4` + -### 2. 모델 다운로드 +### 2. Download Models -모델은 모두 [Comfy-Org VOID 모델 저장소](https://huggingface.co/Comfy-Org/void-model)에 호스팅됩니다. +All models are hosted on the [Comfy-Org VOID model repository](https://huggingface.co/Comfy-Org/void-model). -**확산 모델** — 핵심 2회 통과 인페인팅 모델: +**Diffusion Models** — the core two-pass inpainting model: -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 정밀화 통과, 더 나은 시간적 안정성 -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 기본 통과 +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — Refinement pass, better temporal stability +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — Primary pass **VAE:** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) -**광학 흐름:** +**Optical Flow:** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) -**SAM3 체크포인트** — 세그먼테이션용: +**SAM3 Checkpoint** — for segmentation: -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) -**텍스트 인코더:** +**Text Encoder:** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ @@ -100,34 +116,34 @@ ComfyUI를 최신 버전으로 업데이트한 후, `워크플로우` -> `템플 │ └── void_pass1.safetensors ``` -### 3. 워크플로우 사용하기 +### 3. Using the Workflow -**입력:** +**Inputs:** -- **소스 비디오** — `비디오 로드` 노드를 통해 비디오를 로드하세요(ComfyUI의 `input/` 폴더에 넣으세요) -- **양성 프롬프트(인페인팅 채우기)** — 제거된 이후의 장면을 설명하세요. 무엇이 제거되었는지가 아니라 남아있는 부분과 그 모습에 집중하세요. - - 예시: `empty kitchen counter, daylight, tiles visible` -- **음성 프롬프트** — 선택적 아티팩트 방지 목록; 비워둘 수도 있습니다 -- **SAM3 객체 프롬프트** — 마스크할 **무엇인지**에 대한 짧은 레이블입니다. SAM3는 의미적 이해를 통해 대상 객체의 세그멘테이션 마스크를 생성합니다. - - 예시: `person in blue jacket`, `red cup on table` - - SAM3 프롬프트의 최대 토큰 수는 **32**입니다. 여러 대상을 별도로 프롬프트하려면 쉼표로 구분하고, 각 프롬프트당 감지되는 최대 객체 수를 지정하려면 `:N`을 사용하세요: `eye:2, window panels:4` +- **Source video** — Load a video via the `Load Video` node (place it in the ComfyUI `input/` folder) +- **Positive prompt (inpaint fill)** — Describe the scene **after** removal. Focus on what remains and how it looks, not on what was removed + - Example: `empty kitchen counter, daylight, tiles visible` +- **Negative prompt** — Optional anti-artifact list; can be left empty +- **SAM3 object prompt** — A short label for **what** to mask out. SAM3 uses semantic understanding to create a segmentation mask for the target object. + - Example: `person in blue jacket`, `red cup on table` + - Max tokens for SAM3 prompts is **32**. To prompt multiple subjects separately, separate with commas and use `:N` to specify the max objects detected per prompt: `eye:2, window panels:4` -**모드:** +**Modes:** -| 프롬프트 | 역할 | +| Prompt | Role | |--------|------| -| SAM3 객체 | **무엇**을 제거할지 (SAM3는 의미적 세그멘테이션을 통해 마스크를 생성합니다) | -| 양성(인페인팅) | 시간에 걸쳐 빈 공간을 **어떻게** 채울지 | +| SAM3 object | **What** is removed (SAM3 creates the mask via semantic segmentation) | +| Positive (inpaint) | **How** the hole is filled across time | -시간적 안정성이 중요한 긴 클립이나 질감이 있는 배경에서는 **2회 통과**(정밀화 통과)를 사용하세요. **1회 통과**만으로도 빠르지만, 더 많은 떨림이 나타날 수 있습니다. +Use **Pass 2** (refinement pass) for longer clips or textured backgrounds where temporal stability matters. **Pass 1** alone is faster but may show more jitter. - - 이 워크플로우는 서브그래프 노드를 사용해 모듈식 비디오 처리를 수행합니다. 서브그래프 문서를 확인해 워크플로우를 맞춤화하고 확장하는 방법을 알아보세요. + + This workflow uses Subgraph nodes for modular video processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 추가 참고사항 +## Additional Notes -- **마스크 품질이 중요합니다** — 대상 객체 주변에 깔끔하고 딱 맞는 마스크를 사용하면 최상의 결과를 얻을 수 있습니다. -- **프롬프트 작성 팁** — 제거된 이후의 장면이 자연스럽게 보이는 모습을 묘사하세요. 제거 자체가 아니라요. -- **음성 프롬프트 사용**은 반복적인 결함(워터마크, 흐릿함, 과다한 팔다리)이 보일 때만 사용하세요. -- **2회 통과 워크플로우** — 템플릿은 1회 통과 후 자동으로 2회 통과를 실행합니다; 테스트 중 더 빠른 반복을 위해 1회 통과만 실행할 수도 있습니다. +- **Mask quality matters** — a clean, tight mask around the target object produces the best results +- **Prompt writing tip** — describe the scene as it should appear _naturally_ after removal, not the removal itself +- **Use negative prompt** only when you see repeating defects (watermarks, blur, extra limbs) +- **Two-pass workflow** — the template runs Pass 1 then Pass 2 automatically; you can also run just Pass 1 for faster iterations during testing diff --git a/ko/tutorials/video/bytedance/bernini-r.mdx b/ko/tutorials/video/bytedance/bernini-r.mdx index 649485dbb..073af8240 100644 --- a/ko/tutorials/video/bytedance/bernini-r.mdx +++ b/ko/tutorials/video/bytedance/bernini-r.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Community Resources": 9b4a0aaf --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' # ComfyUI Bernini-R 소개 @@ -82,19 +80,27 @@ ComfyUI/ **기능 설명:** 조명이 일치하는 편집된 이미지를 생성하고, 전후 비교를 나란히 표시합니다. 인물 및 제품 재조명, 사진 컬렉션의 일관된 조명, 전자상거래 카탈로그 촬영에 적합합니다. +Bernini-R 이미지 편집 워크플로 미리보기 + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Bernini-R" 검색 - Comfy Cloud에서 열기 + + JSON 다운로드 또는 템플릿 라이브러리에서 "Bernini-R" 검색 + -
- Bernini-R 이미지 편집 출력 - Bernini-R 이미지 편집 비교 -
+**입력 자료** + + + + 기본 입력 이미지를 다운로드하거나 자체 이미지를 사용하세요. + + + 기본 참조 이미지를 다운로드하거나 자체 이미지를 사용하세요. + + ### 실행 단계 @@ -114,16 +120,27 @@ ComfyUI/ **기능 설명:** Bernini-R로 일관된 재조명이 적용된 편집 비디오를 생성합니다. 소스 비디오, 선택적 참조 이미지 또는 참조 비디오를 연결하고, 작업 유형을 선택한 후 프롬프트를 작성하여 실행합니다. +Bernini-R 비디오 편집 워크플로 미리보기 + - - JSON 다운로드 또는 템플릿 라이브러리에서 "Bernini-R" 검색 - Comfy Cloud에서 열기 + + JSON 다운로드 또는 템플릿 라이브러리에서 "Bernini-R" 검색 + -![Bernini-R 비디오 편집 미리보기](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_bernini_r_video_editing-1.webp) +**입력 자료** + + + + 기본 입력 비디오를 다운로드하거나 자체 비디오를 사용하세요. + + + 기본 참조 이미지를 다운로드하거나 자체 이미지를 사용하세요. + + ### 실행 단계 diff --git a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 2bcdd7102..40572f4bb 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Super-resolution upscaler": 89338328 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; [HunyuanVideo 1.5](https://github.com/Tencent/HunyuanVideo)는 텐센트의 Hunyuan팀이 개발한 경량 83억 파라미터 모델입니다. 이 모델은 소비자용 GPU(24GB VRAM)에서 플래그십 수준의 비디오 생성 성능을 제공하며, 품질을 저하시키지 않으면서 진입 장벽을 대폭 낮췄습니다. diff --git a/ko/tutorials/video/hunyuan/hunyuan-video.mdx b/ko/tutorials/video/hunyuan/hunyuan-video.mdx index ad5f0cf7a..a055e3cbb 100644 --- a/ko/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/ko/tutorials/video/hunyuan/hunyuan-video.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Try It Yourself": 076fb43f --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; + + Comfy Cloud에서 열기 + ## 프롬프트 작성 팁 diff --git a/ko/tutorials/video/ltxv.mdx b/ko/tutorials/video/ltxv.mdx index 2c4a28824..74bcfe74a 100644 --- a/ko/tutorials/video/ltxv.mdx +++ b/ko/tutorials/video/ltxv.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Requirements": bfab2742 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; [LTX-Video](https://huggingface.co/Lightricks/LTX-Video)는 lightricks가 개발한 매우 효율적인 비디오 모델입니다. 이 모델의 중요한 점은 길고 상세한 프롬프트를 제공하는 것입니다. @@ -33,6 +32,8 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; 첫 번째 [프레임 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/ltxv/i2v/girl1.png)를 통해 비디오를 제어할 수 있습니다. +Workflow preview + Comfy Cloud에서 열기 JSON 다운로드 또는 템플릿 라이브러리에서 "LTX-Video" 검색 @@ -46,7 +47,20 @@ import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## 텍스트에서 비디오로 -LTX-Video 텍스트에서 비디오로 +### LTXV 텍스트 기반 비디오 생성（`ltxv_text_to_video`） + +텍스트 프롬프트에서 비디오를 생성합니다. + +LTXV 텍스트 기반 비디오 생성 워크플로 미리보기 + + + + Comfy Cloud에서 열기 + + + JSON 다운로드 또는 템플릿 라이브러리에서 "LTXV Text to Video" 검색 + + 비디오를 직접 ComfyUI로 드래그하여 워크플로우를 실행하세요. diff --git a/ko/tutorials/video/wan/fun-camera.mdx b/ko/tutorials/video/wan/fun-camera.mdx index ac4bb8765..137fe77c4 100644 --- a/ko/tutorials/video/wan/fun-camera.mdx +++ b/ko/tutorials/video/wan/fun-camera.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Performance Reference": 2dc2f8f0 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## Wan2.1 Fun Camera 소개 @@ -41,6 +39,10 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' 1.3B 또는 14B 중 하나를 선택하세요: +Workflow preview + +Workflow preview + Wan2.1 Fun Camera 1.3B 디퓨전 모델 diff --git a/ko/tutorials/video/wan/fun-control.mdx b/ko/tutorials/video/wan/fun-control.mdx index b72c3d8bc..c4610b0eb 100644 --- a/ko/tutorials/video/wan/fun-control.mdx +++ b/ko/tutorials/video/wan/fun-control.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Usage Tips": 9e74c201 --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## Wan2.1-Fun-Control 소개 diff --git a/ko/tutorials/video/wan/fun-inp.mdx b/ko/tutorials/video/wan/fun-inp.mdx index 319f8544a..4f3baeac7 100644 --- a/ko/tutorials/video/wan/fun-inp.mdx +++ b/ko/tutorials/video/wan/fun-inp.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Other Wan2.1 Fun InP or video-related custom node packages": 4c96c810 --- - - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; ## Wan2.1-Fun-InP 소개 diff --git a/ko/tutorials/video/wan/wan-alpha.mdx b/ko/tutorials/video/wan/wan-alpha.mdx index 55423d5ca..36e8f4794 100644 --- a/ko/tutorials/video/wan/wan-alpha.mdx +++ b/ko/tutorials/video/wan/wan-alpha.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Wan-Alpha Text-to-Video Workflow (14B)": 32a5a973 --- - import UpdateReminder from "/snippets/ko/tutorials/update-reminder.mdx"; Wan-Alpha는 알파 채널 투명도가 적용된 고품질 동영상을 생성하는 특수한 텍스트 기반 비디오 생성 모델입니다. Wan2.1-14B-T2V 베이스 모델을 기반으로 하며, 투명한 배경과 반투명 객체를 포함한 동영상을 만들어 합성 작업에 완벽합니다. diff --git a/ko/tutorials/video/wan/wan-ati.mdx b/ko/tutorials/video/wan/wan-ati.mdx index f5b9de21f..0e2cc75ef 100644 --- a/ko/tutorials/video/wan/wan-ati.mdx +++ b/ko/tutorials/video/wan/wan-ati.mdx @@ -10,11 +10,8 @@ translationBlockHashes: "WAN ATI Trajectory Control Workflow Example": f704f2c5 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - **ATI (Any Trajectory Instruction)**는 ByteDance 팀에서 제안한 제어 가능한 비디오 생성 프레임워크입니다. ATI는 Wan2.1을 기반으로 구현되었으며, 임의의 트래젝터리 명령어를 통해 비디오 내 객체, 국소 영역 및 카메라 모션을 통합적으로 제어할 수 있습니다. 프로젝트 URL: [https://github.com/bytedance/ATI](https://github.com/bytedance/ATI) @@ -26,7 +23,6 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - **Wan2.1 호환**: 공식 Wan2.1 구현을 기반으로 하며, 환경과 모델 구조와 호환됩니다. - **풍부한 시각화 도구**: 입력 트래젝터리, 출력 비디오, 트래젝터리 오버레이의 시각화를 지원합니다. - ## WAN ATI 트래젝터리 제어 워크플로우 예시 diff --git a/ko/tutorials/video/wan/wan-causal-forcing.mdx b/ko/tutorials/video/wan/wan-causal-forcing.mdx index 6ff065a2e..de46618f6 100644 --- a/ko/tutorials/video/wan/wan-causal-forcing.mdx +++ b/ko/tutorials/video/wan/wan-causal-forcing.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Model downloads": c6ce079e --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Causal Forcing**은 추론 중 **반복적 조건화**를 적용하는 비디오 생성 기술로, 생성된 각 프레임이 모델에 다시 입력되어 다음 프레임을 예측합니다. 이를 통해 단일 시작 이미지에서 **1~4회의 추론 단계**만으로 부드럽고 시간적으로 일관된 비디오를 생성할 수 있습니다. diff --git a/ko/tutorials/video/wan/wan-dancer.mdx b/ko/tutorials/video/wan/wan-dancer.mdx index 1f134c5d7..f5e7d4ad6 100644 --- a/ko/tutorials/video/wan/wan-dancer.mdx +++ b/ko/tutorials/video/wan/wan-dancer.mdx @@ -39,6 +39,8 @@ Wan Dancer 워크플로는 두 가지 입력(캐릭터의 참조 이미지와 ComfyUI를 최신 버전으로 업데이트한 후 워크플로 파일을 다운로드하여 ComfyUI로 드래그하거나, 템플릿 라이브러리에서 `워크플로` → `템플릿 탐색` → `비디오`로 이동하여 "Wan Dancer"를 찾으세요. +Workflow preview + Comfy Cloud에서 열기 diff --git a/ko/tutorials/video/wan/wan-flf.mdx b/ko/tutorials/video/wan/wan-flf.mdx index 76ae0d4d9..3f7f18581 100644 --- a/ko/tutorials/video/wan/wan-flf.mdx +++ b/ko/tutorials/video/wan/wan-flf.mdx @@ -45,7 +45,6 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 ![Wan2.1 FLF2V 720P f16 워크플로우](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1_flf2v/wan2.1_flf2v_720_f16.webp) - 아래 두 장의 이미지를 다운로드해 동영상의 시작 프레임과 종료 프레임으로 사용하겠습니다. ![start_image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/wan2.1_flf2v/input/start_image.png) @@ -75,7 +74,6 @@ Wan FLF2V (첫 번째와 마지막 프레임 동영상 생성)는 알리바바 **CLIP Vision** - [clip_vision_h.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors?download=true) - 파일 저장 위치 ``` ComfyUI/ diff --git a/ko/tutorials/video/wan/wan-move.mdx b/ko/tutorials/video/wan/wan-move.mdx index d3630d0cb..7bf5ea8ca 100644 --- a/ko/tutorials/video/wan/wan-move.mdx +++ b/ko/tutorials/video/wan/wan-move.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Model links": 8e0ef166 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Wan-Move**는 알리바바의 퉁이 랩에서 개발한 모션 제어가 가능한 비디오 생성 프레임워크입니다. 사용자는 입력 이미지 위에 점의 궤적을 지정함으로써 생성된 비디오에서 객체의 움직임을 제어할 수 있어 이미지에서 비디오 생성을 더욱 정밀하고 제어 가능하게 만듭니다. diff --git a/ko/tutorials/video/wan/wan-video.mdx b/ko/tutorials/video/wan/wan-video.mdx index 127b4bcb1..db40b08a2 100644 --- a/ko/tutorials/video/wan/wan-video.mdx +++ b/ko/tutorials/video/wan/wan-video.mdx @@ -12,9 +12,6 @@ translationBlockHashes: "Wan2.1 Image-to-Video Workflow (14B)": dfb19590 --- - - - Wan2.1 Video 시리즈는 2025년 2월 알리바바가 [Apache 2.0 라이선스](https://github.com/Wan-Video/Wan2.1?tab=Apache-2.0-1-ov-file)로 오픈소스로 공개한 동영상 생성 모델입니다. 두 가지 버전을 제공합니다: - 14B (140억 파라미터) diff --git a/ko/tutorials/video/wan/wan2-2-animate.mdx b/ko/tutorials/video/wan/wan2-2-animate.mdx index a4581fa4c..c398b9f71 100644 --- a/ko/tutorials/video/wan/wan2-2-animate.mdx +++ b/ko/tutorials/video/wan/wan2-2-animate.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "About Wan2.2 Animate workflow": ea4c1d9a --- - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 위한 통합 프레임워크입니다. @@ -41,7 +40,6 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 - ## Wan2.2 Animate 워크플로우 소개 이 문서에서는 두 가지 워크플로우를 제공합니다: @@ -55,12 +53,13 @@ Wan-Animate는 WAN 팀에서 개발한 캐릭터 애니메이션 및 교체를 다음 워크플로우 파일을 다운로드해 ComfyUI로 끌어다 놓으면 워크플로우가 로드됩니다. +Workflow preview + Comfy Cloud에서 열기 JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Animate" 검색 - 아래 자료를 입력으로 다운로드하세요: **참조 이미지:** diff --git a/ko/tutorials/video/wan/wan2-2-fun-control.mdx b/ko/tutorials/video/wan/wan2-2-fun-control.mdx index ac9c7507f..ec3ea38b8 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-control.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Wan2.2 Fun Control Video Generation Workflow Example": cbbb7456 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Wan2.2-Fun-Control**은 알리바바 PAI팀이 출시한 차세대 비디오 생성 및 제어 모델입니다. 혁신적인 Control Codes 메커니즘과 딥러닝, 다중 모달 조건 입력을 결합해 미리 설정된 제어 조건에 부합하는 고품질 비디오를 생성할 수 있습니다. 이 모델은 **Apache 2.0 라이선스**로 공개되며 상업적 사용도 지원합니다. @@ -40,7 +38,6 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' - 이 워크플로우는 두 가지 버전을 제공합니다: 1. lightx2v의 [Wan2.2-Lightning](https://huggingface.co/lightx2v/Wan2.2-Lightning) 4단계 LoRA를 사용한 버전: 비디오 역동성에 약간의 손실이 있을 수 있지만 속도가 더 빠릅니다. 2. 가속화되지 않은 fp8_scaled 버전 @@ -64,6 +61,8 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/wan2.2_14B_fun_inp.mp4" > +Workflow preview + Comfy Cloud에서 열기 JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Control" 검색 @@ -114,7 +113,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. 워크플로우 안내 ![Wan2.2 펀 컨트롤 워크플로우 단계](/images/tutorial/video/wan/wan2_2/wan_2.2_14b_fun_control.jpg) diff --git a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx index 4f035f807..0051e1ffc 100644 --- a/ko/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/ko/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 692b42a8 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **Wan2.2-Fun-Inp**는 알리바바 PAI 팀이 출시한 시작-끝 프레임 제어 동영상 생성 모델입니다. 이 모델은 **시작 및 끝 프레임 이미지**를 입력해 중간 전환 동영상을 생성할 수 있어 크리에이터들이 더욱 창의적인 제어를 할 수 있도록 지원합니다. 이 모델은 **Apache 2.0 라이선스**로 배포되며 상업적 사용도 가능합니다. @@ -62,6 +60,7 @@ ComfyUI를 최신 버전으로 업데이트한 후, 메뉴 `워크플로우` -> 또는 ComfyUI를 최신 버전으로 업데이트한 후 아래 워크플로우를 다운로드해 ComfyUI에 드래그하여 로드하세요. +Wan2.2 Fun Inp 워크플로 미리보기 JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 Fun Inp" 검색 diff --git a/ko/tutorials/video/wan/wan2-2-s2v.mdx b/ko/tutorials/video/wan/wan2-2-s2v.mdx index 2d7079bec..65e211680 100644 --- a/ko/tutorials/video/wan/wan2-2-s2v.mdx +++ b/ko/tutorials/video/wan/wan2-2-s2v.mdx @@ -20,7 +20,6 @@ Wan2.2-S2V, 고급 오디오 기반 비디오 생성 모델이 이제 ComfyUI에 Wan2.2 S2V 코드: [GitHub](https://github.com/aigc-apps/VideoX-Fun) Wan2.2 S2V 모델: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) - ## Wan2.2 S2V ComfyUI 네이티브 워크플로우 @@ -35,16 +34,16 @@ Wan2.2 S2V 모델: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > +Workflow preview + Comfy Cloud에서 열기 JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 S2V" 검색 - 다음 이미지와 오디오를 입력으로 다운로드하세요: ![입력](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/input.jpg) - 기본 입력 이미지를 다운로드하거나 자신의 이미지를 사용하세요 @@ -69,7 +68,6 @@ Wan2.2 S2V 모델: [Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) **text_encoders** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - ``` ComfyUI/ ├───📂 models/ @@ -84,7 +82,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. 워크플로우 지침 ![워크플로우 지침](/images/tutorial/video/wan/wan_2.2_14b_s2v.jpg) diff --git a/ko/tutorials/video/wan/wan2_2.mdx b/ko/tutorials/video/wan/wan2_2.mdx index 7c85fffa8..1e27726d1 100644 --- a/ko/tutorials/video/wan/wan2_2.mdx +++ b/ko/tutorials/video/wan/wan2_2.mdx @@ -79,7 +79,6 @@ ComfyUI Wan2.2 사용법에 대해 라이브 스트림을 진행했으니, 이 allowFullScreen > - 이 튜토리얼에서는 [🤗 Comfy-Org/Wan_2.2_ComfyUI_Repackaged](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged) 버전을 사용하겠습니다. @@ -102,12 +101,19 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan_2_2_5B_t2v.mp4" > +Workflow preview + +Workflow preview + +Workflow preview + +Workflow preview + JSON 다운로드 또는 템플릿 라이브러리에서 "Wan2.2 5B" 검색 Comfy Cloud에서 열기 - ### 2. 모델 수동 다운로드 **디퓨전 모델** @@ -160,7 +166,6 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` Comfy Cloud에서 열기 - ### 2. 모델 수동 다운로드 **디퓨전 모델** @@ -173,7 +178,6 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` **텍스트 인코더** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - ``` ComfyUI/ ├───📂 models/ @@ -215,7 +219,6 @@ ComfyUI를 최신 버전으로 업데이트하신 후, 메뉴 `워크플로우` Comfy Cloud에서 열기 - 다음 이미지를 입력으로 사용할 수 있습니다: ![입력 이미지](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) @@ -272,7 +275,6 @@ ComfyUI/ Comfy Cloud에서 열기 - 다음 이미지를 입력 자료로 다운로드하세요: ![입력 자료](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v_start_image.png) diff --git a/ko/tutorials/video/zai/scail2.mdx b/ko/tutorials/video/zai/scail2.mdx index 8bd4b33b6..992d34d8f 100644 --- a/ko/tutorials/video/zai/scail2.mdx +++ b/ko/tutorials/video/zai/scail2.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Model Installation": d4fa2215 --- - - import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' **SCAIL-2**는 Wan2.1 기반의 엔드투엔드 캐릭터 애니메이션 모델로, 구동 비디오의 움직임을 참조 캐릭터 이미지에 적용하여 캐릭터 애니메이션(캐릭터가 동작을 수행하도록 함)과 비디오 내 캐릭터 교체(추적된 인물을 참조 캐릭터로 대체)를 모두 가능하게 합니다. @@ -31,6 +29,8 @@ import UpdateReminder from '/snippets/ko/tutorials/update-reminder.mdx' ## SCAIL-2 캐릭터 교체 워크플로우 +Workflow preview + Comfy Cloud에서 열기 diff --git a/tutorials/3d/hunyuan3D-2.mdx b/tutorials/3d/hunyuan3D-2.mdx index 3e371ce92..249813f86 100644 --- a/tutorials/3d/hunyuan3D-2.mdx +++ b/tutorials/3d/hunyuan3D-2.mdx @@ -17,7 +17,6 @@ Hunyuan3D 2.0 adopts a two-stage generation approach, first generating a geometr 1. **Geometry Generation Model (Hunyuan3D-DiT)**: Based on a flow diffusion Transformer architecture, it generates untextured geometric models that precisely match input conditions. 2. **Texture Generation Model (Hunyuan3D-Paint)**: Combines geometric conditions and multi-view diffusion techniques to add high-resolution textures to models, supporting PBR materials. - **Key Advantages** - **High-Precision Generation**: Sharp geometric structures, rich texture colors, support for PBR material generation, achieving near-realistic lighting effects. @@ -47,12 +46,37 @@ In the Hunyuan3D-2mv workflow, we'll use multi-view images to generate a 3D mode +### HY 3D 2.0 MV (`3d_hunyuan3d_multiview_to_model`) + +Generate 3D models from multiple views using Hunyuan3D 2.0 MV. + +HY 3D 2.0 MV workflow preview + Run this workflow instantly on Comfy Cloud - Download the workflow JSON file + Download JSON or search "HY 3D 2.0 MV" in Template Library + + + +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 56 · front view + + + `LoadImage` node 78 · left view + + + `LoadImage` node 80 · back view + + + `LoadImage` node 87 · right view @@ -67,7 +91,6 @@ Download the images below we will use them as input images. ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/left.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/back.png) - In this example, the input images have already been preprocessed to remove excess background. In actual use, you can use custom nodes like [ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) to automatically remove excess background. @@ -99,12 +122,37 @@ If you need to add more views, make sure to load other view images in the `Hunyu In the Hunyuan3D-2mv-turbo workflow, we'll use the Hunyuan3D-2mv-turbo model to generate 3D models. This model is a step distillation version of Hunyuan3D-2mv, allowing for faster 3D model generation. In this version of the workflow, we set `cfg` to 1.0 and add a `flux guidance` node to control the `distilled cfg` generation. +### HY 3D 2.0 MV Turbo (`3d_hunyuan3d_multiview_to_model_turbo`) + +Generate 3D models from multiple views using Hunyuan3D 2.0 MV Turbo. + +HY 3D 2.0 MV Turbo workflow preview + Run this workflow instantly on Comfy Cloud - Download the workflow JSON file + Download JSON or search "HY 3D 2.0 MV Turbo" in Template Library + + + +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 56 · front view + + + `LoadImage` node 82 · back view + + + `LoadImage` node 85 · left view + + + `LoadImage` node 87 · right view @@ -116,11 +164,9 @@ Please download the images below and drag into ComfyUI to load the workflow. Download the images below we will use them as input images. - ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) - ### 2. Manual Model Installation Download the model below and save it to the corresponding ComfyUI folder @@ -146,12 +192,28 @@ ComfyUI/ In the Hunyuan3D-2 workflow, we'll use the Hunyuan3D-2 model to generate 3D models. This model is not a multi-view model. In this workflow, we use the `Hunyuan3Dv2Conditioning` node instead of the `Hunyuan3Dv2ConditioningMultiView` node. +### HY 3D 2.0 (`3d_hunyuan3d_image_to_model`) + +Generate 3D models from single images using Hunyuan3D 2.0. + +HY 3D 2.0 workflow preview + Run this workflow instantly on Comfy Cloud - Download the workflow JSON file + Download JSON or search "HY 3D 2.0" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 56 · `3d_hunyuan3d_image_to_model_input_image.png` diff --git a/tutorials/3d/triposplat.mdx b/tutorials/3d/triposplat.mdx index 0e56235b8..4ba998c77 100644 --- a/tutorials/3d/triposplat.mdx +++ b/tutorials/3d/triposplat.mdx @@ -10,7 +10,11 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' Unlike traditional 3D reconstruction methods that require multiple views or generate meshes as the primary output, TripoSplat creates **Gaussian splat** representations: a rendering technique where thousands of colored 3D Gaussians are placed in space to represent a scene. This approach enables fast, high-quality rendering with controllable density and budget. -TripoSplat workflow +### TripoSplat: Image to Gaussian Splat (`3d_triposplat_image_to_gaussian_splat`) + +Upload a single 2D image. Generate a high-quality 3D Gaussian splat representation with controllable density and budget for rendering. + +TripoSplat workflow preview @@ -19,10 +23,24 @@ Unlike traditional 3D reconstruction methods that require multiple views or gene Run this workflow instantly on Comfy Cloud - Download JSON or search "TripoSplat" in Template Library + Download JSON or search "TripoSplat: Image to Gaussian Splat" in Template Library +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 99 · `white-hotel-on-rocky-island.png` + + + +
+ Input image +
+ ## How it works TripoSplat uses a **feed-forward architecture** that takes a single RGB image and directly predicts a set of 3D Gaussian primitives. The pipeline involves: diff --git a/tutorials/audio/ace-step/ace-step-v1-5.mdx b/tutorials/audio/ace-step/ace-step-v1-5.mdx index 125df1019..e14a0242f 100644 --- a/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -22,19 +22,22 @@ ACE-Step 1.5 is a major update to the open-source music generation model, now na The AIO version packages all models into a single checkpoint file, making it easier to download and manage. -### AIO Workflow +### ACE-Step 1.5 Music Generation AIO (`audio_ace_step_1_5_checkpoint`) - - Run the AIO workflow directly on Comfy Cloud. - +Input style tags and lyrics to generate a full song. The workflow uses the ACE-Step 1.5 model to produce commercial-grade music in under 10 seconds on consumer hardware. - - Download the all-in-one checkpoint workflow for local use. + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation AIO" in Template Library + ### AIO Model Download - + All-in-one checkpoint file (recommended for most users). @@ -51,33 +54,38 @@ The AIO version packages all models into a single checkpoint file, making it eas The split version allows you to download individual model components separately. -### Split Workflow +### ACE-Step 1.5 Music Generation Workflow (`audio_ace_step_1_5_split`) - - Run the split models workflow directly on Comfy Cloud. - +Input a text prompt describing the music style and optional lyrics. Generate a full, high-quality audio song in under 10 seconds on consumer hardware. - - Download the split models workflow for local use. + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation Workflow" in Template Library + ### Split Model Downloads - + + Diffusion model. - + Text encoder (0.6B). - + Text encoder (1.7B). - + VAE model. + **Split Models Storage Location** diff --git a/tutorials/audio/ace-step/ace-step-v1.mdx b/tutorials/audio/ace-step/ace-step-v1.mdx index dfdf62861..c818ef6ad 100644 --- a/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/tutorials/audio/ace-step/ace-step-v1.mdx @@ -60,7 +60,6 @@ Download the following audio file as the input audio:

Download Example Audio File for Input

- ### 2. Complete the Workflow Step by Step ![ACE-Step Step Guide](/images/tutorial/audio/ace_step/ace_step_1_m2m_step_guide.jpg) @@ -79,7 +78,6 @@ You can also implement the lyrics modification and editing functionality from th 1. In the example workflow, you can change the `tags` in `TextEncodeAceStepAudio` from `male voice` to `female voice` to generate female vocals. 2. You can also modify the `lyrics` in `TextEncodeAceStepAudio` to change the lyrics and thus the generated audio. Refer to the examples on the ACE-Step project page for more details. - ## ACE-Step Prompt Guide ACE currently uses two types of prompts: `tags` and `lyrics`. diff --git a/tutorials/audio/stable-audio/stable-audio-1.mdx b/tutorials/audio/stable-audio/stable-audio-1.mdx index 7ba0a8e55..a8cbeaf92 100644 --- a/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -15,15 +15,20 @@ import UpdateReminder from "/snippets/tutorials/update-reminder.mdx" ## Workflow - - Download JSON or search "Stable Audio 1.0" in Template Library - +### Stable Audio 1.0: Text to Audio (`audio_stable_audio_example`) + +Generate audio from text prompts using Stable Audio. + +Stable Audio 1.0 text to audio workflow preview - + + Open in Comfy Cloud - -![Stable Audio 1.0 workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_example-1.webp) + + Download JSON or search "Stable Audio 1.0: Text to Audio" in Template Library + + The workflow uses **standard ComfyUI nodes** — no custom nodes required. It loads the Stable Audio 1.0 checkpoint, encodes your prompt via a CLIP text encoder (t5-base), denoises the latent audio with a KSampler, and decodes it to audio through the model's VAE. @@ -39,7 +44,7 @@ When loading the workflow, ComfyUI will prompt you with download links for any m ### Checkpoint - + 2.3GB. Place in models/checkpoints/ @@ -54,7 +59,7 @@ Place the checkpoint in: ### Text encoder - + Text encoder for prompt conditioning. Place in models/text_encoders/ diff --git a/tutorials/audio/stable-audio/stable-audio-3.mdx b/tutorials/audio/stable-audio/stable-audio-3.mdx index d97a273cb..f3db10439 100644 --- a/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -22,17 +22,20 @@ Stable Audio 3 comes in three variants: ## Available workflows -### Stable Audio 3 Medium +### Stable Audio 3.0 Medium (`audio_stable_audio_3_medium`) - - Download JSON or search "Stable Audio 3 Medium" in Template Library - +Input a short text idea, optional duration, seed, and category. Generate stereo audio (music, SFX, or instruments) using Stable Audio 3 with optional AI-driven text expansion. + +Stable Audio 3 Medium workflow preview - + + Open in Comfy Cloud - -![Stable Audio 3 Medium workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium-1.webp) + + Download JSON or search "Stable Audio 3.0 Medium" in Template Library + + The **Stable Audio 3 Medium** workflow is a full-featured text-to-audio generation pipeline. You provide a short text idea, optional duration, seed, and category — the workflow expands your prompt using Qwen with a **category-aware reprompt template**, then generates stereo audio via the Stable Audio 3 checkpoint. @@ -44,17 +47,20 @@ The **Stable Audio 3 Medium** workflow is a full-featured text-to-audio generati 5. **Enable reprompt** — Toggle `use_reprompt` on to let Qwen expand your short idea into a detailed prompt before generation 6. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -### Stable Audio 3 Medium Base +### Stable Audio 3.0 Medium Base (`audio_stable_audio_3_medium_base`) - - Download JSON or search "Stable Audio 3 Medium Base" in Template Library - +Input a short text description of a sound, music, or effect. The workflow expands your prompt with Qwen and generates a stereo audio clip from Stable Audio 3. + +Stable Audio 3 Medium Base workflow preview - + + Open in Comfy Cloud - -![Stable Audio 3 Medium Base workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium_base-1.webp) + + Download JSON or search "Stable Audio 3.0 Medium Base" in Template Library + + A simplified version of Stable Audio 3 without Qwen reprompt expansion. It expects a complete text prompt and passes it directly to the model. Use this when you already have a detailed prompt and want faster generation. @@ -70,13 +76,15 @@ When loading the workflow, ComfyUI will prompt you with download links for any m ### Checkpoints - + + For the Medium workflow. Place in models/checkpoints/ - + For the Medium Base workflow. Place in models/checkpoints/ + Place checkpoints in: @@ -90,13 +98,15 @@ Place checkpoints in: ### Text encoders - + + Required for all Stable Audio 3 workflows. Place in models/text_encoders/ - + Required for the Medium workflow (Qwen reprompt). Place in models/text_encoders/ + Place text encoders in: diff --git a/tutorials/basic/multiple-loras.mdx b/tutorials/basic/multiple-loras.mdx index 34a425255..b566822e5 100644 --- a/tutorials/basic/multiple-loras.mdx +++ b/tutorials/basic/multiple-loras.mdx @@ -31,7 +31,6 @@ Download the [MoXinV1.safetensors](https://civitai.com/api/download/models/14856 Download the image below and **drag it into ComfyUI** to load the workflow: ![ComfyUI Workflow - Multiple LoRAs](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/multiple_loras.png) - Images containing workflow JSON in their metadata can be directly dragged into ComfyUI or loaded using the menu `Workflows` -> `Open (ctrl+o)`. diff --git a/tutorials/basic/upscale.mdx b/tutorials/basic/upscale.mdx index 4a93da2e4..c0545d6ac 100644 --- a/tutorials/basic/upscale.mdx +++ b/tutorials/basic/upscale.mdx @@ -46,7 +46,6 @@ Save the model file (.pth) in `ComfyUI/models/upscale_models` directory - ### Workflow and Assets Download and drag the following image into ComfyUI to load the basic upscaling workflow: diff --git a/tutorials/flux/flux-1-controlnet.mdx b/tutorials/flux/flux-1-controlnet.mdx index 301e186d3..2264daa8c 100644 --- a/tutorials/flux/flux-1-controlnet.mdx +++ b/tutorials/flux/flux-1-controlnet.mdx @@ -40,24 +40,36 @@ For image preprocessors, you can use the following custom nodes to complete imag ## FLUX.1-Canny-dev Complete Version Workflow +### Flux.1 Canny Model (`flux_canny_model_example`) + +Generate images guided by edge detection using Flux.1 Canny. + +Flux.1 Canny workflow preview + - - Download JSON or search "Flux.1 Canny" in Template Library - Open in Comfy Cloud + + Download JSON or search "Flux.1 Canny" in Template Library + -### 1. Workflow and Asset +**Input materials** -Please download the workflow image below and drag it into ComfyUI to load the workflow +Upload this file to the matching `LoadImage` node: -![ComfyUI Workflow - ControlNet](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-canny-dev.png) + + + `LoadImage` node 17 · `flux_canny_model_example_input_image.png` + + -Please download the image below, which we will use as the input image +
+ flux_canny_model_example_input_image.png +
-![ComfyUI Flux.1 Canny Controlnet input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-canny-dev-input.png) +### 1. Workflow and Asset ### 2. Manual Models Installation @@ -111,26 +123,36 @@ Or use the following custom nodes to complete image preprocessing: ## FLUX.1-Depth-dev-lora Workflow +### Flux.1 Depth Lora (`flux_depth_lora_example`) + +Generate images guided by depth information using Flux.1 LoRA. + +Flux.1 Depth LoRA workflow preview + - - Download JSON or search "Flux.1 Depth LoRA" in Template Library - Open in Comfy Cloud + + Download JSON or search "Flux.1 Depth LoRA" in Template Library + -The LoRA version workflow builds on the complete version by adding the LoRA model. Compared to the [complete version of the Flux workflow](/tutorials/flux/flux-1-text-to-image), it adds nodes for loading and using the corresponding LoRA model. +**Input materials** -### 1. Workflow and Asset - -Please download the workflow image below and drag it into ComfyUI to load the workflow +Upload this file to the matching `LoadImage` node: -![ComfyUI Workflow - ControlNet](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-depth-dev-lora.png) + + + `LoadImage` node 17 · `flux_depth_lora_example_input_image.png` + + -Please download the image below, which we will use as the input image +
+ flux_depth_lora_example_input_image.png +
-![ComfyUI Flux.1 Depth Controlnet input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-depth-dev-lora-input.png) +### 1. Workflow and Asset ### 2. Manual Model Download diff --git a/tutorials/flux/flux-1-fill-dev.mdx b/tutorials/flux/flux-1-fill-dev.mdx index c613cea72..541aad887 100644 --- a/tutorials/flux/flux-1-fill-dev.mdx +++ b/tutorials/flux/flux-1-fill-dev.mdx @@ -51,22 +51,43 @@ ComfyUI/ ## Flux.1 Fill dev inpainting workflow -### 1. Inpainting workflow and asset +### Flux.1 Inpaint (`flux_fill_inpaint_example`) + +Fill missing parts of images using Flux.1 Fill Inpainting. + +Flux.1 inpaint workflow preview - - Download JSON or search "flux_fill_inpaint" in Template Library - Open in Comfy Cloud + + Download JSON or search "Flux.1 Inpaint" in Template Library + -Please download the image below and drag it into ComfyUI to load the corresponding workflow -![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_fill_inpaint_example_input_image.png` + + + +
+ flux_fill_inpaint_example_input_image.png +
+ +**Example output** + +
+ Input image + Flux.1 inpaint example output +
-Please download the image below, we will use it as the input image -![ComfyUI Flux.1 inpaint input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input.png) +### 1. Inpainting workflow and asset The corresponding image already contains an alpha channel, so you don't need to draw a mask separately. @@ -87,22 +108,36 @@ If you want to draw your own mask, please [click here](https://raw.githubusercon ## Flux.1 Fill dev Outpainting Workflow -### 1. Outpainting workflow and asset +### Flux.1 Outpaint (`flux_fill_outpaint_example`) + +Extend images beyond boundaries using Flux.1 outpainting. + +Flux.1 outpaint workflow preview - - Download JSON or search "flux_fill_outpaint" in Template Library - Open in Comfy Cloud + + Download JSON or search "Flux.1 Outpaint" in Template Library + -Please download the image below and drag it into ComfyUI to load the corresponding workflow -![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) +**Input materials** -Please download the image below, we will use it as the input image -![ComfyUI Flux.1 outpaint input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint_input.png) +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_fill_outpaint_example_input_image.png` + + + +
+ flux_fill_outpaint_example_input_image.png +
+ +### 1. Outpainting workflow and asset ### 2. Steps to run the workflow diff --git a/tutorials/flux/flux-1-kontext-dev.mdx b/tutorials/flux/flux-1-kontext-dev.mdx index c7dfbf460..d8c01f732 100644 --- a/tutorials/flux/flux-1-kontext-dev.mdx +++ b/tutorials/flux/flux-1-kontext-dev.mdx @@ -72,26 +72,45 @@ Model save location ## Flux.1 Kontext Dev Workflow +### Flux Kontext Dev Image Edit (`flux_kontext_dev_basic`) + +Smart image editing that keeps characters consistent, edits specific parts without affecting others, and preserves original styles. + +Flux Kontext Dev workflow preview + - - Download JSON or search "Flux Kontext Dev" in Template Library - Open in Comfy Cloud + + Download JSON or search "Flux Kontext Dev" in Template Library + -This workflow uses the `Load Image(from output)` node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing. +**Input materials** -### 1. Workflow and Input Image Download +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 190 · `flux_kontext_dev_basic_input_image.jpg` + + + +
+ flux_kontext_dev_basic_input_image.jpg +
-Download the following files and drag them into ComfyUI to load the corresponding workflow +**Example output** -![ComfyUI Flux.1 Kontext Pro Image Partner Nodes Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/kontext/dev/flux_1_kontext_dev_basic.png) +
+ Input image + Flux Kontext Dev example output +
-**Input Image** +This workflow uses the `Load Image(from output)` node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing. -![ComfyUI Flux Kontext Native Workflow Input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/kontext/dev/rabbit.jpg) +### 1. Workflow and Input Image Download ### 2. Complete the workflow step by step diff --git a/tutorials/flux/flux-1-text-to-image.mdx b/tutorials/flux/flux-1-text-to-image.mdx index d0eeee46b..980550bc0 100644 --- a/tutorials/flux/flux-1-text-to-image.mdx +++ b/tutorials/flux/flux-1-text-to-image.mdx @@ -41,19 +41,26 @@ If you can't download models from [black-forest-labs/FLUX.1-dev](https://hugging ### Flux.1 Dev -#### 1. Workflow File +### Flux.1 Dev fp8: Text to Image (`flux_dev_checkpoint_example`) + +Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices with limited VRAM, requires only one model file, but image quality is slightly reduced compared to the full version. + +Flux.1 Dev fp8 workflow preview Run this workflow on Comfy Cloud - Download JSON or search "Flux.1 Dev" in Template Library + Download JSON or search "Flux.1 Dev fp8" in Template Library -Please download the image below and drag it into ComfyUI to load the workflow. -![Flux Dev Original Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) +**Example output** + +![Flux.1 Dev fp8 example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_checkpoint_example.png) + +#### 1. Workflow File #### 2. Manual Model Installation @@ -99,20 +106,22 @@ Thanks to Flux's excellent prompt following capability, we don't need any negati ### Flux.1 Schnell -#### 1. Workflow File +### Flux.1 Schnell FP8 (`flux_schnell`) + +Quickly generate images with Flux.1 Schnell fp8 quantized version. Ideal for low-end hardware, requires only 4 steps to generate images. + +Flux.1 Schnell FP8 workflow preview Run this workflow on Comfy Cloud - Download JSON or search "Flux.1 Schnell" in Template Library + Download JSON or search "Flux.1 Schnell" in Template Library -Please download the image below and drag it into ComfyUI to load the workflow. - -![Flux Schnell Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) +#### 1. Workflow File #### 2. Manual Models Installation @@ -160,18 +169,26 @@ but it also requires less VRAM, and you only need to install one model file to t ### Flux.1 Dev +### Flux.1 Dev: Text to Image (`flux_dev_full_text_to_image`) + +Generate high-quality images with Flux Dev full version. Requires larger VRAM and multiple model files, but provides the best prompt following capability. + +Flux.1 Dev text-to-image workflow preview + Run this workflow on Comfy Cloud - Download JSON or search "Flux.1 Dev FP8" in Template Library + Download JSON or search "Flux.1 Dev FP8" in Template Library -Please download the image below and drag it into ComfyUI to load the workflow. +**Example output** -![Flux Dev fp8 Checkpoint Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_fp8.png) +![Flux.1 Dev FP8 checkpoint example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_full_text_to_image.png) + +Please download the image below and drag it into ComfyUI to load the workflow. Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. @@ -179,19 +196,23 @@ Ensure that the corresponding `Load Checkpoint` node loads `flux1-dev-fp8.safete ### Flux.1 Schnell +### Flux.1 Schnell FP8 (`flux_schnell`) + +Quickly generate images with Flux.1 Schnell fp8 quantized version. Ideal for low-end hardware, requires only 4 steps to generate images. + +Flux.1 Schnell FP8 checkpoint workflow preview + Run this workflow on Comfy Cloud - Download JSON or search "Flux.1 Schnell FP8" in Template Library + Download JSON or search "Flux.1 Schnell FP8" in Template Library Please download the image below and drag it into ComfyUI to load the workflow. -![Flux Schnell fp8 Checkpoint Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) - Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. Ensure that the corresponding `Load Checkpoint` node loads `flux1-schnell-fp8.safetensors`, and you can try to run the workflow. \ No newline at end of file diff --git a/tutorials/flux/flux-1-uso.mdx b/tutorials/flux/flux-1-uso.mdx index bec397bc0..530700808 100644 --- a/tutorials/flux/flux-1-uso.mdx +++ b/tutorials/flux/flux-1-uso.mdx @@ -23,27 +23,45 @@ USO supports three main approaches: -### 1. Workflow and input +### Flux.1 Dev USO Reference Image Generation (`flux1_dev_uso_reference_image_gen`) -Download the image below and drag it into ComfyUI to load the corresponding workflow. +Use reference images to control both style and subject. Keep your character's face while changing artistic style, or apply artistic styles to new scenes. -![Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) +Flux.1 Dev USO reference image workflow preview - - Download the workflow JSON and drag it into ComfyUI - Run this workflow on Comfy Cloud + + Download the workflow JSON and drag it into ComfyUI + -Use the image below as an input image. +**Input materials** -![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/input.png) +Upload this file to the matching `LoadImage` node: -### 2. Model links + + + `LoadImage` node 47 · `flux1_dev_uso_reference_image_gen_input_image.png` + + + +
+ flux1_dev_uso_reference_image_gen_input_image.png +
+ +**Example output** + +
+ Input image + Flux.1 Dev USO example output +
+### 1. Workflow and input + +### 2. Model links **checkpoints** @@ -75,7 +93,6 @@ Please download all models and place them in the following directories: │ │ └── sigclip_vision_patch14_384.safetensors ``` - ### 3. Workflow instructions ![Workflow instructions](/images/tutorial/flux/flux1_uso_reference_image_gen.jpg) diff --git a/tutorials/flux/flux-2-dev.mdx b/tutorials/flux/flux-2-dev.mdx index 0cb271140..4d21b3774 100644 --- a/tutorials/flux/flux-2-dev.mdx +++ b/tutorials/flux/flux-2-dev.mdx @@ -28,9 +28,11 @@ We are using quantized weights in this workflow. The original FLUX.2 repository -## Single image generation workflow +## Flux.2 Dev (`image_flux2`) -Basic text-to-image workflow for generating single images with FLUX.2 Dev. +Generate photorealistic images with multi-reference consistency and professional text rendering. + +Flux.2 Dev workflow preview @@ -41,19 +43,77 @@ Basic text-to-image workflow for generating single images with FLUX.2 Dev. -## Multi-image reference workflow +**Input materials** + +Upload this file to `LoadImage` node **46**: + + + + `LoadImage` node 46 · `image_flux2_input_image.png` + + + +
+ Input Image for Flux.2 Dev + Flux.2 Dev example output +
+ +## Flux.2 Dev Text to Image (`image_flux2_text_to_image`) -A 2-image reference workflow example. You can extend this implementation to support more reference images. +Text-to-image with enhanced lighting, materials, and realistic details. No input images are required. + +Flux.2 Dev text-to-image workflow preview + + + + Open this workflow directly in Comfy Cloud + + + Download the JSON workflow file for local use + + + +**Example output** + +![Flux.2 Dev text-to-image example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_text_to_image.png) + +## Product Mockup (`image_flux2_fp8`) + +Create product mockups by applying design patterns to packaging, mugs, and other products using multi-reference consistency. + +Flux.2 Dev product mockup workflow preview Open this workflow directly in Comfy Cloud - + Download the JSON workflow file for local use +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 42 · `image_flux2_input_ref_image.png` + + + `LoadImage` node 46 · `image_flux2_input_Illustration.png` + + + +
+ Input Ref Image for Flux.2 Dev + Input Illustration for Flux.2 Dev +
+ +**Example output** + +![Flux.2 Dev product mockup output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_fp8.png) + ## Model links **text_encoders** diff --git a/tutorials/flux/flux-2-klein.mdx b/tutorials/flux/flux-2-klein.mdx index 660502295..05fc52f56 100644 --- a/tutorials/flux/flux-2-klein.mdx +++ b/tutorials/flux/flux-2-klein.mdx @@ -21,6 +21,10 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a ## Flux.2 Klein 4B Workflows +### Flux.2 [Klein] 4B: Text to Image (`image_flux2_klein_text_to_image`) + +Flux.2 Klein 4B text-to-image workflow preview + Run this workflow on Comfy Cloud @@ -30,6 +34,14 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a +**Example output** + +![Flux.2 Klein 4B text-to-image example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_klein_text_to_image.png) + +### Flux.2 [Klein] 4B: Image Edit (`image_flux2_klein_image_edit_4b_base`) + +Flux.2 Klein 4B image edit base workflow preview + Run this workflow on Comfy Cloud @@ -39,6 +51,32 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 76 · `robed_women.png` + + + `LoadImage` node 81 · `pink_tone_chair.png` + + + +
+ robed_women.png + pink_tone_chair.png +
+ +**Example output** + +![Flux.2 Klein 4B image edit base example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/image_flux2_klein_image_edit_4b_base.png) + +### Flux.2 [Klein] 4B Distilled: Image Edit (`image_flux2_klein_image_edit_4b_distilled`) + +Flux.2 Klein 4B image edit distilled workflow preview + Run this workflow on Comfy Cloud @@ -48,6 +86,28 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 76 · `handbag_white.png` + + + `LoadImage` node 81 · `comfy_logo_blue.png` + + + +
+ handbag_white.png + comfy_logo_blue.png +
+ +**Example output** + +![Flux.2 Klein 4B image edit distilled example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_klein_image_edit_4b_distilled.png) + ## Flux.2 Klein 4B Model Downloads @@ -64,6 +124,7 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a VAE for 4B models. + **4B Model Storage Location** ``` @@ -80,6 +141,10 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a ## Flux.2 Klein 9B Workflows +### Flux.2 [Klein] 9B: Text to Image (`image_flux2_text_to_image_9b`) + +Flux.2 Klein 9B text-to-image workflow preview + Run this workflow on Comfy Cloud @@ -89,6 +154,14 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a +**Example output** + +![Flux.2 Klein 9B text-to-image example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_text_to_image_9b.png) + +### Flux.2 [Klein] 9B: Image Edit (`image_flux2_klein_image_edit_9b_base`) + +Flux.2 Klein 9B image edit base workflow preview + Run this workflow on Comfy Cloud @@ -98,6 +171,32 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 76 · `car_interior_white.jpeg` + + + `LoadImage` node 81 · `comfy_logo_blue.png` + + + +
+ car_interior_white.jpeg + comfy_logo_blue.png +
+ +**Example output** + +![Flux.2 Klein 9B image edit base example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/image_flux2_klein_image_edit_9b_base.png) + +### Flux.2 [Klein] 9B Distilled: Image Edit (`image_flux2_klein_image_edit_9b_distilled`) + +Flux.2 Klein 9B image edit distilled workflow preview + Run this workflow on Comfy Cloud @@ -107,6 +206,28 @@ FLUX.2 [Klein] is the fastest model in the Flux family, unifying text-to-image a +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 76 · `bold_outfit_woman.jpeg` + + + `LoadImage` node 121 · `handbag_white.png` + + + +
+ bold_outfit_woman.jpeg + handbag_white.png +
+ +**Example output** + +![Flux.2 Klein 9B image edit distilled example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_klein_image_edit_9b_distilled.png) + ## Flux.2 Klein 9B Model Downloads @@ -127,6 +248,7 @@ For diffusion models, please visit BFL's repo, accept the agreement, and then do VAE for 9B models. + **9B Model Storage Location** ``` diff --git a/tutorials/flux/flux1-krea-dev.mdx b/tutorials/flux/flux1-krea-dev.mdx index 6923ca645..5f435eef2 100644 --- a/tutorials/flux/flux1-krea-dev.mdx +++ b/tutorials/flux/flux1-krea-dev.mdx @@ -24,20 +24,27 @@ This model is released under the [flux-1-dev-non-commercial-license](https://hug -#### 1. Workflow Files +### Flux.1 Krea Dev (`flux1_krea_dev`) + +A fine-tuned FLUX model pushing photorealism to the max. -Download the image or JSON below and drag it into ComfyUI to load the corresponding workflow -![Flux Krea Dev Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) +Flux.1 Krea Dev workflow preview Run this workflow on Comfy Cloud - Download JSON or search "Flux.1 Krea Dev" in Template Library + Download JSON or search "Flux.1 Krea Dev" in Template Library +**Example output** + +![Flux.1 Krea Dev example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux1_krea_dev.png) + +#### 1. Workflow Files + #### 2. Manual Model Installation Please download the following model files: diff --git a/tutorials/image/anima/anima.mdx b/tutorials/image/anima/anima.mdx index 8b93d5ffb..1441b40ea 100644 --- a/tutorials/image/anima/anima.mdx +++ b/tutorials/image/anima/anima.mdx @@ -30,15 +30,24 @@ Both workflows use a **Subgraph** node to manage the text-to-image generation pi This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -### Anima Base v1: Text to Image +### Anima Base v1: Text to Image (`image_anima_base_v1`) - - Download JSON or search "Anima Base v1" in Template Library - +Input a text prompt describing an anime or artistic illustration. Generate a non-photorealistic image focused on anime concepts, characters, or styles. + +Anima Base v1 text-to-image workflow preview - + + Open in Comfy Cloud + + Download JSON or search "Anima Base v1" in Template Library + + + +**Example output** + +![Anima Base v1 example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_base_v1.png) #### Get started @@ -47,19 +56,24 @@ Both workflows use a **Subgraph** node to manage the text-to-image generation pi 3. Select the **Anima Base v1: Text to Image** workflow 4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** -#### Example output - -Anima Base v1 example output +### Anima Preview: Anime Text-to-Image Generation (`image_anima_preview`) -### Anima Preview: Anime Text-to-Image Generation +Input a text prompt to generate an anime-style image using the Anima model. Configure settings like steps and CFG scale to control the output. - - Download JSON or search "Anima Preview" in Template Library - +Anima Preview text-to-image workflow preview - + + Open in Comfy Cloud + + Download JSON or search "Anima Preview" in Template Library + + + +**Example output** + +![Anima Preview example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_preview.png) #### Get started @@ -68,23 +82,19 @@ Both workflows use a **Subgraph** node to manage the text-to-image generation pi 3. Select the **Anima Anime Text-to-Image Generation** workflow 4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** -#### Example output - -Anima Preview example output - ## Anima model downloads You can find all model files at [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima) on Hugging Face. - + Diffusion model (2B) for Anima Base v1. - + Text encoder (Qwen-3 0.6B) shared by both workflows. - + VAE shared by both workflows. @@ -105,7 +115,7 @@ You can find all model files at [circlestone-labs/Anima](https://huggingface.co/ If you are using the Preview workflow, download the preview diffusion model instead: - + Diffusion model (2B) for Anima Preview. diff --git a/tutorials/image/boogu/boogu-image-0.1.mdx b/tutorials/image/boogu/boogu-image-0.1.mdx index 8f88dacb4..ab1549dd3 100644 --- a/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/tutorials/image/boogu/boogu-image-0.1.mdx @@ -28,36 +28,41 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Boogu-Image-0.1-Turbo text-to-image workflow + + +### Boogu Turbo: Text to Image (`image_boogu_image_0_1_turbo_t2i`) + +Generate high-quality images from text prompts using the Boogu Turbo model, a distilled 10B-parameter text-to-image pipeline that produces photorealistic outputs in just 4 steps. + +Boogu Turbo text-to-image workflow preview + - + Open in Comfy Cloud - Download JSON or search "Boogu image 0.1" in Template Library + Download JSON or search "Boogu Turbo" in Template Library - - The Boogu-Image-0.1-Turbo workflow uses a subgraph that encapsulates the diffusion, text encoding, and VAE decode pipeline. You provide a text prompt and resolution settings, and the subgraph handles the rest. ### Boogu-Image-0.1-Turbo model downloads - + + Diffusion model for Boogu-Image-0.1-Turbo. - - + Text encoder for Boogu-Image-0.1-Turbo. - - + VAE for Boogu-Image-0.1-Turbo. - - + LoRA module (rank-128) for Boogu-Image-0.1-Turbo. + **Model Storage Location** @@ -76,34 +81,48 @@ The Boogu-Image-0.1-Turbo workflow uses a subgraph that encapsulates the diffusi ## Boogu-Image-0.1-Edit image editing workflow +### Boogu image 0.1 Edit (`image_boogu_image_0_1_edit`) + +Edit images using Boogu's instruction-driven model, taking one input image and generating an edited output alongside a side-by-side comparison view. + +Boogu image 0.1 Edit workflow preview + - + Open in Comfy Cloud - Download JSON or search "Boogu image 0.1 Edit" in Template Library + Download JSON or search "Boogu image 0.1 Edit" in Template Library - - - Get the example input image for this workflow +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 32 · `tech_cowboy.png` +
+ tech_cowboy.png +
+ ### Boogu-Image-0.1-Edit model downloads - + + Diffusion model for Boogu-Image-0.1-Edit. - - + Text encoder for Boogu-Image-0.1-Edit. - - + VAE for Boogu-Image-0.1-Edit. + **Model Storage Location** diff --git a/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index bf6bb30cb..a851af406 100644 --- a/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -14,7 +14,6 @@ Cosmos-Predict2 supports various generation methods including Text-to-Image (Tex GitHub:[Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) - This guide will walk you through completing **text-to-image** workflow in ComfyUI. For the video generation section, please refer to the following part: @@ -35,7 +34,6 @@ Please download the image below and drag it into ComfyUI to load the workflow. T ![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/cosmos/predict2/cosmos_predict2_2B_t2i.png) - ### 2. Manual Model Installation If the model download wasn't successful, you can try to download them manually by yourself in this section. @@ -54,7 +52,6 @@ For other weights, please visit [Cosmos_Predict2_repackaged](https://huggingface [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - File Storage Location ``` 📂 ComfyUI/ diff --git a/tutorials/image/ernie-image/ernie-image.mdx b/tutorials/image/ernie-image/ernie-image.mdx index 40ed8a63b..cf8796860 100644 --- a/tutorials/image/ernie-image/ernie-image.mdx +++ b/tutorials/image/ernie-image/ernie-image.mdx @@ -22,22 +22,32 @@ The model includes a built-in **Prompt Enhancer** (3B) that expands short inputs - [Hugging Face (ERNIE-Image)](https://huggingface.co/Baidu/ERNIE-Image) - [Hugging Face (ComfyUI Support)](https://huggingface.co/Comfy-Org/ERNIE-Image) - ## ERNIE-Image text-to-image workflow - - Download the ERNIE-Image text-to-image workflow JSON file. - + + +### Ernie Image: Text to Image (`image_ernie_image`) + +Generate images from text prompts using the ERNIE-Image model. Input a text description to produce detailed, structured visuals with a broad stylistic range. + +ERNIE-Image text-to-image workflow preview - - Run this workflow directly on Comfy Cloud. + + + Run this workflow directly on Comfy Cloud + + Download the ERNIE-Image text-to-image workflow JSON file + + - +**Example output** + +![ERNIE-Image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image.png) ### Get started -1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_ernie_image&utm_source=docs) +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_ernie_image&utm_source=docs&utm_medium=referral&utm_campaign=ernie-image) 2. Go to **Template** and search for **ERNIE-Image** 3. Select the **ERNIE-Image** workflow 4. Download any missing models, update the prompt, and click **Run** @@ -46,21 +56,20 @@ The model includes a built-in **Prompt Enhancer** (3B) that expands short inputs You can find all repackaged model files at [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) on Hugging Face. - + + Diffusion model for ERNIE-Image. - - + Text encoder for ERNIE-Image. - - + Prompt Enhancer text encoder for ERNIE-Image. - - + VAE for ERNIE-Image. + **Model storage location** @@ -80,31 +89,41 @@ You can find all repackaged model files at [Comfy-Org/ERNIE-Image](https://huggi [ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo) is a faster variant optimized with DMD and RL, generating images in just **8 steps** compared to the ~50 steps required by the standard model. - - Download the ERNIE-Image-Turbo text-to-image workflow JSON file. - +### Ernie Image Turbo: Text To Image (`image_ernie_image_turbo`) + +Generate images from text prompts using the ERNIE-Image turbo model. Input a text description and receive a high-quality image with precise text rendering. - - Run this workflow directly on Comfy Cloud. +ERNIE-Image-Turbo text-to-image workflow preview + + + + Run this workflow directly on Comfy Cloud + + + Download the ERNIE-Image-Turbo text-to-image workflow JSON file + + +**Example output** + +![ERNIE-Image-Turbo example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image_turbo.png) ### ERNIE-Image-Turbo model downloads - + + Diffusion model for ERNIE-Image-Turbo. - - + Text encoder for ERNIE-Image-Turbo. - - + Prompt Enhancer text encoder for ERNIE-Image-Turbo. - - + VAE for ERNIE-Image-Turbo. + **Model storage location** diff --git a/tutorials/image/hidream/hidream-e1.mdx b/tutorials/image/hidream/hidream-e1.mdx index 1c6f04768..b5b91709f 100644 --- a/tutorials/image/hidream/hidream-e1.mdx +++ b/tutorials/image/hidream/hidream-e1.mdx @@ -63,28 +63,36 @@ Model Save Location │ └── hidream_e1_full_bf16.safetensors ``` - ## HiDream E1.1 ComfyUI Native Workflow Example +### HiDream E1.1 Image Editing (`hidream_e1_1`) + +Edit images with HiDream E1.1. Superior image quality and editing accuracy compared to HiDream-E1-Full. + +HiDream E1.1 image editing workflow preview + - + Open in Comfy Cloud - Download JSON or search "HiDream E1.1" in Template Library + Download JSON or search "HiDream E1.1" in Template Library -E1.1 is an updated version released on July 16, 2025. This version supports dynamic 1-megapixel resolution, and the workflow uses the `Scale Image to Total Pixels` node to dynamically adjust the input image to 1 million pixels. +**Input materials** - -Here are the VRAM usage references during testing: -1. A100 40GB (VRAM usage 95%): First generation: 211s, second generation: 73s +Upload this file to the matching `LoadImage` node: -2. 4090D 24GB (VRAM usage 98%) -- Full version: Out of memory -- FP8_e4m3fn_fast (VRAM 98%) First generation: 120s, second generation: 91s - + + + `LoadImage` node 13 · `hidream_e1_1_input_image.jpg` + + + +
+ hidream_e1_1_input_image.jpg +
### 1. HiDream E1.1 Workflow and Related Materials @@ -115,18 +123,44 @@ Follow these steps to run the workflow: - Since HiDream E1.1 supports dynamic input with a total of 1 million pixels, the workflow uses `Scale Image to Total Pixels` to process and convert all input images, which may cause the aspect ratio to differ from the original input image. - When using the fp16 version of the model, in actual tests, the full version ran out of memory on both A100 40GB and 4090D 24GB, so the workflow is set by default to use `fp8_e4m3fn_fast` for inference. - ## HiDream E1 ComfyUI Native Workflow Example +### HiDream E1 Image Edit (`hidream_e1_full`) + +Edit images with HiDream E1. Professional natural language image editing model. + +HiDream E1 image editing workflow preview + - + Open in Comfy Cloud - Download JSON or search "HiDream E1 Full" in Template Library + Download JSON or search "HiDream E1 Full" in Template Library +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 13 · `hidream_e1_full_input_image.jpg` + + + +
+ hidream_e1_full_input_image.jpg +
+ +**Example output** + +
+ Input image + HiDream E1 example output +
+ E1 is a model released on April 28, 2025. diff --git a/tutorials/image/hidream/hidream-i1.mdx b/tutorials/image/hidream/hidream-i1.mdx index 8984a8f8a..d52664e51 100644 --- a/tutorials/image/hidream/hidream-i1.mdx +++ b/tutorials/image/hidream/hidream-i1.mdx @@ -92,7 +92,13 @@ Model file save location │ └── 📂 diffusion_models/ │ └── ... # We will guide you to install in the corresponding version workflow ``` -### HiDream-I1 Full Version Workflow +### HiDream I1 Full Version Workflow + +### HiDream I1 Full (`hidream_i1_full`) + +Generate images with HiDream I1 Full. Complete version with 50 inference steps for highest quality output. + +HiDream I1 Full workflow preview @@ -103,6 +109,10 @@ Model file save location +**Example output** + +![HiDream I1 Full example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_full.png) + #### 1. Model File Download Please select the appropriate version based on your hardware. Click the link and download the corresponding model file to save it to the `ComfyUI/models/diffusion_models/` folder. @@ -137,6 +147,12 @@ Complete the workflow execution step by step ### HiDream-I1 Dev Version Workflow +### HiDream I1 Dev (`hidream_i1_dev`) + +Generate images with HiDream I1 Dev. Balanced version with 28 inference steps, suitable for medium-range hardware. + +HiDream I1 Dev workflow preview + Run this workflow on Comfy Cloud with zero setup @@ -146,6 +162,10 @@ Complete the workflow execution step by step +**Example output** + +![HiDream I1 Dev example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_dev.png) + #### 1. Model File Download Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. @@ -178,6 +198,12 @@ Complete the workflow execution step by step ### HiDream-I1 Fast Version Workflow +### HiDream I1 Fast (`hidream_i1_fast`) + +Generate images quickly with HiDream I1 Fast. Lightweight version with 16 inference steps, ideal for rapid previews on lower-end hardware. + +HiDream I1 Fast workflow preview + Run this workflow on Comfy Cloud with zero setup @@ -187,6 +213,10 @@ Complete the workflow execution step by step +**Example output** + +![HiDream I1 Fast example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_fast.png) + #### 1. Model File Download Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. diff --git a/tutorials/image/hidream/hidream-o1.mdx b/tutorials/image/hidream/hidream-o1.mdx index ad23f65a7..c812dad4d 100644 --- a/tutorials/image/hidream/hidream-o1.mdx +++ b/tutorials/image/hidream/hidream-o1.mdx @@ -25,37 +25,44 @@ HiDream-O1-Image is released under the [MIT License](https://github.com/HiDream- ## HiDream-O1-Image Full Workflow -### 1. Download Workflow +### HiDream O1 Full: Image generation (`image_hidream_o1`) -Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` -> `Image` and find "HiDream O1 Full: Image generation". +Input a text prompt and optionally upload reference images. Generate a high-resolution image up to 2048x2048 with text-to-image, editing, or subject-driven personalization. -![HiDream-O1-Image Full Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) +HiDream O1 Full workflow preview - - Download workflow + + + Open in Comfy Cloud - - - Open in cloud + + Download JSON or search "HiDream O1 Full" in Template Library + + +**Example output** + +![HiDream O1 Full example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) + +### 1. Download Workflow ### 2. Download Models **Checkpoint** — re-packaged and quantized. All use bf16 for worst outliers, unused deepstack layers removed: -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, uses fp8/mxfp8 matmuls on safe MLP layers for speedup on supported hardware -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 quantized variant -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — Full bf16 precision (largest) +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8, uses fp8/mxfp8 matmuls on safe MLP layers for speedup on supported hardware +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 quantized variant +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — Full bf16 precision (largest) **Text Encoder** (prompt enhancement) — shared across all versions: -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) **LoRA (optional)** — the Dev distillation can also be applied to the Full model as a LoRA, allowing you to adjust the distillation strength (by [Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy)): -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — Full rank -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — Pruned variant -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — Alternative checkpoint-based distillation +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — Full rank +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — Pruned variant +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — Alternative checkpoint-based distillation ``` 📂 ComfyUI/ @@ -81,31 +88,55 @@ Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Temp ## HiDream-O1-Image Dev Workflow -### 1. Download Workflow +### HiDream O1 Dev (`image_hidream_o1_dev`) -Go to `Workflow` -> `Browse Templates` -> `Image` and find "HiDream O1 Dev". +Input a text prompt and optional reference images. Generate a high-resolution image (up to 2048x2048) with support for text-to-image, image editing, and subject-driven personalization. -![HiDream-O1-Image Dev Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1_dev.png) +HiDream O1 Dev workflow preview - - Download workflow + + + Open in Comfy Cloud + + + Download JSON or search "HiDream O1 Dev" in Template Library + - - Open in cloud +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 213 · `noir_portrait.png` + + +
+ noir_portrait.png +
+ +**Example output** + +
+ Input image + HiDream O1 Dev example output +
+ +### 1. Download Workflow ### 2. Download Models **Checkpoint (Dev)** — re-packaged and quantized. All use bf16 for worst outliers, unused deepstack layers removed: -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, uses fp8/mxfp8 matmuls on safe MLP layers for speedup on supported hardware -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 quantized variant -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — Full bf16 precision (largest) +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8, uses fp8/mxfp8 matmuls on safe MLP layers for speedup on supported hardware +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 quantized variant +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — Full bf16 precision (largest) **Text Encoder** (prompt enhancement) — shared across all versions: -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) ``` 📂 ComfyUI/ diff --git a/tutorials/image/ideogram/ideogram-v4.mdx b/tutorials/image/ideogram/ideogram-v4.mdx index 813590145..42675a534 100644 --- a/tutorials/image/ideogram/ideogram-v4.mdx +++ b/tutorials/image/ideogram/ideogram-v4.mdx @@ -12,16 +12,24 @@ Ideogram 4.0 is the latest text-to-image model from Ideogram, released as an ope ## Ideogram 4.0 Text-to-Image Workflow - +### Ideogram v4: Text to Image (`image_ideogram4_t2i`) + +Input a text prompt or structured JSON description. Generate an image with precise layout, color, and style control using Ideogram 4.0. + +Ideogram 4.0 text-to-image workflow preview + + + Open in Comfy Cloud - - - Download JSON or search "Ideogram v4: Text to Image" in Template Library + + Download JSON or search "Ideogram v4: Text to Image" in Template Library + + +**Example output** -![Ideogram 4.0 Example Output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) -*Example output from the Ideogram 4.0 model* +![Ideogram 4.0 example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) ### Prompt Format @@ -39,25 +47,23 @@ A note inside the workflow explains: You can find all repackaged model files at [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) on Hugging Face. - + + Diffusion model for Ideogram 4.0 (~13.8 GB). Place in models/diffusion_models/ - - + Unconditional diffusion model for Ideogram 4.0 (~13.8 GB). Place in models/diffusion_models/ - - + Text encoder for Ideogram 4.0 (~8 GB). Place in models/text_encoders/ - - + Text encoder for Ideogram 4.0 (~2 GB). Place in models/text_encoders/ - - + VAE for Ideogram 4.0 (~335 MB). Place in models/vae/ + **Model storage location** @@ -107,4 +113,3 @@ A special live conversation with Mohammad Norouzi (CEO, Ideogram) and Yoland Yan allowFullScreen > - diff --git a/tutorials/image/krea/krea-2.mdx b/tutorials/image/krea/krea-2.mdx index b137edea0..6fdb14e13 100644 --- a/tutorials/image/krea/krea-2.mdx +++ b/tutorials/image/krea/krea-2.mdx @@ -28,7 +28,6 @@ Krea 2 is available in two variants: - **Krea 2 RAW**: the base model with full-step sampling (52 steps). Designed for diversity and malleability, best for fine-tuning and LoRA training. - **Krea 2 Turbo**: an 8-step distilled checkpoint built for fast, high-quality generation. LoRAs trained on RAW apply seamlessly to Turbo. - **Related Links**: - [Krea 2 on Hugging Face (RAW)](https://huggingface.co/krea/Krea-2-Raw) - [Krea 2 on Hugging Face (Turbo)](https://huggingface.co/krea/Krea-2-Turbo) @@ -38,14 +37,18 @@ Krea 2 is available in two variants: ## Krea-2 Turbo text-to-image workflow -Krea-2 Turbo text-to-image workflow +### Krea-2: Text to Image (`image_krea2_turbo_t2i`) + +Generate images from text prompts using Krea 2, a foundation model built for aesthetic quality and creative control. It focuses on rendering expressive, stylistically diverse images. + +Krea-2 Turbo text-to-image workflow preview - + Open in Comfy Cloud - Download JSON or search "Krea-2" in Template Library + Download JSON or search "Krea-2" in Template Library @@ -87,15 +90,15 @@ Krea also released a collection of style LoRAs for Krea 2. Select one in the **C | LoRA | Trigger Word | Recommended Strength | Download | |------|-------------|:-------------------:|----------| -| krea2_darkbrush | monochrome ink wash style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_darkbrush.safetensors) | -| krea2_dotmatrix | monochrome stippling style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_dotmatrix.safetensors) | -| krea2_kidsdrawing | naive expressive sketch style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_kidsdrawing.safetensors) | -| krea2_neondrip | textured abstract style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_neondrip.safetensors) | -| krea2_rainywindow | rainy window style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_rainywindow.safetensors) | -| krea2_retroanime | purple retro anime style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_retroanime.safetensors) | -| krea2_softwatercolor | art deco watercolor style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_softwatercolor.safetensors) | -| krea2_sunsetblur | ethereal motion blur style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_sunsetblur.safetensors) | -| krea2_vintagetarot | vintage tarot style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/resolve/main/loras/krea2_vintagetarot.safetensors) | +| krea2_darkbrush | monochrome ink wash style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_darkbrush.safetensors) | +| krea2_dotmatrix | monochrome stippling style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_dotmatrix.safetensors) | +| krea2_kidsdrawing | naive expressive sketch style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_kidsdrawing.safetensors) | +| krea2_neondrip | textured abstract style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_neondrip.safetensors) | +| krea2_rainywindow | rainy window style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_rainywindow.safetensors) | +| krea2_retroanime | purple retro anime style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_retroanime.safetensors) | +| krea2_softwatercolor | art deco watercolor style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_softwatercolor.safetensors) | +| krea2_sunsetblur | ethereal motion blur style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_sunsetblur.safetensors) | +| krea2_vintagetarot | vintage tarot style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_vintagetarot.safetensors) | Place the `.safetensors` files in `ComfyUI/models/loras/`. @@ -104,13 +107,13 @@ Place the `.safetensors` files in `ComfyUI/models/loras/`. For local use, download the ComfyUI-optimized model files from [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2). - + krea2_turbo_fp8_scaled.safetensors: Turbo FP8 (recommended for most users) - + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B text encoder - + qwen_image_vae.safetensors @@ -137,17 +140,39 @@ Other model variants (BF16, NVFP4, MXFP8) are also available for users with high ## Krea-2 Turbo style reference workflow -Krea-2 Turbo style reference workflow +### Krea-2 Int8: Image Style Reference (`image_krea2_turbo_int8_image_style_reference`) + +Generate images with the Krea-2 Turbo model while referencing the style of 1–2 uploaded images, using the high-performance Int8 Convrot format for fast inference. + +Krea-2 Turbo style reference workflow preview - + Open in Comfy Cloud - Download JSON or search "Krea-2 Style Reference" in Template Library + Download JSON or search "Krea-2 Style Reference" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 69 · `krea2_reference_image.png` +
+ krea2_reference_image.png +
+ +**Example output** + +![Krea-2 style reference example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_krea2_turbo_int8_image_style_reference.png) + The style reference workflow builds on the Krea-2 Turbo pipeline by adding reference image conditioning. Upload one or more reference images to influence the aesthetic style, mood, and visual direction of the generated output. The workflow is organized into a few parts: @@ -168,16 +193,16 @@ The workflow is organized into a few parts: This style reference workflow uses a dedicated diffusion model and LoRA. The text encoder and VAE are shared with the text-to-image workflow. - + krea2_turbo_int8_convrot.safetensors - + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B text encoder - + qwen_image_vae.safetensors - + krea2_style_reference.safetensors diff --git a/tutorials/image/lens/lens.mdx b/tutorials/image/lens/lens.mdx index 578619123..42d94b703 100644 --- a/tutorials/image/lens/lens.mdx +++ b/tutorials/image/lens/lens.mdx @@ -29,11 +29,15 @@ Both the standard and Turbo variants use a **Subgraph** node to manage the text- This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -### Lens +### Lens: Text to Image (`image_lens_t2i`) + +Input a text prompt and select resolution and aspect ratio. Generate a high-quality image using the efficient Lens text-to-image model. + +Lens text-to-image workflow preview - Download JSON or search "Lens" in Template Library + Download JSON or search "Lens" in Template Library {/* TODO: Enable Cloud template when Lens is available on Comfy Cloud */} {/**/} @@ -41,6 +45,10 @@ Both the standard and Turbo variants use a **Subgraph** node to manage the text- {/**/} +**Example output** + +![Lens text-to-image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_t2i.png) + #### Get started @@ -51,17 +59,15 @@ Both the standard and Turbo variants use a **Subgraph** node to manage the text- 3. Select the **Lens** workflow 4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** -#### Example output - -Lens text-to-image example output +### Lens Turbo: Text to Image (`image_lens_turbo_t2i`) -### Lens Turbo +Input a text prompt and select resolution, aspect ratio, and inference steps. Generate a high-quality image using the Lens text-to-image model. -Lens Turbo is a distilled variant that generates images in fewer sampling steps for faster inference. +Lens Turbo text-to-image workflow preview - Download JSON or search "Lens Turbo" in Template Library + Download JSON or search "Lens Turbo" in Template Library {/* TODO: Enable Cloud template when Lens Turbo is available on Comfy Cloud */} {/**/} @@ -69,6 +75,10 @@ Lens Turbo is a distilled variant that generates images in fewer sampling steps {/**/} +**Example output** + +![Lens Turbo text-to-image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_turbo_t2i.png) + #### Get started 1. Update ComfyUI to the latest version @@ -77,28 +87,24 @@ Lens Turbo is a distilled variant that generates images in fewer sampling steps 3. Select the **Lens Turbo** workflow 4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** -#### Example output - -Lens Turbo text-to-image example output - ## Lens model downloads You can find all model files at [Comfy-Org/Lens](https://huggingface.co/Comfy-Org/Lens) on Hugging Face. - + Diffusion model for Lens (BF16). - + Diffusion model for Lens Turbo (BF16). - + Text encoder (GPT-OSS-20B) shared by both Lens and Lens Turbo. - + VAE (FLUX.2) shared by both Lens and Lens Turbo. diff --git a/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 703c71d71..13d39940f 100644 --- a/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -22,15 +22,25 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## NewBie-image text-to-image workflow +### NewBie Exp0.1: Anime Generation (`image_newbieimage_exp0_1-t2i`) + +Generate detailed anime-style images with NewBie Exp0.1's Next-DiT architecture. Supports XML structured prompts for better multi-character scenes and attribute binding. + +NewBie-image text-to-image workflow preview + - - Download JSON or search "NewBie-image" in Template Library + + Open in Comfy Cloud - - Open in cloud + + Download JSON or search "NewBie-image" in Template Library +**Example output** + +![NewBie-image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_newbieimage_exp0_1-t2i.png) + ## Model links diff --git a/tutorials/image/omnigen/omnigen2.mdx b/tutorials/image/omnigen/omnigen2.mdx index b99d1d4a4..b15c5b4c1 100644 --- a/tutorials/image/omnigen/omnigen2.mdx +++ b/tutorials/image/omnigen/omnigen2.mdx @@ -55,15 +55,26 @@ File save location: ## ComfyUI OmniGen2 Text-to-Image Workflow -### 1. Download Workflow File +### OmniGen2: Text to Image (`image_omnigen2_t2i`) + +Generate high-quality images from text prompts using OmniGen2's unified 7B multimodal model with dual-path architecture. + +OmniGen2 text-to-image workflow preview - + - Open and run this workflow directly in Comfy Cloud. + Open and run this workflow directly in Comfy Cloud + + + Download JSON or search "OmniGen2" in Template Library -![Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) +**Example output** + +![OmniGen2 text-to-image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_omnigen2_t2i.png) + +### 1. Download Workflow File ### 2. Complete Workflow Step by Step @@ -83,20 +94,45 @@ Please follow the numbered steps in the image for step-by-step confirmation to e ## ComfyUI OmniGen2 Image Editing Workflow -OmniGen2 has rich image editing capabilities and supports adding text to images +OmniGen2 has rich image editing capabilities and supports adding text to images. -### 1. Download Workflow File +### OmniGen2 Image Edit (`image_omnigen2_image_edit`) + +Edit images with natural language instructions using OmniGen2's advanced image editing capabilities and text rendering support. + +OmniGen2 image edit workflow preview - + - Open and run this workflow directly in Comfy Cloud. + Open and run this workflow directly in Comfy Cloud + + + Download JSON or search "OmniGen2 Image Edit" in Template Library -![Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 16 · `image_omnigen2_image_edit_input_image.png` + + -Download the image below, which we will use as the input image. -![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/input_fairy.png) +
+ image_omnigen2_image_edit_input_image.png +
+ +**Example output** + +
+ Input image + OmniGen2 image edit example output +
+ +### 1. Download Workflow File ### 2. Complete Workflow Step by Step diff --git a/tutorials/image/ovis/ovis-image.mdx b/tutorials/image/ovis/ovis-image.mdx index 2a2b546c0..39c5f259c 100644 --- a/tutorials/image/ovis/ovis-image.mdx +++ b/tutorials/image/ovis/ovis-image.mdx @@ -20,12 +20,18 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Ovis-Image text-to-image workflow +### Ovis-Image Text to Image (`image_ovis_text_to_image`) + +Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering in generated images. Designed to operate efficiently under stringent computational constraints. + +Ovis-Image text-to-image workflow preview + - + Open in Comfy Cloud - Download JSON or search "Ovis image" in Template Library + Download JSON or search "Ovis image" in Template Library diff --git a/tutorials/image/qwen/qwen-image-2512.mdx b/tutorials/image/qwen/qwen-image-2512.mdx index 24c3b16e9..ff8a9feb4 100644 --- a/tutorials/image/qwen/qwen-image-2512.mdx +++ b/tutorials/image/qwen/qwen-image-2512.mdx @@ -36,6 +36,12 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' +### Qwen Image 2512 (`image_qwen_Image_2512`) + +Text-to-image model with enhanced human realism, finer natural details, and improved text rendering. + +Qwen-Image-2512 workflow preview + Open in Comfy Cloud @@ -45,9 +51,13 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' +**Example output** + +![Qwen-Image-2512 example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_qwen_Image_2512.png) + ### 1. Workflow file -After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow below into ComfyUI to load it. +After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow preview above into ComfyUI to load it. The workflow includes two subgraphs: - **Text to Image (Qwen-Image 2512)**: Standard 50-step generation @@ -88,4 +98,3 @@ The workflow includes two subgraphs: │ └── qwen_image_vae.safetensors ``` - diff --git a/tutorials/image/qwen/qwen-image-edit-2511.mdx b/tutorials/image/qwen/qwen-image-edit-2511.mdx index 0691c6a5f..113028dbd 100644 --- a/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -27,9 +27,11 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' -### 1. Workflow file +### Qwen Image Edit 2511 - Material Replacement (`image_qwen_image_edit_2511`) + +Replace materials in objects (for example, furniture) by combining reference images with Qwen-Image-Edit-2511. -After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow below into ComfyUI to load it. +Qwen-Image-Edit-2511 workflow preview @@ -40,6 +42,35 @@ After updating ComfyUI, you can find the workflow file from the templates, or dr +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 41 · `leather_sofa.png` + + + `LoadImage` node 83 · `texture_fur.png` + + + +
+ leather_sofa.png + texture_fur.png +
+ +**Example output** + +
+ Input image + Qwen-Image-Edit-2511 example output +
+ +### 1. Workflow file + +After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow preview above into ComfyUI to load it. + ### 2. Model download **Text Encoders** @@ -73,4 +104,3 @@ After updating ComfyUI, you can find the workflow file from the templates, or dr │ └── qwen_image_vae.safetensors ``` - diff --git a/tutorials/image/qwen/qwen-image-edit.mdx b/tutorials/image/qwen/qwen-image-edit.mdx index ffd046d2d..20123238a 100644 --- a/tutorials/image/qwen/qwen-image-edit.mdx +++ b/tutorials/image/qwen/qwen-image-edit.mdx @@ -35,22 +35,43 @@ Features include: -### 1. Workflow File +### Qwen Image Edit (`image_qwen_image_edit`) + +Edit images with precise bilingual text editing and dual semantic/appearance editing capabilities using Qwen-Image-Edit's 20B MMDiT model. -After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow below into ComfyUI to load it. -![Qwen-image Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) +Qwen-Image-Edit workflow preview - - Download JSON or search "image_qwen_image_edit" in Template Library - Run this workflow on Cloud GPUs with zero setup + + Download JSON or search "Qwen Image Edit" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 78 · `image_qwen_image_edit_input_image.png` + -Download the image below as input -![Qwen-image Text-to-Image Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) +
+ image_qwen_image_edit_input_image.png +
+ +**Example output** + +
+ Input image + Qwen-Image-Edit example output +
+ +### 1. Workflow File ### 2. Model Download diff --git a/tutorials/image/qwen/qwen-image-layered.mdx b/tutorials/image/qwen/qwen-image-layered.mdx index c337dc653..e6644097d 100644 --- a/tutorials/image/qwen/qwen-image-layered.mdx +++ b/tutorials/image/qwen/qwen-image-layered.mdx @@ -21,23 +21,40 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Qwen-Image-Layered workflow -| -| -| Download the JSON workflow file -| -| -| -| Run ComfyUI online with zero setup -| -| - +### Qwen-Image-Layered Decomposition (`image_qwen_image_layered`) + +Decompose an image into editable RGBA layers for recolor, replace, resize, and reposition workflows. + +Qwen-Image-Layered workflow preview + + + + Download the JSON workflow file + + + Run ComfyUI online with zero setup + + + +**Input materials** + +Upload this file to `LoadImage` node **74**: + + + + `LoadImage` node 74 · `coastal_smiling_woman.png` + + + +![coastal_smiling_woman.png](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/input/coastal_smiling_woman.png) + ## Model links **text_encoders** -|- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) +- [qwen_2.5_vl_7b_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors) **diffusion_models** diff --git a/tutorials/image/qwen/qwen-image.mdx b/tutorials/image/qwen/qwen-image.mdx index 1dc15807a..35360f03b 100644 --- a/tutorials/image/qwen/qwen-image.mdx +++ b/tutorials/image/qwen/qwen-image.mdx @@ -43,18 +43,29 @@ Currently Qwen-Image has multiple ControlNet support options available: allowFullScreen > - ## Qwen-Image Native Workflow Example +### Qwen-Image: Text to Image (`image_qwen_image`) + +Generate images with exceptional multilingual text rendering and editing capabilities using Qwen-Image's 20B MMDiT model. + +Qwen-Image text-to-image workflow preview + + Open in Comfy Cloud - + + Download JSON or search "Qwen-Image" in Template Library +**Example output** + +![Qwen-Image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_qwen_image.png) + There are three different models used in the workflow attached to this document: 1. Qwen-Image original model fp8_e4m3fn 2. 8-step accelerated version: Qwen-Image original model fp8_e4m3fn with lightx2v 8-step LoRA @@ -69,7 +80,6 @@ GPU: RTX4090D 24GB | fp8_e4m3fn with lightx2v 8-step LoRA | 86% | ≈ 55s | ≈ 34s | | Distilled fp8_e4m3fn | 86% | ≈ 69s | ≈ 36s | - ### 1. Workflow File After updating ComfyUI, you can find the workflow file in the templates, or drag the workflow below into ComfyUI to load it. @@ -151,20 +161,36 @@ Qwen_image_distill This is a ControlNet model, so you can use it as normal ControlNet. +### Qwen-Image InstantX Union ControlNet (`image_qwen_image_instantx_controlnet`) + +Generate images with Qwen-Image InstantX ControlNet, supporting canny, soft edge, depth, and pose. + +Qwen-Image InstantX ControlNet workflow preview + + Open in Comfy Cloud - + + Download JSON or search "Qwen-Image InstantX ControlNet" in Template Library -### 1. Workflow and Input Images +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 71 · `image_qwen_image_instantx_controlnet_input_image.jpg` + + -Download the image below and drag it into ComfyUI to load the workflow -![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) +
+ image_qwen_image_instantx_controlnet_input_image.jpg +
-Download the image below as input -![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/input.jpg) +**## 1. Workflow and Input Images ### 2. Model Links 1. InstantX Controlnet @@ -205,27 +231,38 @@ ComfyUI/ ## Qwen Image ControlNet DiffSynth-ControlNets Model Patches Workflow +### Qwen-Image ControlNet Model Patch (`image_qwen_image_controlnet_patch`) + +Control image generation using Qwen-Image ControlNet models. Supports canny, depth, and inpainting controls through model patching. + +Qwen-Image ControlNet model patch workflow preview + + Open in Comfy Cloud - + + Download JSON or search "Qwen-Image ControlNet Patch" in Template Library -This model is actually not a ControlNet, but a Model patch that supports three different control modes: canny, depth, and inpaint. - -Original model address: [DiffSynth-Studio/Qwen-Image ControlNet](https://www.modelscope.cn/collections/Qwen-Image-ControlNet-6157b44e89d444) -Comfy Org rehost address: [Qwen-Image-DiffSynth-ControlNets/model_patches](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/tree/main/split_files/model_patches) +**Input materials** +Upload this file to the matching `LoadImage` node: -### 1. Workflow and Input Images + + + `LoadImage` node 71 · `image_qwen_image_controlnet_patch_input_image.png` + + -Download the image below and drag it into ComfyUI to load the corresponding workflow -![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/image_qwen_image_controlnet_patch.png) +
+ image_qwen_image_controlnet_patch_input_image.png +
-Download the image below as input: +### 1. Workflow and Input Images -![input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-controlnet-model-patch/input.png) +After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow preview above into ComfyUI to load it. ### 2. Model Links @@ -235,7 +272,6 @@ Other models are the same as the Qwen-Image basic workflow. You only need to dow - [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) - [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) - ### 3. Workflow Usage Instructions Currently, diffsynth has three patch models: Canny, Depth, and Inpaint. @@ -275,24 +311,48 @@ For the Inpaint model, it requires using the [Mask Editor](/interface/maskeditor ## Qwen Image Union ControlNet LoRA Workflow +### Qwen-Image Union Control (`image_qwen_image_union_control_lora`) + +Generate images with precise structural control using Qwen-Image's unified ControlNet LoRA. Supports multiple control types including canny, depth, lineart, softedge, normal, and openpose. + +Qwen-Image Union Control workflow preview + + Open in Comfy Cloud + + + Download JSON or search "Qwen-Image Union Control" in Template Library - + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 73 · `image_qwen_image_union_control_lora_input_image.png` +
+ image_qwen_image_union_control_lora_input_image.png +
+ +**Example output** + +
+ Input image + Qwen-Image Union Control example output +
+ Original model address: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) Comfy Org rehost address: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): Image structure control LoRA supporting canny, depth, pose, lineart, softedge, normal, openpose ### 1. Workflow and Input Images -Download the image below and drag it into ComfyUI to load the workflow -![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/image_qwen_image_union_control_lora.png) - -Download the image below as input - -![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-union-control-lora/input.png) +After updating ComfyUI, you can find the workflow file from the templates, or drag the workflow preview above into ComfyUI to load it. ### 2. Model Links diff --git a/tutorials/image/z-image/z-image-turbo.mdx b/tutorials/image/z-image/z-image-turbo.mdx index 4757e5f6c..867695ab0 100644 --- a/tutorials/image/z-image/z-image-turbo.mdx +++ b/tutorials/image/z-image/z-image-turbo.mdx @@ -25,18 +25,28 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - [GitHub](https://github.com/Tongyi-MAI/Z-Image) - [Hugging Face](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo) + + ## Z-Image-Turbo text-to-image workflow +### Z-Image-Turbo: Text to Image (`image_z_image_turbo`) + +An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer, supports English & Chinese. + +Z-Image-Turbo text-to-image workflow preview + - - Download the Z-Image-Turbo text-to-image workflow JSON file. + + Run this workflow directly on Comfy Cloud - - Run this workflow directly on ComfyUI Cloud. + + Download the Z-Image-Turbo text-to-image workflow JSON file - +**Example output** + +![Z-Image-Turbo example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_z_image_turbo.png) ### Z-Image-Turbo model downloads @@ -66,16 +76,35 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Z-Image-Turbo Fun Union ControlNet workflow -This workflow uses the Z-Image-Turbo Fun Union ControlNet model to generate images with ControlNet guidance. It applies Canny edge detection to a reference image and uses the ControlNet to guide the generation process. +### Z-Image-Turbo Fun Union ControlNet (`image_z_image_turbo_fun_union_controlnet`) + +Multi-control ControlNet supporting Canny, HED, Depth, Pose, and MLSD for Z-Image-Turbo. + +Z-Image-Turbo Fun Union ControlNet workflow preview - - Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file. + + Run this workflow directly on Comfy Cloud - - Run this workflow directly on ComfyUI Cloud. + + Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 58 · `image_z_image_turbo_fun_union_controlnet_input_image.png` + + + +
+ image_z_image_turbo_fun_union_controlnet_input_image.png +
+ ### Additional model for ControlNet diff --git a/tutorials/image/z-image/z-image.mdx b/tutorials/image/z-image/z-image.mdx index 041d56798..78efa25db 100644 --- a/tutorials/image/z-image/z-image.mdx +++ b/tutorials/image/z-image/z-image.mdx @@ -23,29 +23,40 @@ Z-Image (Base) is the non-distilled foundation model designed for community-driv ## Z-Image text-to-image workflow - - Download the Z-Image text-to-image workflow JSON file. - + + +### Z-Image: Text to Image (`image_z_image`) + +Foundation for creative freedom. Diverse aesthetics with exceptional photorealistic quality; ideal for fine-tuning; responsive to negative prompts; high generation diversity. + +Z-Image text-to-image workflow preview - - Run this workflow directly on ComfyUI Cloud. + + + Run this workflow directly on Comfy Cloud + + Download the Z-Image text-to-image workflow JSON file + + - +**Example output** + +![Z-Image example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_z_image.png) ## Z-Image model downloads - + + Text encoder for Z-Image. - - + Diffusion model for Z-Image. - - + VAE for Z-Image. + **Model Storage Location** diff --git a/tutorials/llm/gemma4/gemma4.mdx b/tutorials/llm/gemma4/gemma4.mdx index 6c000333d..0c76ed157 100644 --- a/tutorials/llm/gemma4/gemma4.mdx +++ b/tutorials/llm/gemma4/gemma4.mdx @@ -30,17 +30,36 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## Available workflow -### Gemma 4: Text Generation +### Gemma4: Text Generation (`llm_gemma4_text_gen`) - - Download JSON or search "Gemma 4 Text Generation" in Template Library - +Input your text prompt and optionally an image, audio, or video. Generate text output with configurable reasoning, coding, and multilingual support. + +Gemma 4 text generation workflow preview - + + Open in Comfy Cloud + + Download JSON or search "Gemma4: Text Generation" in Template Library + + + +**Input materials** -![Gemma 4 Text Generation Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_gemma4_text_gen-1.webp) +Upload these optional files to the matching nodes: + + + + `LoadImage` node 2 · `the_lily_veil.png` + + + `LoadAudio` node 5 · `voice_demo.mp3` + + + `LoadVideo` node 6 · `video_wan_vace_inpainting_input_video.mp4` + + This workflow demonstrates the core **text generation** capabilities of Gemma 4. It accepts an optional image, audio file, or video as additional context alongside your text prompt, and generates natural language output — with support for reasoning, coding, and multilingual prompts. @@ -67,17 +86,19 @@ This workflow demonstrates the core **text generation** capabilities of Gemma 4. Gemma 4 models are loaded as text encoders in ComfyUI. Download the relevant model file and place it in the correct directory: - + + Fast, lightweight. Recommended for consumer GPUs. - + Balanced performance. The default model in the workflow. Browse all Gemma 4 model weights. + Place the downloaded `.safetensors` file in: diff --git a/tutorials/llm/qwen/qwen3.mdx b/tutorials/llm/qwen/qwen3.mdx index 07d821261..4df6a3d00 100644 --- a/tutorials/llm/qwen/qwen3.mdx +++ b/tutorials/llm/qwen/qwen3.mdx @@ -17,7 +17,6 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - **ComfyUI native** — works with the built-in `TextGenerate` node, no custom nodes needed - **Lightweight** — shares the same text encoder format as Qwen3.5, available in 2B, 4B, and 9B variants to fit different hardware - ## Use Cases Qwen 3.0 is well-suited for tasks that require structured text generation and intelligent reasoning within a ComfyUI workflow: @@ -29,17 +28,20 @@ Qwen 3.0 is well-suited for tasks that require structured text generation and in ## Available workflow -### Qwen 3.0: Text Generation +### Qwen3.0: Text Generation (`llm_qwen3_text_gen`) - - Download JSON or search "Qwen 3.0 Text Generation" in Template Library - +Input a text prompt to generate detailed, reasoned responses using the Qwen3-4B-Thinking model. + +Qwen 3.0 text generation workflow preview - + + Open in Comfy Cloud - -![Qwen 3.0 Text Generation Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_text_gen-1.webp) + + Download JSON or search "Qwen3.0: Text Generation" in Template Library + + This workflow demonstrates the core **text generation** capabilities of Qwen 3.0. It accepts a text prompt and generates detailed, structured responses using the model's built-in reasoning capabilities. @@ -63,17 +65,19 @@ This workflow demonstrates the core **text generation** capabilities of Qwen 3.0 Qwen 3.0 models are loaded as text encoders in ComfyUI. The model files are shared with Qwen3.5 — download the variant that best fits your hardware: - + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - + Balanced size and quality. Recommended for most consumer GPUs. - + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + Place the downloaded `.safetensors` file in: diff --git a/tutorials/llm/qwen/qwen3_5.mdx b/tutorials/llm/qwen/qwen3_5.mdx index 884dd1083..00f680aff 100644 --- a/tutorials/llm/qwen/qwen3_5.mdx +++ b/tutorials/llm/qwen/qwen3_5.mdx @@ -18,7 +18,6 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' - **ComfyUI native** — works with the built-in `TextGenerate` node, no custom nodes needed - **Lightweight** — 4B parameter model, suitable for consumer GPUs - ## Use Cases Qwen3.5 excels in scenarios where combining visual understanding with text generation adds value to a ComfyUI workflow: @@ -31,17 +30,34 @@ Qwen3.5 excels in scenarios where combining visual understanding with text gener ## Available workflow -### Qwen3.5: Text Generation +### Qwen3.5: Text Generation (`llm_qwen3_5_text_gen`) - - Download JSON or search "Qwen3.5 Text Generation" in Template Library - +Use the Qwen3.5 model to analyze an input image and generate descriptive text prompts. This workflow performs image captioning and reverse prompt engineering. - +Qwen3.5 text generation workflow preview + + + Open in Comfy Cloud + + Download JSON or search "Qwen3.5: Text Generation" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 2 · `man_with_red_hat.png` + + -![Qwen3.5 Text Generation Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_5_text_gen-1.webp) +
+ Input image +
This workflow demonstrates the **text generation and image understanding** capabilities of Qwen3.5. It accepts a text prompt and an optional image, and generates descriptive text or structured analysis based on the input. @@ -65,17 +81,19 @@ This workflow demonstrates the **text generation and image understanding** capab Qwen3.5 models are loaded as text encoders in ComfyUI. Choose the variant that best suits your hardware: - + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - + Balanced size and quality. Recommended for most consumer GPUs. - + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + Place the downloaded `.safetensors` file in: diff --git a/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/tutorials/partner-nodes/beeble/beeble-switchx.mdx index 569670d5c..949c8c91d 100644 --- a/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -21,7 +21,7 @@ Replace the background environment and relight your subject with a reference ima Beeble SwitchX Image Edit Workflow - + Open in Comfy Cloud @@ -62,7 +62,7 @@ Apply environment relighting to a video while preserving the original motion and Beeble SwitchX Video Edit Workflow - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/bria/background-removal.mdx b/tutorials/partner-nodes/bria/background-removal.mdx index 9ba6b7bc9..8af904caa 100644 --- a/tutorials/partner-nodes/bria/background-removal.mdx +++ b/tutorials/partner-nodes/bria/background-removal.mdx @@ -12,71 +12,143 @@ Bria's AI background processing models are now available in ComfyUI through Part -## Image Background Removal +### Bria: Remove Image Background (`utility_bria_remove_image_background`) -Remove the background from any image using Bria's AI service. The result is returned as an image with a transparent background. +Upload an image to automatically remove its background using the Bria API. The result is returned as an image with a transparent background. + +Bria: Remove Image Background workflow preview - - Download the workflow. + + Open in Comfy Cloud + + + Download JSON or search "Bria: Remove Image Background" in Template Library - - Try it on Comfy Cloud. + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 26 · `denim_girl.png` +
+ Input image + Bria: Remove Image Background example output +
+ ## Video Background Processing Bria's video background processing nodes let you remove, replace, or apply chroma-key effects to video backgrounds. The input video must be 60 seconds or shorter. -### Remove Video Background +### Bria: Remove Video Background (`api_bria_remove_video_background`) Replace a video's background with a solid color of your choice. +Bria: Remove Video Background workflow preview + - - Download the workflow. + + Open in Comfy Cloud - - Try it on Comfy Cloud. + + Download JSON or search "Bria: Remove Video Background" in Template Library -### Transparent Video Background +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 4 · `spear_warrior.mp4` + + + +### Bria: Remove Video Background (Transparent) (`api_bria_remove_video_background_transparent`) Remove the background from a video and output it with transparency. +Bria: Remove Video Background (Transparent) workflow preview + - - Download the workflow. + + Open in Comfy Cloud + + + Download JSON or search "Bria: Remove Video Background (Transparent)" in Template Library - - Try it on Comfy Cloud. + + +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 5 · `ad_video_demo.mp4` -### Video Green Screen +### Bria: Video Green Screen (`api_bria_video_green_screen`) Apply a professional chroma-key green or blue screen effect to your video, with multiple preset options including broadcast_green, chroma_green, and blue_screen. +Bria: Video Green Screen workflow preview + - - Download the workflow. + + Open in Comfy Cloud + + + Download JSON or search "Bria: Video Green Screen" in Template Library - - Try it on Comfy Cloud. + + +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 5 · `investigator.mp4` -### Replace Video Background +### Bria: Video Replace Background (`api_bria_video_replace_background`) Replace a video's background with a custom image or another video. +Bria: Video Replace Background workflow preview + - - Download the workflow. + + Open in Comfy Cloud - - Try it on Comfy Cloud. + + Download JSON or search "Bria: Video Replace Background" in Template Library + +**Input materials** + +Upload these files to the matching `LoadVideo` and `LoadImage` nodes: + + + + `LoadVideo` node 4 · `stained_window_vintage_woman.mp4` + + + `LoadImage` node 8 · `gothic_hall_light_rays.png` + + + +
+ gothic_hall_light_rays.png +
diff --git a/tutorials/partner-nodes/bria/fibo.mdx b/tutorials/partner-nodes/bria/fibo.mdx index 3c76ba933..29324de7e 100644 --- a/tutorials/partner-nodes/bria/fibo.mdx +++ b/tutorials/partner-nodes/bria/fibo.mdx @@ -21,23 +21,64 @@ FIBO Edit is Bria AI's JSON-native image editing model, now available in ComfyUI -## FIBO Edit Workflows +### Bria: Image Edit (`api_bria_image_edit`) - - Download the Bria FIBO Image Edit workflow. +FIBO enables precise, predictable image editing through structured JSON control over lighting, composition, and camera settings. + +Bria: Image Edit workflow preview + + + + Open in Comfy Cloud + + Download JSON or search "Bria: Image Edit" in Template Library + + + +**Input materials** - - Download the Bria FIBO Image Outpainting workflow. +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 5 · `sunlit_rooftop_capture.png` + + +
+ Input image + Bria: Image Edit example output +
+ +### Bria: Image Outpainting (`api_bria_image_outpainting`) - - Try the Image Edit workflow instantly on Comfy Cloud. +A professional Bria-powered outpainting template designed for seamless image extension and edge-aware scene expansion. + +Bria: Image Outpainting workflow preview + + + + Open in Comfy Cloud + + Download JSON or search "Bria: Image Outpainting" in Template Library + + + +**Input materials** - - Try the Image Outpainting workflow instantly on Comfy Cloud. +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 11 · `baby_otter.png` + + +
+ Input image +
## Example outputs @@ -50,4 +91,3 @@ FIBO Edit excels at various editing tasks: - **Text adjustment**: Adjust the text in an image - **Art style transformation**: Re-imagine your image in a different art style - diff --git a/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index 31bd7503d..e221ceee6 100644 --- a/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -26,51 +26,75 @@ With these capabilities, Seed Audio 1.0 fits a wide range of use cases including ## Available workflows -### Seed Audio 1.0: Text to Audio +### Seed Audio 1.0: Text to Audio (`api_bytedance_seed_audio1_0_t2a`) -Generate audio directly from a text prompt. Describe the voice, ambience, sound effects, and dialogue in the prompt. +Input a text prompt to generate speech, dialogue, background music, and sound effects in one audio file. -![Seed Audio 1.0 Text to Audio preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/api_bytedance_seed_audio1_0_t2a.png) +Seed Audio 1.0: Text to Audio workflow preview - + Open in Comfy Cloud - Download JSON or search "Seed Audio 1.0" in Template Library + Download JSON or search "Seed Audio 1.0: Text to Audio" in Template Library -### Seed Audio 1.0: Text + Audio to Audio +### Seed Audio 1.0: Text + Audio to Audio (`api_bytedance_seed_audio1_0_ta2a`) -Generate audio with voice cloning. Connect a reference audio clip and the model clones the voice and style, applying it to your text prompt. +Upload a reference audio clip and write a text prompt to clone the voice into a new scene. -![Seed Audio 1.0 Text + Audio to Audio preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/api_bytedance_seed_audio1_0_ta2a.png) +Seed Audio 1.0: Text + Audio to Audio workflow preview - + Open in Comfy Cloud - Download JSON or search "Seed Audio 1.0" in Template Library + Download JSON or search "Seed Audio 1.0: Text + Audio to Audio" in Template Library -### Seed Audio 1.0: Text + Image to Audio +**Input materials** -Derive a voice from a character image. Connect a character image and the model generates audio with a voice style matching the visual subject. +Upload this file to the matching `LoadAudio` node: -![Seed Audio 1.0 Text + Image to Audio preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/api_bytedance_seed_audio1_0_ti2a.png) + + + `LoadAudio` node 8 · `seed_audio_ref_audio1.mp3` + + + +### Seed Audio 1.0: Text + Image to Audio (`api_bytedance_seed_audio1_0_ti2a`) + +Upload a character image and write a text prompt to generate audio with a voice style matching the visual subject. + +Seed Audio 1.0: Text + Image to Audio workflow preview - + Open in Comfy Cloud - Download JSON or search "Seed Audio 1.0" in Template Library + Download JSON or search "Seed Audio 1.0: Text + Image to Audio" in Template Library +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 6 · `girl_and_fish.png` + + + +
+ girl_and_fish.png +
+ ## How to use Seed Audio 1.0 in ComfyUI Seed Audio 1.0 ships as the **ByteDanceSeedAudio** built-in node. You can find it in the node menu under ByteDance. diff --git a/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index 0683d1737..15b2b7b89 100644 --- a/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -24,7 +24,7 @@ Seedream 5.0 lite is the latest image generation model from BytePlus. It is the Seedream 5.0 Lite Image Edit workflow preview - + Open in Comfy Cloud @@ -77,7 +77,7 @@ Download these sample input images to try the workflow: Seedream 5.0 Lite Text-to-Image workflow preview - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index 114b12811..de6cadeed 100644 --- a/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -27,7 +27,7 @@ Generate a high-quality image from a text prompt, with Seedream 5.0 Pro handling Seedream 5.0 Pro Text-to-Image workflow preview - + Open in Comfy Cloud @@ -44,7 +44,7 @@ Edit existing images with text instructions. Change objects, swap styles, adjust Seedream 5.0 Pro Image Edit workflow preview - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/tutorials/partner-nodes/google/gemini-omni-flash.mdx index 51ae47d33..3bd90b9c1 100644 --- a/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -25,7 +25,7 @@ Gemini Omni Flash is Google DeepMind's high-quality, cost-efficient video genera ### Text to Video - + Open in Comfy Cloud @@ -40,7 +40,7 @@ Generate cinematic video from natural language prompts. Transform text descripti ### Image to Video - + Open in Comfy Cloud @@ -61,7 +61,7 @@ Generate a video from two images using Gemini Omni Flash. Interpret natural lang ### Video Edit - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/google/gemini.mdx b/tutorials/partner-nodes/google/gemini.mdx index 093b5aae4..81cc72861 100644 --- a/tutorials/partner-nodes/google/gemini.mdx +++ b/tutorials/partner-nodes/google/gemini.mdx @@ -14,22 +14,36 @@ In this guide, we will walk you through completing the corresponding conversatio -## Google Gemini Chat Workflow +### Google Gemini (`api_google_gemini`) -### 1. Workflow File Download +Experience Google's multimodal AI with Gemini's reasoning capabilities. -Please download the Json file below and drag it into ComfyUI to load the corresponding workflow. +Google Gemini workflow preview - - Download Json Format Workflow File - - + Open in Comfy Cloud + + Download JSON or search "Google Gemini" in Template Library + -### 2. Complete the Workflow Execution Step by Step +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 2 · `example.png` + + + +
+ example.png +
+ +### Complete the workflow execution step by step ![OpenAI Chat Step Guide](/images/tutorial/api_nodes/google/tripo_image_to_model_step_guide.jpg) diff --git a/tutorials/partner-nodes/google/nano-banana-2-lite.mdx b/tutorials/partner-nodes/google/nano-banana-2-lite.mdx index f506b46d7..89f5302f4 100644 --- a/tutorials/partner-nodes/google/nano-banana-2-lite.mdx +++ b/tutorials/partner-nodes/google/nano-banana-2-lite.mdx @@ -20,42 +20,53 @@ Nano Banana 2 Lite is Google DeepMind's fastest and most cost-efficient Gemini I - **In-image text rendering**: Draft copy and render legible text directly into generated images for localized ad variations - **Configurable model selection**: Choose between Nano Banana 2 Lite, Nano Banana 2, and Nano Banana Pro within the same node -## Workflows +### Nano Banana 2 Lite: Text to Image (`api_nano_banana_2_lite_t2i`) -### Text to Image +Generate images from text descriptions using the ultra-fast Gemini 3.1 Flash-Lite Image model. No file inputs required. + +Nano Banana 2 Lite: Text to Image workflow preview - + Open in Comfy Cloud - Download JSON or search "Nano Banana 2 Lite" in Template Library + Download JSON or search "Nano Banana 2 Lite: Text to Image" in Template Library -Generate images from text descriptions. No file inputs required — just enter your prompt and the workflow produces one or more generated images with support for interleaved text and image inputs. +**Example output** + +![Nano Banana 2 Lite: Text to Image example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_nano_banana_2_lite_t2i.png) + +### Nano Banana 2 Lite: Image Edit (`api_nano_banana_2_lite_image_edit`) -![Nano Banana 2 Lite Text to Image](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_nano_banana_2_lite_t2i.png) +Edit an image using Nano Banana 2 Lite. Upload an image and provide a text instruction to produce a side-by-side comparison of the original and edited result. -### Image Edit +Nano Banana 2 Lite: Image Edit workflow preview - + Open in Comfy Cloud - Download JSON or search "Nano Banana 2 Lite" in Template Library - - - Get the example input image for this workflow + Download JSON or search "Nano Banana 2 Lite: Image Edit" in Template Library -Upload an image and provide a text instruction to edit it. The workflow produces a side-by-side comparison of the original and edited result. +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 8 · `red_f1_track_car.png` + + -
- Nano Banana 2 Lite Image Edit - Input - Nano Banana 2 Lite Image Edit - Output +
+ Input image + Nano Banana 2 Lite: Image Edit example output
## Which model should you pick? diff --git a/tutorials/partner-nodes/google/nano-banana-2.mdx b/tutorials/partner-nodes/google/nano-banana-2.mdx index d6f0cb7b5..8e43ac4b7 100644 --- a/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -23,7 +23,7 @@ Nano Banana 2 is now available in ComfyUI through Partner Nodes. This release fu ## Nano Banana 2 image edit workflow - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/google/nano-banana-pro.mdx b/tutorials/partner-nodes/google/nano-banana-pro.mdx index ec0b3c0e1..b64ef93d7 100644 --- a/tutorials/partner-nodes/google/nano-banana-pro.mdx +++ b/tutorials/partner-nodes/google/nano-banana-pro.mdx @@ -21,26 +21,35 @@ Nano Banana Pro is Google DeepMind's flagship image model (Gemini 3 Pro Image) f - **Character and subject consistency**: Maintain character resemblance and object fidelity across complex multi-subject compositions - **Advanced reasoning**: Configurable thinking levels — choose Minimal for fast exploration or High/Dynamic for complex layouts and precise control -## Workflows +### Nano Banana Pro (`api_nano_banana_pro`) -### Image Edit +Nano-banana Pro (Gemini 3.0 Pro Image) for studio-quality 4K image generation and editing with enhanced text rendering and character consistency. + +Nano Banana Pro workflow preview - + Open in Comfy Cloud - Download JSON or search "Nano Banana Pro" in Template Library + Download JSON or search "Nano Banana Pro" in Template Library - - Get the first example input image + + +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 11 · `api_nano_banana_pro_input_image_1.png` - - Get the second example input image + + `LoadImage` node 12 · `api_nano_banana_pro_input_image_2.png` -Upload images and provide text instructions to generate or edit. The workflow accepts multiple input images and produces a high-fidelity result with precise prompt adherence. +**Example output** ![Nano Banana Pro example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_nano_banana_pro.png) diff --git a/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx b/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx index d398c0aac..543bb488c 100644 --- a/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx +++ b/tutorials/partner-nodes/grok/grok-imagine-video-1-5.mdx @@ -24,7 +24,7 @@ Both variants generate **native audio**: sound effects, ambience, and dialogue a Grok Imagine Video 1.5 Workflow - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index 052701160..22df955be 100644 --- a/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -33,7 +33,7 @@ Version 1.1 targets five core production-critical capabilities: dynamic expressi Build a complete scene from scratch. You control style, shot size, lighting, action, and audio entirely through the prompt. HappyHorse 1.1 returns a single video with dialogue, sound effects, and music baked in. - + Try the Text-to-Video workflow instantly on Comfy Cloud. @@ -48,7 +48,7 @@ Build a complete scene from scratch. You control style, shot size, lighting, act Animate a static first frame. The image already carries the look, so you describe the motion and the camera move. HappyHorse 1.1 returns a video with audio baked in. - + Try the Image-to-Video workflow instantly on Comfy Cloud. @@ -66,7 +66,7 @@ Animate a static first frame. The image already carries the look, so you describ Orchestrate a multi-character stage play. Map characters and scenes to reference images, then direct them through a timestamped storyboard with per-character dialogue. Up to 9 reference images per generation, with character and scene references separated so characters stay consistent across background changes. - + Try the Reference-to-Video workflow instantly on Comfy Cloud. diff --git a/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index aea8038e9..f011246d6 100644 --- a/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -40,111 +40,101 @@ Currently, ComfyUI's Partner nodes support the following Hunyuan 3D model genera 2. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute model generation 3. After the workflow completes, preview and export your 3D model -## Text-to-3D workflow +### HunYuan3D: Text to Model (`api_hunyuan3d_text_to_model`) -Enter a text description to generate a 3D model in one click. The model accurately follows style, shape, and material details for consistent results. +Input a text prompt or upload a reference image to generate a detailed 3D model asset. **Example prompt**: "Mechanical dolphin with gears, steampunk" -Hunyuan 3D Text-to-3D workflow preview +HunYuan3D: Text to Model workflow preview - - Run on Comfy Cloud - -## Image-to-3D workflow - -Upload one or more images to generate high-quality 3D models. Support for 2–4 multi-view images improves geometry and material fidelity. + + + Open in Comfy Cloud + + + Download JSON or search "HunYuan3D: Text to Model" in Template Library + + -Hunyuan 3D Image-to-3D workflow preview +### HY 3D: Image to Model (`api_hunyuan3d_image_to_model`) - - Run on Comfy Cloud - -
-Input materials +Upload an image to generate a 3D model with geometry and PBR textures. Upload 2–4 multi-view images from different angles for improved geometry and material fidelity. -Download this sample input image to try the workflow: +HY 3D: Image to Model workflow preview - - - Download sample input image + + + Open in Comfy Cloud + + + Download JSON or search "HY 3D: Image to Model" in Template Library -
- -## Multi-view-to-3D workflow -Provide multiple view images (front, back, side) to generate 3D models with improved accuracy and detail. This uses the same workflow as Image-to-3D—simply upload 2–4 images from different angles. +**Input materials** -Hunyuan 3D Multi-view-to-3D workflow preview - - - Run on Comfy Cloud - -
-Input materials - -Download these sample multi-view input images to try the workflow: +Upload these files to the matching `LoadImage` nodes: - - Download front view - - - Download back view - - - Download left view + + `LoadImage` node 1 · `pink_robot_front.png` - - Download right view + + `LoadImage` node 10 · `pink_robot_back.png` -
+ +
+ pink_robot_front.png + pink_robot_back.png +
## Advanced features Following the [initial HY 3D 3.0 integration](https://blog.comfy.org/p/hunyuan-3d-30-in-comfyui-state-of), Hunyuan 3D's advanced processing features are now available via Partner Nodes. These workflows help close the gap between generation and production by bringing key post-processing steps into ComfyUI. -### 3D parts decomposition +### HY 3D: 3D Parts Decomposition (`api_hunyuan3d_part`) -Split a complete 3D model into meaningful structural parts, such as armor pieces, accessories, wheels, or other distinct components. This makes it easier to edit specific regions of an asset, swap parts for variations, and prepare models for modular workflows, animation, or downstream assembly. +Upload a 3D model to automatically segment it into its constituent parts, generating a fully decomposed 3D asset for reuse and editing. -Hunyuan 3D parts decomposition workflow preview +HY 3D: 3D Parts Decomposition workflow preview - - Run the workflow + + Open in Comfy Cloud - - Get the JSON file + + Download JSON or search "HY 3D: 3D Parts Decomposition" in Template Library -### UV unwrapping -Automatically generate UV layouts for supported 3D models, turning raw geometry into assets that are much easier to texture. Instead of manually cutting seams and organizing UV islands, creators can move more quickly into painting, baking, and material work with a cleaner starting point. +### HY 3D: UV Unwrapping (`api_hunyuan3d_model2uv`) + +Upload a 3D model to perform UV unwrapping. Generate a processed model with optimized UV layout for texturing. -Hunyuan 3D UV unwrapping workflow preview +HY 3D: UV Unwrapping workflow preview - - Run the workflow + + Open in Comfy Cloud - - Get the JSON file + + Download JSON or search "HY 3D: UV Unwrapping" in Template Library -### Smart topology -Convert dense geometry into cleaner meshes with more organized edge flow, helping generated models become easier to optimize and reuse in real production pipelines. This is especially useful when preparing assets for game engines, real-time rendering, or any workflow that benefits from lower-density, better-structured geometry. +### HY 3D: Smart Topology (`api_hunyuan3d_smart_topology`) + +Upload a high-poly 3D model file. Generate a lower-polygon, topologically optimized 3D model with a specified reduction level. -Hunyuan 3D smart topology workflow preview +HY 3D: Smart Topology workflow preview - - Run the workflow + + Open in Comfy Cloud - - Get the JSON file + + Download JSON or search "HY 3D: Smart Topology" in Template Library \ No newline at end of file diff --git a/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index 5630c754b..31dd2e1dc 100644 --- a/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -40,64 +40,52 @@ Currently, ComfyUI's Partner nodes support the following Hunyuan 3D model genera 2. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute model generation 3. After the workflow completes, preview and export your 3D model -## Text-to-3D workflow +### HunYuan3D: Text to Model (`api_hunyuan3d_text_to_model`) -Enter a text description to generate a 3D model in one click. The model accurately follows style, shape, and material details for consistent results. +Input a text prompt or upload a reference image to generate a detailed 3D model asset. **Example prompt**: "Mechanical dolphin with gears, steampunk" -Hunyuan 3D Text-to-3D workflow preview +HunYuan3D: Text to Model workflow preview - - Run on Comfy Cloud - -## Image-to-3D workflow - -Upload one or more images to generate high-quality 3D models. Support for 2–4 multi-view images improves geometry and material fidelity. + + + Open in Comfy Cloud + + + Download JSON or search "HunYuan3D: Text to Model" in Template Library + + -Hunyuan 3D Image-to-3D workflow preview +### HY 3D: Image to Model (`api_hunyuan3d_image_to_model`) - - Run on Comfy Cloud - -
-Input materials +Upload an image to generate a 3D model with geometry and PBR textures. Upload 2–4 multi-view images from different angles for improved geometry and material fidelity. -Download this sample input image to try the workflow: +HY 3D: Image to Model workflow preview - - - Download sample input image + + + Open in Comfy Cloud + + + Download JSON or search "HY 3D: Image to Model" in Template Library -
-## Multi-view-to-3D workflow +**Input materials** -Provide multiple view images (front, back, side) to generate 3D models with improved accuracy and detail. This uses the same workflow as Image-to-3D—simply upload 2–4 images from different angles. - -Hunyuan 3D Multi-view-to-3D workflow preview - - - Run on Comfy Cloud - -
-Input materials - -Download these sample multi-view input images to try the workflow: +Upload these files to the matching `LoadImage` nodes: - - Download front view + + `LoadImage` node 1 · `pink_robot_front.png` - - Download back view - - - Download left view - - - Download right view + + `LoadImage` node 10 · `pink_robot_back.png` -
+ +
+ pink_robot_front.png + pink_robot_back.png +
diff --git a/tutorials/partner-nodes/kling/kling-3-0.mdx b/tutorials/partner-nodes/kling/kling-3-0.mdx index 648110e02..941a76c6b 100644 --- a/tutorials/partner-nodes/kling/kling-3-0.mdx +++ b/tutorials/partner-nodes/kling/kling-3-0.mdx @@ -20,23 +20,86 @@ Kling 3.0 is one of the most advanced multi-modal generation systems, now availa -## Kling 3.0 workflows +### Kling 3.0: Text to Image (`api_kling_v3_t2i`) - - Download the Kling 3.0 Image-to-Video workflow. +Generate an image from a text prompt using the Kling 3.0 model. + +Kling 3.0: Text to Image workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Kling 3.0: Text to Image" in Template Library + + + +**Example output** + +![Kling 3.0: Text to Image example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_kling_v3_t2i.png) + +### Kling 3.0: Video Generation (`api_kling_v3_video`) + +Input text or image prompts to generate 15-second video sequences with multi-shot compositions, complex camera movements, and consistent subjects. + +Kling 3.0: Video Generation workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Kling 3.0: Video Generation" in Template Library + - - Download the Kling 3.0 Omni Video Edit workflow. +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 1 · `boxing.png` + + +
+ boxing.png +
+ +### Kling 3.0: First Last Frame to Video (`api_kling_v3_flf2v`) + +Input a first and last frame image to generate a continuous video sequence with multi-shot generation and precise element control. + +Kling 3.0: First Last Frame to Video workflow preview - - Try the Image-to-Video workflow instantly on Comfy Cloud. + + + Open in Comfy Cloud + + Download JSON or search "Kling 3.0: First Last Frame to Video" in Template Library + + + +**Input materials** + +Upload these files to the matching `LoadImage` nodes: - - Try the Omni Video Edit workflow instantly on Comfy Cloud. + + + `LoadImage` node 1 · `goldenfish_in_bag.png` + + `LoadImage` node 13 · `man_in_daydream.png` + + + +
+ goldenfish_in_bag.png + man_in_daydream.png +
## Learn more diff --git a/tutorials/partner-nodes/kling/kling-motion-control.mdx b/tutorials/partner-nodes/kling/kling-motion-control.mdx index add428521..0d4107a21 100644 --- a/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -38,32 +38,37 @@ The `character_orientation` parameter determines how the model interprets spatia -## Kling 2.6 Motion Control workflow +### Kling2.6: Motion Control (`api_kling_motion_control`) -Kling 2.6 Motion Control workflow preview +Apply precise character actions and expressions from a reference video to your character image with synchronized motion control. + +Kling2.6: Motion Control workflow preview - - Run the Kling 2.6 Motion Control workflow on Comfy Cloud. + + Open in Comfy Cloud - - Download the workflow JSON file for local use. + + Download JSON or search "Kling2.6: Motion Control" in Template Library -
-Input materials -Download these sample input files to try the workflow: +**Input materials** + +Upload these files to the matching `LoadImage` and `LoadVideo` nodes: - - Download sample reference image + + `LoadImage` node 11 · `streetwear_fox.jpg` - - Download sample motion reference video + + `LoadVideo` node 2 · `street_dancer.mp4` -
+ +
+ streetwear_fox.jpg +
## Input requirements diff --git a/tutorials/partner-nodes/krea2/krea2-t2i.mdx b/tutorials/partner-nodes/krea2/krea2-t2i.mdx index 43b1998d8..b283c52ec 100644 --- a/tutorials/partner-nodes/krea2/krea2-t2i.mdx +++ b/tutorials/partner-nodes/krea2/krea2-t2i.mdx @@ -26,7 +26,7 @@ Check the following documentation for detailed node parameter settings: Krea 2 Text to Image Workflow - + Open in Comfy Cloud @@ -47,7 +47,7 @@ Check the following documentation for detailed node parameter settings: The Style Reference workflow adds an extra step: upload one or more reference images that define the aesthetic style, mood, and visual direction. The generated image will match your prompt while applying the style from your reference images. - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index c293456f5..46e566028 100644 --- a/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -8,7 +8,6 @@ sidebarTitle: "Moonvalley" **Service unavailable**: The Moonvalley API service is no longer available. These nodes have been deprecated and may not function as expected. - import ReqHint from "/snippets/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/tutorials/update-reminder.mdx"; @@ -49,7 +48,6 @@ Currently, Moonvalley-related Partner nodes are natively supported in ComfyUI. Y

Download the workflow file in JSON format

- ### 2. Follow the Steps to Run the Workflow ![Text-to-Video Workflow](/images/tutorial/api_nodes/moonvalley/api_moonvalley_text_to_video.jpg) @@ -60,7 +58,6 @@ Currently, Moonvalley-related Partner nodes are natively supported in ComfyUI. Y 4. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to start video generation 5. After the API returns the result, you can view the generated video in the `Save Video` node. The video will also be saved in the `ComfyUI/output/` directory - ## Moonvalley Image-to-Video Workflow ### 1. Download the Workflow File @@ -90,7 +87,6 @@ Download the image below as the input image 5. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to start video generation 6. After the API returns the result, you can view the generated video in the `Save Video` node. The video will also be saved in the `ComfyUI/output/` directory - ## Moonvalley Video-to-Video Workflow The `Moonvalley Marey Video to Video` node allows you to input a reference video for video re-drawing. You can use the reference video's motion or character poses for video generation. @@ -115,7 +111,6 @@ Download the video below as the input video: src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video_input.mp4" > - ### 2. Follow the Steps to Run the Workflow ![Video-to-Video Workflow](/images/tutorial/api_nodes/moonvalley/api_moonvalley_video_to_video.jpg) diff --git a/tutorials/partner-nodes/openai/chat.mdx b/tutorials/partner-nodes/openai/chat.mdx index e96926d54..ed78bbc28 100644 --- a/tutorials/partner-nodes/openai/chat.mdx +++ b/tutorials/partner-nodes/openai/chat.mdx @@ -14,22 +14,36 @@ In this guide, we will walk you through completing the corresponding conversatio -## OpenAI Chat Workflow +### OpenAI: ChatGPT (`api_openai_chat`) -### 1. Workflow File Download +Engage with OpenAI's advanced language models for intelligent conversations. -Please download the Json file below and drag it into ComfyUI to load the corresponding workflow. +OpenAI: ChatGPT workflow preview - + Open in Comfy Cloud - - Download the JSON format workflow file. + + Download JSON or search "OpenAI: ChatGPT" in Template Library -### 2. Complete the Workflow Execution Step by Step +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 5 · `example.png` + + + +
+ example.png +
+ +### Complete the workflow execution step by step ![OpenAI Chat Step Guide](/images/tutorial/api_nodes/openai/openai_chat_step_guide.jpg) diff --git a/tutorials/partner-nodes/openai/dall-e-3.mdx b/tutorials/partner-nodes/openai/dall-e-3.mdx index cfcd4524b..959a2b51c 100644 --- a/tutorials/partner-nodes/openai/dall-e-3.mdx +++ b/tutorials/partner-nodes/openai/dall-e-3.mdx @@ -48,7 +48,6 @@ Since the corresponding workflow is very simple, you can also directly add the * ![ComfyUI openai-dall-e-3 workflow](/images/tutorial/api_nodes/openai/openai-dall-e-3/text2image.jpg) - 1. Add the **OpenAI DALL·E 3** node in ComfyUI 2. Enter the description of the image you want to generate in the prompt text box 3. Adjust optional parameters as needed (quality, style, size, etc.) diff --git a/tutorials/partner-nodes/openrouter/llm.mdx b/tutorials/partner-nodes/openrouter/llm.mdx index 6c6cef122..50b1e8381 100644 --- a/tutorials/partner-nodes/openrouter/llm.mdx +++ b/tutorials/partner-nodes/openrouter/llm.mdx @@ -59,17 +59,34 @@ The **model** dropdown is a dynamic combo. When you change the model, the node u Text-only models hide image and video inputs. -## Example workflow (`api_openrouter_llm`) +### OpenRouter LLM (`api_openrouter_llm`) -The template below is **one example**, not the only way to use the node. Download it to see how inputs connect, then adapt the graph for your task—pure text chat, captioning, prompt rewriting, feeding strings into image nodes, or anything else that fits your pipeline. +Select a model from OpenRouter's curated list (Claude, GPT, Gemini, etc.). Generate a text response with optional media uploads for vision-capable models. - +OpenRouter LLM workflow preview + + + Open in Comfy Cloud + + Download JSON or search "OpenRouter LLM" in Template Library + + + +**Input materials** - - Download JSON or search "OpenRouter LLM" in Template Library +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 2 · `contact_sheet_app_input_1.png` + + +
+ contact_sheet_app_input_1.png +
In this sample, the workflow analyzes a reference image and returns a text-to-image prompt for downstream image nodes. You can omit **Load Image**, change the prompts, or rebuild the graph entirely for your own use case. diff --git a/tutorials/partner-nodes/overview.mdx b/tutorials/partner-nodes/overview.mdx index 9f8eac3f5..1b36aa82e 100644 --- a/tutorials/partner-nodes/overview.mdx +++ b/tutorials/partner-nodes/overview.mdx @@ -27,7 +27,6 @@ Learn how to log in with ComfyUI Account API Key ![Select Comfy API Key Login](/images/interface/setting/user/user-login-api-1.jpg) - ## Use ComfyUI Account API Key Integration to call paid model Partner nodes Currently, we support accessing our services through ComfyUI API Key Integration to call paid model Partner nodes. Please refer to the API Key Integration section to learn how to use API Key Integration to call paid model Partner nodes. @@ -40,8 +39,6 @@ Currently, we support accessing our services through ComfyUI API Key Integration Please refer to the API Key Integration section to learn how to use API Key Integration to call paid model Partner nodes - - ## Advantages of Partner Nodes Partner Nodes provide several important advantages for ComfyUI users: @@ -51,7 +48,6 @@ Partner Nodes provide several important advantages for ComfyUI users: - **Simplified experience**: No need to manage API keys or handle complex API requests - **Controlled costs**: The prepaid system ensures you have complete control over your spending with no unexpected charges - ## Pricing diff --git a/tutorials/partner-nodes/recraft/recraft-v4.mdx b/tutorials/partner-nodes/recraft/recraft-v4.mdx index 47d3758c8..6b7fa1306 100644 --- a/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -25,7 +25,7 @@ Recraft V4 is a new image generation model built for professional design work. I ## Recraft V4 text to image workflow - + Open in Comfy Cloud @@ -59,7 +59,7 @@ Recraft V4 is a new image generation model built for professional design work. I Recraft V4 can generate production-ready SVG vector outputs directly. This is useful for logos, icons, brand assets, or anything that needs to scale. SVG outputs are compatible with Illustrator, Figma, and Sketch. - + Open in Comfy Cloud diff --git a/tutorials/partner-nodes/reve/reve-image.mdx b/tutorials/partner-nodes/reve/reve-image.mdx index 53bd62cc3..bfad54228 100644 --- a/tutorials/partner-nodes/reve/reve-image.mdx +++ b/tutorials/partner-nodes/reve/reve-image.mdx @@ -20,17 +20,24 @@ Reve is an image generation model available in ComfyUI through Partner Nodes. It -## Reve Create +### Reve: Text to Image (`api_reve_image_create`) -The Reve Image Create node generates images from text prompts. The model performs well with prompts describing fashion editorial styles, specific lighting conditions, natural textures, and motion. +Input text prompts to generate high-quality images with detailed aesthetics and accurate text rendering. - - Run the Reve Image Create workflow on Comfy Cloud. - +Reve: Text to Image workflow preview - - Download the workflow JSON file for local use. + + + Open in Comfy Cloud + + + Download JSON or search "Reve: Text to Image" in Template Library + + +**Example output** + +![Reve: Text to Image example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_reve_image_create.png) ### Example outputs @@ -72,17 +79,35 @@ A wide-angle landscape photograph of a solitary Jaguar F-Type car in British gre A minimalist wide shot of a vast white sand dune desert during golden hour. A white owl with its wings spread is in the foreground. Soft warm sunlight grazing the tops of the dunes, casting long shadows. Sparse, golden-brown desert shrubs are scattered across the ripples, with a prominent patch of dry, cracked, hexagonal mud in the bottom right foreground. In the far distance, a hazy, ethereal range of mountains rests against a pale, washed-out sky. ``` -## Reve Edit +### Reve: Image Edit (`api_reve_image_edit`) + +Upload an image and provide text instructions to edit it. Generate a modified output image using Reve's advanced editing model. -The Reve Image Edit node modifies existing images based on text instructions. It supports editing lighting conditions, color grading, and adding motion effects. +Reve: Image Edit workflow preview - - Run the Reve Image Edit workflow on Comfy Cloud. + + + Open in Comfy Cloud + + Download JSON or search "Reve: Image Edit" in Template Library + + + +**Input materials** - - Download the workflow JSON file for local use. +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 4 · `woman_running_in_field.png` + + +
+ Input image + Reve: Image Edit example output +
### Example outputs @@ -104,17 +129,40 @@ The lighting is cold, soft glow emits from the window. Side profile shot of the subject, his hand is grabbing the sunglasses. ``` -## Reve Remix +### Reve: Image Remix And Edit (`api_reve_image_remix`) -The Reve Image Remix node uses elements from reference images as layers and applies them to a target image. This allows you to transfer specific visual properties such as shadows, color grading, or aesthetic styles between images. +Remix and edit images using Reve's advanced model. Combine multiple reference images into a single output. - - Run the Reve Image Remix workflow on Comfy Cloud. +Reve: Image Remix And Edit workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Reve: Image Remix And Edit" in Template Library + + +**Input materials** + +Upload these files to the matching `LoadImage` nodes: - - Download the workflow JSON file for local use. + + + `LoadImage` node 2 · `iced_drink_cup.png` + + `LoadImage` node 4 · `casual_male_portrait.png` + + + `LoadImage` node 5 · `japanese_street_alley.png` + + + +**Example output** + +![Reve: Image Remix And Edit example](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_reve_image_remix.png) ### Example outputs diff --git a/tutorials/partner-nodes/runway/video-generation.mdx b/tutorials/partner-nodes/runway/video-generation.mdx index 5e43ef55a..1af825f00 100644 --- a/tutorials/partner-nodes/runway/video-generation.mdx +++ b/tutorials/partner-nodes/runway/video-generation.mdx @@ -80,7 +80,6 @@ You can refer to the numbers in the image to complete the basic image-to-video w ## First-Last Frame Video Generation Workflow - ### 1. Workflow File Download The video below contains workflow information in its `metadata`. Please download and drag it into ComfyUI to load the corresponding workflow. diff --git a/tutorials/partner-nodes/sonilo/video-to-music.mdx b/tutorials/partner-nodes/sonilo/video-to-music.mdx index 82d8e2696..bdf7523fe 100644 --- a/tutorials/partner-nodes/sonilo/video-to-music.mdx +++ b/tutorials/partner-nodes/sonilo/video-to-music.mdx @@ -34,26 +34,50 @@ You can drive music generation directly from your ComfyUI node graphs, alongside -## Video-to-music workflow +### Sonilo: Video Soundtrack Generation (`api_sonilo_v2m`) -Sonilo's video-to-music workflow analyzes your video's visuals, pacing, and emotional cues to generate a synchronized soundtrack. +Generate synchronized soundtracks from video footage. Input a video to produce music that matches its pacing and emotional cues. - - Run on Comfy Cloud - +Sonilo: Video Soundtrack Generation workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Sonilo: Video Soundtrack Generation" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 705 · `leopard_rider.mp4` + + 1. Load your video in the input node. 2. (Optional) Add a text prompt to guide the style or mood of the generated music. 3. Click the **Run** button or use the shortcut `Ctrl(Cmd) + Enter` to start music generation. 4. The generated audio will be saved to the `ComfyUI/output/audio` directory. -## Text-to-audio workflow +### Sonilo: Text to Music Generation (`api_sonilo_t2m`) + +Generate high-quality, production-ready music from text prompts. -You can also generate music purely from text prompts describing the style, mood, and instrumentation you want. +Sonilo: Text to Music Generation workflow preview - - Run on Comfy Cloud - + + + Open in Comfy Cloud + + + Download JSON or search "Sonilo: Text to Music Generation" in Template Library + + **Example prompts:** diff --git a/tutorials/partner-nodes/tripo/model-generation.mdx b/tutorials/partner-nodes/tripo/model-generation.mdx index 377ef512d..711a5516b 100644 --- a/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/tutorials/partner-nodes/tripo/model-generation.mdx @@ -89,7 +89,6 @@ You can refer to the numbers in the image to complete the basic image-to-model w 3. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute model generation. After the workflow completes, the corresponding model will be automatically saved to the `ComfyUI/output/` directory 4. For model download, please refer to the instructions in the text-to-model section - ## Multi-view Model Generation Workflow Generate a 3D model from multiple view images for enhanced accuracy. diff --git a/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/tutorials/partner-nodes/tripo/tripo-3-1.mdx index cb2c0870c..493a6712d 100644 --- a/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -84,4 +84,3 @@ Generate a high-detail 3D model from multiple view images using Tripo 3.1. | Material Output | PBR-ready | Standard material maps | | Refine Support | Supported | Supported (v1.4 only for Refine Draft) | - diff --git a/tutorials/utility/depth-anything-3.mdx b/tutorials/utility/depth-anything-3.mdx index 7ba26dc3c..a462c1d72 100644 --- a/tutorials/utility/depth-anything-3.mdx +++ b/tutorials/utility/depth-anything-3.mdx @@ -28,10 +28,10 @@ ComfyUI now natively supports Depth Anything 3 nodes. Make sure you have updated Download the Depth Anything 3 checkpoint(s) and save them to the corresponding ComfyUI folder: -- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_small.safetensors)) — Lightweight, fast inference -- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_base.safetensors)) — Balanced performance -- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — Best for monocular depth, includes sky detection -- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/resolve/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — Metric scale depth in metres, includes sky detection +- **Small** ([depth_anything_3_small.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_small.safetensors)) — Lightweight, fast inference +- **Base** ([depth_anything_3_base.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_base.safetensors)) — Balanced performance +- **Mono-Large** ([depth_anything_3_mono_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_mono_large.safetensors)) — Best for monocular depth, includes sky detection +- **Metric-Large** ([depth_anything_3_metric_large.safetensors](https://huggingface.co/Comfy-Org/Depth-Anything-3/blob/main/geometry_estimation/depth_anything_3_metric_large.safetensors)) — Metric scale depth in metres, includes sky detection ``` ComfyUI/ @@ -45,24 +45,33 @@ ComfyUI/ ## Example Workflows ---- +### Depth Anything 3: Image Depth Estimation (`utility_depth_anything3_image_depth_estimation`) -## 1. Image Depth Estimation +Upload one image and generate a depth map using Depth Anything 3. View a side-by-side comparison of the original image and depth output. -**What it does:** Upload one image and run **Image Depth Estimation (Depth Anything 3)** to produce a depth map. The result is shown in **Depth Preview**, with a side-by-side comparison view of the original image and depth output. +Depth Anything 3 image depth estimation workflow preview + + Open in Comfy Cloud + - Download JSON or search "Depth Anything 3" in Template Library + Download JSON or search "Depth Anything 3: Image Depth Estimation" in Template Library - - Get the example input image for this workflow + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 85 · `retro_futuristic_home.png` -
- Image Depth Estimation output - Image Depth Estimation comparison +
+ Input image
### Steps to Run @@ -76,22 +85,36 @@ ComfyUI/ This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ---- +### Depth Anything 3: Video Depth Estimation (`utility_depth_anything3_video_depth_estimation`) -## 2. Video Depth Estimation +Upload a video to generate a per-frame depth sequence. Inside the subgraph, **GetVideoComponents** splits the input video into frames, **LoadDA3Model** loads the model, and **SetVideoComponents** reassembles the depth frames back into a video output. -**What it does:** Upload a video and run **Video Depth Estimation (Depth Anything 3)** to produce a per-frame depth sequence. Inside the subgraph, **GetVideoComponents** splits the input video into frames, **LoadDA3Model** loads the model, and **SetVideoComponents** reassembles the depth frames back into a video output. +Depth Anything 3 video depth estimation workflow preview - - Download JSON or search "Depth Anything 3" in Template Library - Open in Comfy Cloud + + Download JSON or search "Depth Anything 3: Video Depth Estimation" in Template Library + -![Video Depth Estimation preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_depth_anything3_video_depth_estimation-1.webp) +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 87 · `empty_room_assembly.mp4` + + + +
+ +
### Steps to Run @@ -104,8 +127,6 @@ ComfyUI/ This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ---- - ## Model Variants | Variant | head_type | has_sky | has_confidence | camera_decoder | Best for | diff --git a/tutorials/utility/face-detection/mediapipe.mdx b/tutorials/utility/face-detection/mediapipe.mdx index b15c9dfc0..f792f8573 100644 --- a/tutorials/utility/face-detection/mediapipe.mdx +++ b/tutorials/utility/face-detection/mediapipe.mdx @@ -26,29 +26,44 @@ MediaPipe Face Detection is natively supported in ComfyUI (PR [#14009](https://g ## MediaPipe Face Detection Workflow -### 1. Download the Workflow +### Mediapipe: Image Face Detection (`utility_face_detection_mediapipe`) -Update your ComfyUI to the latest version, then go to `Workflow` → `Browse Templates` and find "Mediapipe: Image Face Detection" under the Utility category. +Input an image and detect up to 6 facial landmarks per face, enabling ultrafast multi-face detection. - - Download workflow - +Mediapipe image face detection workflow preview - - Open in cloud + + + Open in Comfy Cloud + + Download JSON or search "Mediapipe: Image Face Detection" in Template Library + + + +**Input materials** - - Get the example input image for this workflow +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 5 · `soft_neon_girl.png` + + +
+ Input image +
-![MediaPipe Face Detection preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_face_detection_mediapipe-1.webp) +### 1. Download the Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` → `Browse Templates` and find "Mediapipe: Image Face Detection" under the Utility category. ### 2. Download the Model The MediaPipe Face Detection model is hosted on the [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe). -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) Place it in the following directory structure: diff --git a/tutorials/utility/frame-interpolation.mdx b/tutorials/utility/frame-interpolation.mdx index 2ceaa2710..dcdd6920a 100644 --- a/tutorials/utility/frame-interpolation.mdx +++ b/tutorials/utility/frame-interpolation.mdx @@ -32,4 +32,3 @@ Rather than always treating interpolation as a final post-process, it can also b Run on Comfy Cloud - diff --git a/tutorials/utility/moge.mdx b/tutorials/utility/moge.mdx index 10b42eb8a..8eebfc318 100644 --- a/tutorials/utility/moge.mdx +++ b/tutorials/utility/moge.mdx @@ -52,31 +52,30 @@ ComfyUI/ ## Example Workflows ---- - -## 1. Depth Estimation +### MoGe: Depth Estimation (`utility_moge_depth_estimation`) -**What it does:** Takes a single image and estimates its metric depth map, colored depth preview, and mask: outputs the same metric-scale depth that MoGe infers in one forward pass. Useful as a scene depth reference for compositing, depth-based effects, or as preprocessing before mesh generation. +Upload a single RGB image and generate a colored depth preview and raw depth map. -MoGe also estimates the camera's field of view (FOV) from the image, which can be optionally overridden with a ground-truth value for even more accurate results. +MoGe depth estimation workflow preview - - Download JSON or search "MoGe Depth Estimation" in Template Library - - - - Open in Comfy Cloud - + + Open in Comfy Cloud + + + Download JSON or search "MoGe: Depth Estimation" in Template Library + - - Get the example input image for this workflow +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 9 · `alien_world.png` -
- depth estimation color preview - depth estimation raw preview -
+ ### 1.1 Steps to Run @@ -85,26 +84,36 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b 3. Click `Queue` or use `Ctrl(cmd) + Enter` to run 4. The workflow outputs colored depth preview, raw depth preview, and a mask ---- +### MoGe: Perspective Geometry Estimation (`3d_moge_perspective_to_mesh`) -## 2. Perspective to Mesh +Upload an image to estimate its perspective geometry. Generate a 3D depth map and surface normals from the input, then convert to a textured GLB mesh. -**What it does:** Converts a single perspective photo into a textured GLB mesh with normal and depth previews. MoGe estimates point maps, depth, and normals from the visible scene, then converts them to a mesh. This is **monocular geometry estimation**: occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. +MoGe perspective to mesh workflow preview - - Download JSON or search "3D MoGe Perspective to Mesh" in Template Library - - - - Open in Comfy Cloud - + + Open in Comfy Cloud + + + Download JSON or search "MoGe: Perspective Geometry Estimation" in Template Library + - - Get the example input image for this workflow +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 9 · `modern_living_room.png` -![perspective to mesh preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_perspective_to_mesh-1.webp) + + +
+ Input image +
+ +This is **monocular geometry estimation**: occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. ### 2.1 Steps to Run @@ -113,26 +122,36 @@ MoGe also estimates the camera's field of view (FOV) from the image, which can b 3. (Optional) View the OpenGL and DirectX normal previews 4. Click `Queue` or use `Ctrl(cmd) + Enter` to run ---- +### Moge: Panorama to Mesh (`3d_moge_panorama_to_mesh`) -## 3. Panorama to Mesh +Upload an equirectangular 360° panorama image and generate a textured GLB mesh with vertex colors. -**What it does:** Converts an equirectangular (360°) panorama into a textured GLB mesh. The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, runs monocular geometry estimation on each view independently, then merges them into a single mesh. Each segment is still single-view estimation, so the result is a rough scene reconstruction: useful for getting a spatial overview of a 360° scene, but occluded areas and geometry behind surfaces will be missing or fragmented. +MoGe panorama to mesh workflow preview - - Download JSON or search "3D MoGe Panorama to Mesh" in Template Library - - - - Open in Comfy Cloud - + + Open in Comfy Cloud + + + Download JSON or search "Moge: Panorama to Mesh" in Template Library + - - Get the example input image for this workflow +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 28 · `lego_street_panorama.png` -![panorama to mesh preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/3d_moge_panorama_to_mesh-1.webp) + + +
+ Input panorama +
+ +The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, run monocular geometry estimation on each view independently, then merge them into a single mesh. ### 3.1 Steps to Run diff --git a/tutorials/utility/pose-detection-sdpose.mdx b/tutorials/utility/pose-detection-sdpose.mdx index 9f8bd07b1..90031a6d3 100644 --- a/tutorials/utility/pose-detection-sdpose.mdx +++ b/tutorials/utility/pose-detection-sdpose.mdx @@ -39,52 +39,144 @@ Four workflows are available depending on your use case: Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find SDPose workflows under the Utility category. +### SDPose: Image Multi-Person Detection (`utility_sdpose_multi_person`) + +Upload an image to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose image multi-person detection workflow preview + - - Run in Comfy Cloud + + Open in Comfy Cloud - - Download JSON + + Download JSON or search "SDPose: Image Multi-Person Detection" in Template Library +**Input materials** + +Upload this file to the matching `LoadImage` node: + - - Run in Comfy Cloud + + `LoadImage` node 679 · `group_photo.png` - - Download JSON + + +
+ Input image +
+ +### SDPose: Video Multi-Person Detection (`utility_sdpose_multi_person_video`) + +Upload a video to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose video multi-person detection workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose: Video Multi-Person Detection" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 694 · `man_playing_violin.mp4` +
+ +
+ +### SDPose-OOD: Image to Pose Map (`utility_sdpose_ood_image_to_pose`) + +Upload an image to extract pose keypoints and generate a corresponding pose map using the SDPose-OOD model. + +SDPose-OOD image to pose map workflow preview + - - Run in Comfy Cloud + + Open in Comfy Cloud - - Download JSON + + Download JSON or search "SDPose-OOD: Image to Pose Map" in Template Library +**Input materials** + +Upload this file to the matching `LoadImage` node: + - - Run in Comfy Cloud + + `LoadImage` node 667 · `dancer.png` - - Download JSON + + +**Example output** + +
+ Input image + SDPose-OOD image to pose map example output +
+ +### SDPose-OOD: Video to Pose Map (`utility_sdpose_ood_video_to_pose_map`) + +Upload a video to extract pose keypoints and generate a pose map. The workflow supports multiple person detection and uses an enhanced SDPose model for accurate whole-body feature extraction. + +SDPose-OOD video to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Video to Pose Map" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 679 · `man_in_the_rain.mp4` +**Example output** + +
+ + +
+ ### 2. Download Models The SDPose and RT-DETRv4 model checkpoints are hosted on the [Comfy-Org SDPose model repository](https://huggingface.co/Comfy-Org/SDPose). **checkpoints** (SDPose model): -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) **diffusion_models** (RT-DETRv4 detector): -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (recommended) -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (full precision, larger) +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (recommended) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (full precision, larger) Place them in the following directory structure: diff --git a/tutorials/utility/preprocessors.mdx b/tutorials/utility/preprocessors.mdx index 9f8c8aa4f..1bd821fef 100644 --- a/tutorials/utility/preprocessors.mdx +++ b/tutorials/utility/preprocessors.mdx @@ -29,13 +29,36 @@ This workflow emphasizes: Depth outputs can be reused across multiple passes, making it easier to iterate without re-running expensive upstream steps. - - Run on Comfy Cloud - +### Video to Depth Map (`utility-depthAnything-v2-relative-video`) - - Download JSON - +Convert a video to a temporally stable depth map. + +Video to Depth Map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Depth Map" in Template Library + + + +**Input materials** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 6 · `roller_coaster.mp4` + + + +
+ +
## Lineart conversion @@ -48,13 +71,36 @@ This workflow is designed to: Lineart pairs especially well with depth and pose, offering strong structural constraints without overconstraining style. - - Run on Comfy Cloud - +### Video to Lineart / Canny (`utility-lineart-video`) + +Convert a video to lineart or Canny edges for control processors. + +Video to Lineart workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Lineart / Canny" in Template Library + + + +**Input materials** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 2 · `utility-lineart-video-input.mp4` + + - - Download JSON - +
+ +
## Pose detection @@ -67,13 +113,36 @@ This workflow focuses on: By isolating pose extraction into a dedicated workflow, pose data becomes easier to inspect, refine, and reuse. - - Run on Comfy Cloud - +### Video to Pose Map - OpenPose (`utility-openpose-video`) - - Download JSON - +Convert a video to a temporally stable pose control map. + +Video to Pose Map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Pose Map - OpenPose" in Template Library + + + +**Input materials** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 2 · `pose_input.mp4` + + + +
+ +
## Normals extraction @@ -90,12 +159,34 @@ Normal outputs can be used to: - Add a stronger 3D-like structure to stylization and redraw pipelines - Improve consistency across frames when paired with pose/depth for animation work - - Run on Comfy Cloud - +### Video to Normal Map (`utility-normal_crafter-video`) + +Convert a video to a temporally stable normal map. + +Video to Normal Map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Normal Map" in Template Library + + + +**Input materials** + +Upload this file to the matching `VHS_LoadVideo` node: - - Download JSON - + + + `VHS_LoadVideo` node 3 · `normals_input.mp4` + + +
+ +
diff --git a/tutorials/utility/remove-background-birefnet.mdx b/tutorials/utility/remove-background-birefnet.mdx index 95f0a46ac..99cb59335 100644 --- a/tutorials/utility/remove-background-birefnet.mdx +++ b/tutorials/utility/remove-background-birefnet.mdx @@ -29,19 +29,36 @@ BiRefNet is natively supported in ComfyUI (PR [#12747](https://github.com/Comfy- Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "BiRefNet: Remove Background" under the Utility category. - - Download workflow +### BiRefNet: Remove Background (`utility_birefnet_remove_background`) + +Upload an image with any background. Generate a version with the background removed and a precision segmentation mask. + +BiRefNet remove background workflow preview + + + + Open in Comfy Cloud + + Download JSON or search "BiRefNet: Remove Background" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadImage` node: - - Open in cloud + + + `LoadImage` node 17 · `the_lily_veil.png` + ### 2. Download Models The BiRefNet model is hosted on the [Comfy-Org BiRefNet model repository](https://huggingface.co/Comfy-Org/BiRefNet). -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) Place it in the following directory structure: diff --git a/tutorials/utility/seedvr2.mdx b/tutorials/utility/seedvr2.mdx index c428b0e50..0dda4ff1c 100644 --- a/tutorials/utility/seedvr2.mdx +++ b/tutorials/utility/seedvr2.mdx @@ -25,14 +25,14 @@ SeedVR2 comes in two sizes. The [3B model](https://huggingface.co/ByteDance-Seed Download the SeedVR2 checkpoint(s) from the [Comfy-Org SeedVR2 repository](https://huggingface.co/Comfy-Org/SeedVR2) and save them to the corresponding ComfyUI folder: -- [seedvr2_3b_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_3b_fp16.safetensors): 3B FP16 base model, lower VRAM requirement -- [seedvr2_3b_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_3b_fp8_e4m3fn.safetensors): 3B FP8 quantized, further reduced VRAM -- [seedvr2_3b_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_3b_int8_convrot.safetensors): 3B INT8 with convrot -- [seedvr2_7b_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_7b_fp16.safetensors): 7B FP16 base model, higher quality -- [seedvr2_7b_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_7b_fp8_e4m3fn.safetensors): 7B FP8 quantized, balance of quality and VRAM -- [seedvr2_7b_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_7b_int8_convrot.safetensors): 7B INT8 with convrot -- [seedvr2_7b_sharp_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/diffusion_models/seedvr2_7b_sharp_fp16.safetensors): 7B FP16 sharp variant for enhanced detail -- [seedvr2_ema_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/resolve/main/vae/seedvr2_ema_vae_fp16.safetensors): VAE checkpoint (shared across all model variants) +- [seedvr2_3b_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_3b_fp16.safetensors): 3B FP16 base model, lower VRAM requirement +- [seedvr2_3b_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_3b_fp8_e4m3fn.safetensors): 3B FP8 quantized, further reduced VRAM +- [seedvr2_3b_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_3b_int8_convrot.safetensors): 3B INT8 with convrot +- [seedvr2_7b_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_7b_fp16.safetensors): 7B FP16 base model, higher quality +- [seedvr2_7b_fp8_e4m3fn.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_7b_fp8_e4m3fn.safetensors): 7B FP8 quantized, balance of quality and VRAM +- [seedvr2_7b_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_7b_int8_convrot.safetensors): 7B INT8 with convrot +- [seedvr2_7b_sharp_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/diffusion_models/seedvr2_7b_sharp_fp16.safetensors): 7B FP16 sharp variant for enhanced detail +- [seedvr2_ema_vae_fp16.safetensors](https://huggingface.co/Comfy-Org/SeedVR2/blob/main/vae/seedvr2_ema_vae_fp16.safetensors): VAE checkpoint (shared across all model variants) Place the files in the following directories: @@ -53,21 +53,37 @@ ComfyUI/ ## Example Workflows ---- +### SeedVR2 3B Int8: Upscale Image (`utility_seedvr2_3b_int8_upscale_image`) -## 1. Image Upscale (3B INT8) +Upscale images using SeedVR2 3B Int8, a one-step diffusion-based video restoration model that produces high-quality results with improved temporal consistency. -**What it does:** Upscales a single image using the SeedVR2 3B INT8 model. The INT8 quantized variant offers a good balance of quality and reduced VRAM usage. +SeedVR2 3B Int8 upscale image workflow preview - - Download JSON or search "SeedVR2 3B Int8: Upscale Image" in Template Library + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Image" in Template Library + + +**Input materials** + +Upload this file to the matching `LoadImage` node: - - Get the example input image for this workflow + + + `LoadImage` node 1 · `watch_macro_shot.png` + -![SeedVR2 3B INT8 upscale preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_3b_int8_upscale_image.png) +**Example output** + +
+ Input image + SeedVR2 3B Int8 upscale example output +
### 1.1 Steps to Run @@ -75,21 +91,37 @@ ComfyUI/ 2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader 3. Click `Queue` or use `Ctrl(cmd) + Enter` to run ---- +### SeedVR2 7B Int8: Upscale Image (`utility_seedvr2_7b_int8_upscale_image`) -## 2. Image Upscale (7B INT8) +Upscale images using SeedVR2 7B Int8, a one-step diffusion model that enhances resolution through adversarial training and adaptive window attention. -**What it does:** Upscales a single image using the SeedVR2 7B INT8 model. The larger 7B model delivers higher quality results with INT8 quantization for efficient VRAM usage. +SeedVR2 7B Int8 upscale image workflow preview - - Download JSON or search "SeedVR2 7B Int8: Upscale Image" in Template Library + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 7B Int8: Upscale Image" in Template Library + - - Get the example input image for this workflow +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 1 · `indoor_portrait.png` + -![SeedVR2 7B INT8 upscale preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_7b_int8_upscale_image.png) +**Example output** + +
+ Input image + SeedVR2 7B Int8 upscale example output +
### 2.1 Steps to Run @@ -97,19 +129,41 @@ ComfyUI/ 2. Select the `seedvr2_7b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader 3. Click `Queue` or use `Ctrl(cmd) + Enter` to run ---- +### SeedVR2 3B Int8: Upscale Video (`utility_seedvr2_3b_int8_upscale_video`) -## 3. Video Upscale (3B INT8) +Upscale and restore video footage using SeedVR2 3B Int8, a one-step diffusion model that enhances resolution while maintaining temporal consistency across frames. -**What it does:** Upscales a video using the SeedVR2 3B INT8 model. The workflow enhances resolution while maintaining temporal consistency across frames. Ideal for restoring old or degraded footage and upscaling low-resolution videos. +SeedVR2 3B Int8 upscale video workflow preview - - Download JSON or search "SeedVR2 3B Int8: Upscale Video" in Template Library + + + Open in Comfy Cloud + + Download JSON or search "SeedVR2 3B Int8: Upscale Video" in Template Library + + + +**Input materials** + +Upload this file to the matching `LoadVideo` node: - - Get the example input video for this workflow + + + `LoadVideo` node 73 · `grainy_perfume_shot_crf32.mp4` + + +**Example output** + +
+ + +
### 3.1 Steps to Run diff --git a/tutorials/utility/video-segment-sam3.mdx b/tutorials/utility/video-segment-sam3.mdx index ccf362cdf..79abae81c 100644 --- a/tutorials/utility/video-segment-sam3.mdx +++ b/tutorials/utility/video-segment-sam3.mdx @@ -17,7 +17,7 @@ SAM 3.1 is natively supported in ComfyUI (PR [#13408](https://github.com/Comfy-O [SAM 3 GitHub](https://github.com/facebookresearch/sam3) | [Paper (arXiv)](https://arxiv.org/abs/2604.02296) | [🤗 Model Hub](https://huggingface.co/Comfy-Org/sam3.1) SAM 3.1 segments and tracks objects across video frames based on text prompts. The example above shows the segmentation output with masks applied to the target objects throughout the video. @@ -37,31 +37,71 @@ SAM 3.1 segments and tracks objects across video frames based on text prompts. T Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find the SAM 3.1 workflows under the Utility category. -**Video Segmentation:** +### SAM3: Video Segmentation (`utility_video_segment_sam3`) + +Use the SAM3 model to segment the main subject or content from a video, isolating specific objects or regions. + +SAM3 video segmentation workflow preview - - Download video workflow + + + Open in Comfy Cloud + + Download JSON or search "SAM3: Video Segmentation" in Template Library + + + +**Input materials** - - Open in cloud +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 115 · `drinking_unicorn.mp4` + -**Image Segmentation:** +**Example output** - - Download image workflow + + +### SAM3: Image Segmentation (`utility_image_segment_sam3`) + +Use the SAM3 model to segment the main subject or content from a photo or image, isolating specific objects or regions. + +SAM3 image segmentation workflow preview + + + + Open in Comfy Cloud + + Download JSON or search "SAM3: Image Segmentation" in Template Library + + + +**Input materials** - - Open in cloud +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 79 · `neon_guitarist.png` + + +
+ Input image +
### 2. Download Models The SAM 3.1 model is hosted on the [Comfy-Org SAM 3.1 model repository](https://huggingface.co/Comfy-Org/sam3.1). -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) Place it in the following directory structure: diff --git a/tutorials/utility/void-video-inpainting.mdx b/tutorials/utility/void-video-inpainting.mdx index ad0eebf65..4d1646236 100644 --- a/tutorials/utility/void-video-inpainting.mdx +++ b/tutorials/utility/void-video-inpainting.mdx @@ -19,11 +19,11 @@ VOID is natively supported in ComfyUI (PR [#13403](https://github.com/Comfy-Org/
@@ -44,13 +44,30 @@ VOID is natively supported in ComfyUI (PR [#13403](https://github.com/Comfy-Org/ Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "VOID: Video Inpainting" under the Utility category. - - Download workflow +### VOID: Video Inpainting (`utility_void_video_inpainting`) + +Upload a video and mask the object you want to remove. Generate a clean video with the object and its physical interactions deleted. + +VOID video inpainting workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "VOID: Video Inpainting" in Template Library + - - Open in cloud +**Input materials** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 4 · `snowboarder.mp4` + ### 2. Download Models @@ -58,24 +75,24 @@ All models are hosted on the [Comfy-Org VOID model repository](https://huggingfa **Diffusion Models** — the core two-pass inpainting model: -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — Refinement pass, better temporal stability -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — Primary pass +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — Refinement pass, better temporal stability +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — Primary pass **VAE:** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) **Optical Flow:** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) **SAM3 Checkpoint** — for segmentation: -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) **Text Encoder:** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ @@ -114,8 +131,6 @@ All models are hosted on the [Comfy-Org VOID model repository](https://huggingfa Use **Pass 2** (refinement pass) for longer clips or textured backgrounds where temporal stability matters. **Pass 1** alone is faster but may show more jitter. - - This workflow uses Subgraph nodes for modular video processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. diff --git a/tutorials/video/bytedance/bernini-r.mdx b/tutorials/video/bytedance/bernini-r.mdx index 0f2a388cd..915172c11 100644 --- a/tutorials/video/bytedance/bernini-r.mdx +++ b/tutorials/video/bytedance/bernini-r.mdx @@ -71,16 +71,18 @@ ComfyUI/ **What it does:** Generate an edited image with matched lighting and view a side-by-side before/after comparison. Ideal for portrait and product relighting, consistent lighting across photo sets, and e-commerce catalog photography. +Bernini-R Image Editing workflow preview + - - Download JSON or search "Bernini-R" in Template Library - Open in Comfy Cloud + + Download JSON or search "Bernini-R" in Template Library + -#### Input materials +**Input materials** @@ -91,11 +93,6 @@ ComfyUI/ -
- Bernini-R Image Editing output - Bernini-R Image Editing comparison -
- ### Steps to Run 1. **Select Task Type**: choose your task (Image Editing, Subject to Image, etc.) @@ -117,16 +114,18 @@ ComfyUI/ **What it does:** Generate an edited video with consistent relighting using Bernini-R. Connect a source video, optional reference image(s) or reference video, pick the task type, write a prompt, and run. +Bernini-R Video Editing workflow preview + - - Download JSON or search "Bernini-R" in Template Library - Open in Comfy Cloud + + Download JSON or search "Bernini-R" in Template Library + -#### Input materials +**Input materials** @@ -137,8 +136,6 @@ ComfyUI/ -![Bernini-R Video Editing preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_bernini_r_video_editing-1.webp) - ### Steps to Run 1. **Load Source Video**: connect your input video diff --git a/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index a4e8245d3..da132e772 100644 --- a/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -52,12 +52,12 @@ Please download the video below and drag it into ComfyUI to load the workflow. T > - - Download the JSON format workflow file - Run this workflow on Comfy Cloud with pre-installed models + + Download the JSON format workflow file + Please download the following image as input: diff --git a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index c4b3dbe52..ef1961c7c 100644 --- a/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -111,7 +111,7 @@ HunyuanVideo 1.5 Image-to-Video transforms static images into smooth, high-quali -#### Input materials +**Input materials** diff --git a/tutorials/video/hunyuan/hunyuan-video.mdx b/tutorials/video/hunyuan/hunyuan-video.mdx index 6dc8b76e7..f6f3d317f 100644 --- a/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/tutorials/video/hunyuan/hunyuan-video.mdx @@ -147,12 +147,12 @@ Download the workflow image below and drag it into ComfyUI to load the workflow: ![ComfyUI Workflow - Hunyuan Image-to-Video v1](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/i2v/v1_robot.webp) - - Download the workflow image and drag into ComfyUI - Open in Comfy Cloud + + Download the workflow image and drag into ComfyUI + Download the image below and use it as the starting frame for the image-to-video generation: @@ -205,12 +205,12 @@ Download the workflow image below and drag it into ComfyUI to load the workflow: ![ComfyUI Workflow - Hunyuan Image-to-Video v2](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hunyuan-video/i2v/v2_fennec_gril.webp) - - Download the workflow image and drag into ComfyUI - Open in Comfy Cloud + + Download the workflow image and drag into ComfyUI + Download the image below and use it as the starting frame for the image-to-video generation: @@ -278,7 +278,6 @@ Samurai waving sword and hitting the camera. camera angle movement, zoom in, fas flying car fastly moving and flying through the city ``` - --- ![example](/images/tutorial/advanced/hunyuanvideo/cyber_car_race.png) diff --git a/tutorials/video/kandinsky/kandinsky-5.mdx b/tutorials/video/kandinsky/kandinsky-5.mdx index 14032111c..8b5abb642 100644 --- a/tutorials/video/kandinsky/kandinsky-5.mdx +++ b/tutorials/video/kandinsky/kandinsky-5.mdx @@ -37,19 +37,19 @@ All models are available in 5-second and 10-second video generation versions. ## Text-to-Video workflow -### 1. Download workflow file +### Kandinsky 5.0 Video Lite Text to Video (`video_kandinsky5_t2v`) -Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 T2V" to load the workflow. +A lightweight 2B model that generates videos from English and Russian prompts with high visual quality. -![Kandinsky 5.0 T2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_t2v-1.webp) +Kandinsky 5.0 Video Lite Text to Video workflow preview - - Download the T2V workflow to use locally - Open in Comfy Cloud + + Download JSON or search "Kandinsky 5.0 Video Lite Text to Video" in Template Library + ### 2. Manually download models @@ -89,26 +89,35 @@ ComfyUI/ ## Image-to-Video workflow -### 1. Download workflow file +### Kandinsky 5.0 Video Lite Image to Video (`video_kandinsky5_i2v`) -Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Kandinsky 5.0 I2V" to load the workflow. +A lightweight 2B model that generates videos from English and Russian prompts with high visual quality. -![Kandinsky 5.0 I2V Workflow Preview](https://github.com/Comfy-Org/workflow_templates/blob/main/templates/video_kandinsky5_i2v-1.webp) +Kandinsky 5.0 Video Lite Image to Video workflow preview - - Download the I2V workflow to use locally - Open in Comfy Cloud + + Download JSON or search "Kandinsky 5.0 Video Lite Image to Video" in Template Library + -**Input Image** +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 11 · `crystal_flower.png` + + + +
+ crystal_flower.png +
- - Default input image for the I2V workflow. Download and use this image, or replace with your own. - ### 2. Manually download models **Text Encoders** diff --git a/tutorials/video/ltx/ltx-2-3.mdx b/tutorials/video/ltx/ltx-2-3.mdx index 408036a7d..69e841739 100644 --- a/tutorials/video/ltx/ltx-2-3.mdx +++ b/tutorials/video/ltx/ltx-2-3.mdx @@ -44,8 +44,14 @@ Generate videos from text prompts with improved prompt understanding and text re +### LTX-2.3 Text to Video (`video_ltx2_3_t2v`) + +Generate videos from text prompts with improved prompt understanding and text rendering. + +LTX-2.3 Text to Video ( workflow preview + - + Open in Comfy Cloud @@ -103,8 +109,14 @@ Generate videos from an input image with more consistent motion and smoother ani +### LTX-2.3 Image to Video (`video_ltx2_3_i2v`) + +Generate videos from an input image with improved motion consistency. + +LTX-2.3 Image to Video ( workflow preview + - + Open in Comfy Cloud @@ -164,8 +176,14 @@ Interpolate between a start image and an end image to generate a smooth video tr +### LTX-2.3 First-Last Frame to Video (`video_ltx2_3_flf2v`) + +Interpolate between a start and end image. + +LTX-2.3 First-Last Frame to Video ( workflow preview + - + Open in Comfy Cloud @@ -214,8 +232,14 @@ Upload an image and an audio file to generate a high-quality video with synchron +### LTX-2.3 Image and Audio to Video (`video_ltx2_3_ia2v`) + +Generate lip-synced videos from an image and audio. + +LTX-2.3 Image and Audio to Video ( workflow preview + - + Open in Comfy Cloud @@ -279,7 +303,7 @@ Generate LTX-2.3 videos with IC-LoRA using aligned control inputs like depth, po ![LTX-2.3 IC-LoRA preview](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_ltx2_3_ic_lora-1.webp) - + Open in Comfy Cloud @@ -291,7 +315,7 @@ Generate LTX-2.3 videos with IC-LoRA using aligned control inputs like depth, po This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -#### Input materials +**Input materials** @@ -346,8 +370,14 @@ Generate personalized videos with synchronized audio from a text prompt, referen +### LTX-2.3 ID LoRA (`video_ltx2_3_id_lora`) + +Generate personalized videos with synchronized audio from a reference image and audio clip. + +LTX-2.3 ID LoRA ( workflow preview + - + Open in Comfy Cloud diff --git a/tutorials/video/ltx/ltx-2.mdx b/tutorials/video/ltx/ltx-2.mdx index 66a900902..b2dd1fae9 100644 --- a/tutorials/video/ltx/ltx-2.mdx +++ b/tutorials/video/ltx/ltx-2.mdx @@ -51,7 +51,7 @@ Generate videos from text prompts. Download workflow - + Open in cloud @@ -64,7 +64,7 @@ Generate videos from text prompts. Download workflow - + Open in cloud @@ -79,12 +79,12 @@ Generate videos from an input image. Download workflow - + Open in cloud -#### Input materials +**Input materials** @@ -100,12 +100,12 @@ Generate videos from an input image. Download workflow - + Open in cloud -#### Input materials +**Input materials** @@ -125,12 +125,12 @@ Generate videos with structural control using IC-LoRAs. Download workflow - + Open in cloud -#### Input materials +**Input materials** @@ -149,12 +149,12 @@ Generate videos with structural control using IC-LoRAs. Download workflow - + Open in cloud -#### Input materials +**Input materials** @@ -173,12 +173,12 @@ Generate videos with structural control using IC-LoRAs. Download workflow - + Open in cloud -#### Input materials +**Input materials** diff --git a/tutorials/video/ltxv.mdx b/tutorials/video/ltxv.mdx index 850c4ff0b..0f5fc53c6 100644 --- a/tutorials/video/ltxv.mdx +++ b/tutorials/video/ltxv.mdx @@ -22,26 +22,34 @@ Drag the video directly into ComfyUI to run the workflow. ## Image to Video -Allows you to control the video with a first frame image. +### LTXV Image to Video (`ltxv_image_to_video`) + +Generate videos from still images. + +LTXV Image to Video workflow preview Open in Comfy Cloud - Download JSON or search "LTX-Video" in Template Library + Download JSON or search "LTXV Image to Video" in Template Library -#### Input materials +**Input materials** + +Upload this file to the matching `LoadImage` node: - - Download the default input image, or use your own image. + + `LoadImage` node 78 · `ltxv_image_to_video_input_image.jpg` -LTX-Video Image to Video +
+ ltxv_image_to_video_input_image.jpg +
Drag the video directly into ComfyUI to run the workflow. @@ -49,17 +57,21 @@ Drag the video directly into ComfyUI to run the workflow. ## Text to Video +### LTXV Text to Video (`ltxv_text_to_video`) + +Generate videos from text prompts. + +LTXV Text to Video workflow preview + Open in Comfy Cloud - Download JSON or search "LTX-Video" in Template Library + Download JSON or search "LTXV Text to Video" in Template Library -LTX-Video Text to Video - Drag the video directly into ComfyUI to run the workflow. diff --git a/tutorials/video/wan/fun-camera.mdx b/tutorials/video/wan/fun-camera.mdx index bfdb1fbf6..fe707bb56 100644 --- a/tutorials/video/wan/fun-camera.mdx +++ b/tutorials/video/wan/fun-camera.mdx @@ -88,10 +88,14 @@ File Storage Location: ## ComfyUI Wan2.1 Fun Camera 1.3B Workflow -### 1. Download Workflow +### Wan 2.1 Fun Camera 1.3B (`video_wan2.1_fun_camera_v1.1_1.3B`) + +Generate dynamic videos with cinematic camera movements using Wan 2.1 Fun Camera 1.3B model. + +Wan 2.1 Fun Camera 1.3B workflow preview - + Open in Comfy Cloud @@ -111,14 +115,20 @@ Download the video below and drag it into ComfyUI to load the corresponding work If you want to use the 14B version, simply replace the model file with the 14B version, but please be aware of the VRAM requirements. -### 2. Download Input Material +**Input materials** - - - Download the image below to use as the starting frame for the 1.3B workflow +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 52 · `video_wan2.1_fun_camera_v1.1_1.3B_start_image.jpg` +
+ video_wan2.1_fun_camera_v1.1_1.3B_start_image.jpg +
+ ### 3. Complete the Workflow Step by Step ![Wan2.1 Fun Camera Workflow Steps](/images/tutorial/video/wan/wan2-1-fun-camera-1-3b-step-guide.jpg) @@ -136,10 +146,14 @@ If you want to use the 14B version, simply replace the model file with the 14B v ## ComfyUI Wan2.1 Fun Camera 14B Workflow -### 1. Download Workflow +### Wan 2.1 Fun Camera 14B (`video_wan2.1_fun_camera_v1.1_14B`) + +Generate high-quality videos with advanced camera control using the full 14B model. + +Wan 2.1 Fun Camera 14B workflow preview - + Open in Comfy Cloud @@ -153,14 +167,20 @@ If you want to use the 14B version, simply replace the model file with the 14B v src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/fun-camera/v1.1/wan2.1_fun_camera_14B.mp4" > -### 2. Download Input Material +**Input materials** - - - Download the image below to use as the starting frame for the 14B workflow +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 52 · `video_wan2.1_fun_camera_v1.1_14B_start_image.jpg` +
+ video_wan2.1_fun_camera_v1.1_14B_start_image.jpg +
+ ## Performance Reference **1.3B Version**: diff --git a/tutorials/video/wan/fun-control.mdx b/tutorials/video/wan/fun-control.mdx index 265066de6..954de6240 100644 --- a/tutorials/video/wan/fun-control.mdx +++ b/tutorials/video/wan/fun-control.mdx @@ -111,7 +111,7 @@ Thanks to our powerful ComfyUI authors who provide feature-rich nodes. If you wa ![Wan2.1 Fun Control Native Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/wan2.1_fun_control/wan2.1_fun_control_native.webp) - + Open in Comfy Cloud @@ -119,7 +119,7 @@ Thanks to our powerful ComfyUI authors who provide feature-rich nodes. If you wa -#### Input Materials +**Input materials** @@ -165,7 +165,7 @@ You can use [ComfyUI Manager](https://github.com/Comfy-Org/ComfyUI-Manager) to i ![Wan2.1 Fun Control Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/wan2.1_fun_control-1.webp) - + Open in Comfy Cloud @@ -173,7 +173,7 @@ You can use [ComfyUI Manager](https://github.com/Comfy-Org/ComfyUI-Manager) to i -#### Input Materials +**Input materials** diff --git a/tutorials/video/wan/fun-inp.mdx b/tutorials/video/wan/fun-inp.mdx index 8d88335a0..f0a1b6cb4 100644 --- a/tutorials/video/wan/fun-inp.mdx +++ b/tutorials/video/wan/fun-inp.mdx @@ -39,7 +39,7 @@ Download the image below and drag it into ComfyUI to load the workflow: ![Wan2.1 Fun InP Workflow](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/wan2.1_fun_inp-1.webp) - + Open in Comfy Cloud diff --git a/tutorials/video/wan/vace.mdx b/tutorials/video/wan/vace.mdx index 82dbb9fd3..05a9d20b1 100644 --- a/tutorials/video/wan/vace.mdx +++ b/tutorials/video/wan/vace.mdx @@ -124,13 +124,19 @@ Since the models used in the workflows covered in this document are consistent, **What it does:** Transform text descriptions into high-quality videos using the Wan2.1 VACE 14B model. Supports both 480p and 720p resolution. +### Wan2.1 VACE Text to Video (`video_wan_vace_14B_t2v`) + +Transform text descriptions into high-quality videos. Supports both 480p and 720p with VACE-14B model. + +Wan2.1 VACE Text to Video workflow preview + - - Download JSON or search "Wan2.1 VACE Text to Video" in Template Library - Open in Comfy Cloud + + Download JSON or search "Wan2.1 VACE Text to Video" in Template Library + -### 1. Download Workflow File +### Wan 2.2 5B Video Generation (`video_wan2_2_5B_ti2v`) -Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Wan2.2 5B video generation" to load the workflow. +Fast text-to-video and image-to-video generation with 5B parameters. Optimized for rapid prototyping and creative exploration. - +Wan 2.2 5B Video Generation workflow preview + + Open in Comfy Cloud + Download JSON or search "Wan2.2 5B" in Template Library - - Open in Comfy Cloud - ### 2. Manually Download Models @@ -150,25 +145,19 @@ ComfyUI/ ## Wan2.2 14B T2V Text-to-Video Workflow Example -### 1. Workflow File - -Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Wan2.2 14B T2V" to load the workflow. +### Wan 2.2 14B Text to Video (`video_wan2_2_14B_t2v`) -Or update your ComfyUI to the latest version, then download the following video and drag it into ComfyUI to load the workflow. +Generate high-quality videos from text prompts with cinematic aesthetic control and dynamic motion generation using Wan 2.2. - +Wan 2.2 14B Text to Video workflow preview + + Open in Comfy Cloud + Download JSON or search "Wan2.2 14B T2V" in Template Library - - Open in Comfy Cloud - ### 2. Manually Download Models @@ -227,28 +216,37 @@ ComfyUI/ ## Wan2.2 14B I2V Image-to-Video Workflow Example -### 1. Workflow File +### Wan 2.2 14B Image to Video (`video_wan2_2_14B_i2v`) -Please update your ComfyUI to the latest version, and through the menu `Workflow` -> `Browse Templates` -> `Video`, find "Wan2.2 14B I2V" to load the workflow. +Transform static images into dynamic videos with precise motion control and style preservation using Wan 2.2. -Or update your ComfyUI to the latest version, then download the following video and drag it into ComfyUI to load the workflow. - +Wan 2.2 14B Image to Video workflow preview + + Open in Comfy Cloud + Download JSON or search "Wan2.2 14B I2V" in Template Library - - Open in Comfy Cloud + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 97 · `video_wan2_2_14B_i2v_input_image.jpg` -You can use the following image as input: -![Input Image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/input.jpg) +**Example output** + +
+ Input image + Wan 2.2 14B Image to Video example output +
### 2. Manually Download Models @@ -308,28 +306,38 @@ ComfyUI/ The first and last frame workflow uses the same model locations as the I2V section. -### 1. Workflow and Input Material Preparation +### Wan 2.2 14B First-Last Frame to Video (`video_wan2_2_14B_flf2v`) -Download the video or the JSON workflow below and open it in ComfyUI. - +Generate smooth video transitions by defining start and end frames. + +Wan 2.2 14B First-Last Frame to Video workflow preview + + Open in Comfy Cloud + Download JSON or search "Wan2.2 14B FLF2V" in Template Library - - Open in Comfy Cloud - -Download the following images as input materials: +**Input materials** + +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 62 · `video_wan2_2_14B_flf2v_start_image.png` + + + `LoadImage` node 68 · `video_wan2_2_14B_flf2v_end_image.png` + + -![Input Material](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v_start_image.png) -![Input Material](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/2.2/wan22_14B_flf2v_end_image.png) +
+ video_wan2_2_14B_flf2v_start_image.png + video_wan2_2_14B_flf2v_end_image.png +
### 2. Follow the Steps diff --git a/tutorials/video/zai/scail2.mdx b/tutorials/video/zai/scail2.mdx index e57801872..e76532a22 100644 --- a/tutorials/video/zai/scail2.mdx +++ b/tutorials/video/zai/scail2.mdx @@ -20,18 +20,22 @@ import UpdateReminder from '/snippets/tutorials/update-reminder.mdx' ## SCAIL-2 Character Replacement Workflow +### SCAIL-2 Character Replacement (`video_wan21_scail2_character_replacement`) + +Replace a character in a video while preserving motion and scene context. + +SCAIL-2 Character Replacement workflow preview + - + Open in Comfy Cloud - Download JSON or search "SCAIL-2" in Template Library + Download JSON or search "SCAIL-2 Character Replacement" in Template Library -![ComfyUI Workflow - SCAIL-2 Character Replacement](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_wan21_scail2_character_replacement-1.webp) - -#### Input materials +**Input materials** diff --git a/zh/tutorials/3d/hunyuan3D-2.mdx b/zh/tutorials/3d/hunyuan3D-2.mdx index 54416450d..699302167 100644 --- a/zh/tutorials/3d/hunyuan3D-2.mdx +++ b/zh/tutorials/3d/hunyuan3D-2.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Hunyuan3D 2.0 Open-Source Model Series": 363ff037 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' # 混元3D 2.0 简介 @@ -29,7 +27,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 1. **几何生成模型（Hunyuan3D-DiT）**：基于流扩散的Transformer架构，生成无纹理的几何模型，可精准匹配输入条件。 2. **纹理生成模型（Hunyuan3D-Paint）**：结合几何条件和多视图扩散技术，为模型添加高分辨率纹理，支持PBR材质。 - **主要优势** - **高精度生成**：几何结构锐利，纹理色彩丰富，支持PBR材质生成，实现接近真实的光影效果。 @@ -53,179 +50,242 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 可以加载对应的工作流并提示完成模型下载，对应的 `.glb` 格式模型将输出至 `ComfyUI/output/mesh` 文件夹。 -## ComfyUI Hunyuan3D-2mv 工作流示例 +## ComfyUI Hunyuan3D-2mv Workflow Example -Hunyuan3D-2mv 工作流中，我们将使用多视角的图片来生成3D模型，另外多个视角的图片在这个工作流中并不是必须的，你可以只输入 `front` 视角的图片来生成3D模型。 +In the Hunyuan3D-2mv workflow, we'll use multi-view images to generate a 3D model. Note that multiple view images are not mandatory in this workflow - you can use only the `front` view image to generate a 3D model. +### HY 3D 2.0 MV (`3d_hunyuan3d_multiview_to_model`) + +Generate 3D models from multiple views using Hunyuan3D 2.0 MV. + +HY 3D 2.0 MV workflow preview + - 在 Comfy Cloud 上立即运行此工作流 + Run this workflow instantly on Comfy Cloud - 下载工作流 JSON 文件 + Download JSON or search "HY 3D 2.0 MV" in Template Library -### 1. 工作流 +**输入素材** -请下载下面的图片，并拖入 ComfyUI 以加载工作流, +Upload these files to the matching `LoadImage` nodes: -![Hunyuan3D-2mv workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/hunyuan-3d-multiview-elf.webp) + + + `LoadImage` node 56 · front view + + + `LoadImage` node 78 · left view + + + `LoadImage` node 80 · back view + + + `LoadImage` node 87 · right view + + +### 1. Workflow + +Please download the images below and drag into ComfyUI to load the workflow. +![Hunyuan3D-2mv workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/hunyuan-3d-multiview-elf.webp) -下载下面的图片，同时我们将使用这些图片作为图片输入 +Download the images below we will use them as input images. ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/front.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/left.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/back.png) -在本示例中提供的输入图片都已经过提前处理去除了多余的背景，在实际的使用中，你可以借助类似[ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) 这样的自定义来完成多余背景的自动去除。 +In this example, the input images have already been preprocessed to remove excess background. In actual use, you can use custom nodes like [ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) to automatically remove excess background. -### 2. 手动安装模型 +### 2. Manual Model Installation -下载下面的模型，并保存到对应的 ComfyUI 文件夹 +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv.safetensors` +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv.safetensors // 重命名后的文件 +│ │ └── hunyuan3d-dit-v2-mv.safetensors // renamed file ``` -### 3. 按步骤运行工作流 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2mv](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv.jpg) -1. 确保 Image Only Checkpoint Loader(img2vid model) 加载了我们下载并重命名的 `hunyuan3d-dit-v2-mv.safetensors` 模型 -2. 在 `Load Image` 节点的各个视角中加载了对应视角的图片 -3. 点击 `Queue` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 +1. Ensure that the Image Only Checkpoint Loader(img2vid model) has loaded our downloaded and renamed `hunyuan3d-dit-v2-mv.safetensors` model +2. Load the corresponding view images in each of the `Load Image` nodes +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -如果你需要增加更多的视角，请确保 `Hunyuan3Dv2ConditioningMultiView` 节点中加载了其它视角的图片，并确保在 `Load Image` 节点中加载了对应视角的图片。 +If you need to add more views, make sure to load other view images in the `Hunyuan3Dv2ConditioningMultiView` node, and ensure that you load the corresponding view images in the `Load Image` nodes. -## 使用 Hunyuan3D-2mv-turbo 工作流 +## Hunyuan3D-2mv-turbo Workflow -Hunyuan3D-2mv-turbo 工作流中，我们将使用 Hunyuan3D-2mv-turbo 模型来生成3D模型，这个模型是 Hunyuan3D-2mv 的分步蒸馏（Step Distillation）版本，可以更快地生成3D模型，在这个版本的工作流中我们设置 `cfg` 为 1.0 并添加 `flux guidance` 节点来控制 `distilled cfg` 的生成。 +In the Hunyuan3D-2mv-turbo workflow, we'll use the Hunyuan3D-2mv-turbo model to generate 3D models. This model is a step distillation version of Hunyuan3D-2mv, allowing for faster 3D model generation. In this version of the workflow, we set `cfg` to 1.0 and add a `flux guidance` node to control the `distilled cfg` generation. + +### HY 3D 2.0 MV Turbo (`3d_hunyuan3d_multiview_to_model_turbo`) + +Generate 3D models from multiple views using Hunyuan3D 2.0 MV Turbo. + +HY 3D 2.0 MV Turbo workflow preview - 在 Comfy Cloud 上立即运行此工作流 + Run this workflow instantly on Comfy Cloud - 下载工作流 JSON 文件 + Download JSON or search "HY 3D 2.0 MV Turbo" in Template Library -### 1. 工作流 +**输入素材** -请下载下面的图片，并拖入 ComfyUI 以加载工作流, +Upload these files to the matching `LoadImage` nodes: + + + + `LoadImage` node 56 · front view + + + `LoadImage` node 82 · back view + + + `LoadImage` node 85 · left view + + + `LoadImage` node 87 · right view + + + +### 1. Workflow + +Please download the images below and drag into ComfyUI to load the workflow. ![Hunyuan3D-2mv-turbo workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) -我们将使用下面的图片作为多视角的输入 +Download the images below we will use them as input images. ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) -### 2. 手动安装模型 +### 2. Manual Model Installation -下载下面的模型，并保存到对应的 ComfyUI 文件夹 +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2-mv-turbo.safetensors` +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2-mv-turbo.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // 重命名后的文件 +│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // renamed file ``` -### 3. 按步骤运行工作流 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2mv_turbo](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv_turbo.jpg) -1. 确保 `Image Only Checkpoint Loader(img2vid model)` 节点加载了我们重命名后的 `hunyuan3d-dit-v2-mv-turbo.safetensors` 模型 -2. 在 `Load Image` 节点的各个视角中加载了对应视角的图片 -3. 点击 `Queue` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 +1. Ensure that the `Image Only Checkpoint Loader(img2vid model)` node has loaded our renamed `hunyuan3d-dit-v2-mv-turbo.safetensors` model +2. Load the corresponding view images in each of the `Load Image` nodes +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow + +## Hunyuan3D-2 Single View Workflow -## 使用 Hunyuan3D-2 单视图工作流 +In the Hunyuan3D-2 workflow, we'll use the Hunyuan3D-2 model to generate 3D models. This model is not a multi-view model. In this workflow, we use the `Hunyuan3Dv2Conditioning` node instead of the `Hunyuan3Dv2ConditioningMultiView` node. -Hunyuan3D-2 工作流中，我们将使用 Hunyuan3D-2 模型来生成3D模型，这个模型不是一个多视角的模型，在这个工作流中，我们使用`Hunyuan3Dv2Conditioning` 节点替换掉 `Hunyuan3Dv2ConditioningMultiView` 节点。 +### HY 3D 2.0 (`3d_hunyuan3d_image_to_model`) + +Generate 3D models from single images using Hunyuan3D 2.0. + +HY 3D 2.0 workflow preview - 在 Comfy Cloud 上立即运行此工作流 + Run this workflow instantly on Comfy Cloud - 下载工作流 JSON 文件 + Download JSON or search "HY 3D 2.0" in Template Library -### 1. 工作流 +**输入素材** -请下载下面的图片，并拖入 ComfyUI 以加载工作流 +Upload this file to the matching `LoadImage` node: -![Hunyuan3D-2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) + + + `LoadImage` node 56 · `3d_hunyuan3d_image_to_model_input_image.png` + + + +### 1. Workflow -同时我们将使用这张图片作为图片输入 +Please download the image below and drag it into ComfyUI to load the workflow. + +![Hunyuan3D-2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) +Download the image below we will use it as input image. ![ComfyUI Hunyuan 3D 2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan_3d_v2_non_multiview_train.png) -### 2. 手动安装模型 +### 2. Manual Model Installation -下载下面的模型，并保存到对应的 ComfyUI 文件夹 +Download the model below and save it to the corresponding ComfyUI folder -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) 下载后可重命名为 `hunyuan3d-dit-v2.safetensors` +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - after downloading, you can rename it to `hunyuan3d-dit-v2.safetensors` ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2.safetensors // 重命名后的文件 +│ │ └── hunyuan3d-dit-v2.safetensors // renamed file ``` -### 3. 按步骤运行工作流 +### 3. Steps to Run the Workflow ![ComfyUI hunyuan3d_2](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2_non_multiview.jpg) -1. 确保 `Image Only Checkpoint Loader(img2vid model)` 节点加载了我们重命名后的 `hunyuan3d-dit-v2.safetensors` 模型 -2. 在 `Load Image` 节点中加载了对应视角的图片 -3. 点击 `Queue` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 +1. Ensure that the `Image Only Checkpoint Loader(img2vid model)` node has loaded our renamed `hunyuan3d-dit-v2.safetensors` model +2. Load the image in the `Load Image` node +3. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -## 社区资源 +## Community Resources -下面是 Hunyuan3D-2 的相关的 ComfyUI 社区资源 +Below are ComfyUI community resources related to Hunyuan3D-2 - [ComfyUI-Hunyuan3DWrapper](https://github.com/kijai/ComfyUI-Hunyuan3DWrapper) - [Kijai/Hunyuan3D-2_safetensors](https://huggingface.co/Kijai/Hunyuan3D-2_safetensors/tree/main) - [ComfyUI-3D-Pack](https://github.com/MrForExample/ComfyUI-3D-Pack) -## 混元3D 2.0 开源模型系列 +## Hunyuan3D 2.0 Open-Source Model Series -目前混元3D 2.0 开源了多个模型，覆盖了完整的3D生成流程，你可以访问 [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2) 了解更多。 +Currently, Hunyuan3D 2.0 has open-sourced multiple models covering the complete 3D generation process. You can visit [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2) for more information. -**Hunyuan3D-2mini 系列** +**Hunyuan3D-2mini Series** -| 模型 | 描述 | 日期 | 参数 | Huggingface | -|--------------------------|-------------------------------|------------|-------|------------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-mini | Mini 图像到形状模型 | 2025-03-18 | 0.6B | [前往](https://huggingface.co/tencent/Hunyuan3D-2mini/tree/main/hunyuan3d-dit-v2-mini) | +| Model | Description | Date | Parameters | Huggingface | +|-----------------------|---------------------------|------------|------------|--------------------------------------------------------------------------------------------| +| Hunyuan3D-DiT-v2-mini | Mini Image to Shape Model | 2025-03-18 | 0.6B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mini/tree/main/hunyuan3d-dit-v2-mini) | -**Hunyuan3D-2mv 系列** +**Hunyuan3D-2mv Series** -| 模型 | 描述 | 日期 | 参数 | Huggingface | -|--------------------------|----------------------------------------------------|------------|-------|-----------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-mv-Fast | 指导蒸馏版本，可以将 DIT 推理时间减半 | 2025-03-18 | 1.1B | [前往](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv-fast) | -| Hunyuan3D-DiT-v2-mv | 多视角图像到形状模型，适合需要用多个角度理解场景的 3D 创作 | 2025-03-18 | 1.1B | [前往](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv) | +| Model | Description | Date | Parameters | Huggingface | +|--------------------------|-------------------------------------------------------------------------------------------------------------|------------|------------|------------------------------------------------------------------------------------------| +| Hunyuan3D-DiT-v2-mv-Fast | Guidance Distillation Version, can halve DIT inference time | 2025-03-18 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv-fast) | +| Hunyuan3D-DiT-v2-mv | Multi-view Image to Shape Model, suitable for 3D creation requiring multiple angles to understand the scene | 2025-03-18 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2mv/tree/main/hunyuan3d-dit-v2-mv) | -**Hunyuan3D-2 系列** +**Hunyuan3D-2 Series** -| 模型 | 描述 | 日期 | 参数 | Huggingface | -|--------------------------|-------------------------------|------------|-------|-------------------------------------------------------------------------------------| -| Hunyuan3D-DiT-v2-0-Fast | 指导蒸馏模型 | 2025-02-03 | 1.1B | [前往](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast) | -| Hunyuan3D-DiT-v2-0 | 图像到形状模型 | 2025-01-21 | 1.1B | [前往](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0) | -| Hunyuan3D-Paint-v2-0 | 纹理生成模型 | 2025-01-21 | 1.3B | [前往](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-paint-v2-0) | -| Hunyuan3D-Delight-v2-0 | 图像去光影模型 | 2025-01-21 | 1.3B | [前往](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-delight-v2-0) | +| Model | Description | Date | Parameters | Huggingface | +|-------------------------|-----------------------------|------------|------------|---------------------------------------------------------------------------------------| +| Hunyuan3D-DiT-v2-0-Fast | Guidance Distillation Model | 2025-02-03 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0-fast) | +| Hunyuan3D-DiT-v2-0 | Image to Shape Model | 2025-01-21 | 1.1B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-dit-v2-0) | +| Hunyuan3D-Paint-v2-0 | Texture Generation Model | 2025-01-21 | 1.3B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-paint-v2-0) | +| Hunyuan3D-Delight-v2-0 | Image Delight Model | 2025-01-21 | 1.3B | [Visit](https://huggingface.co/tencent/Hunyuan3D-2/tree/main/hunyuan3d-delight-v2-0) | diff --git a/zh/tutorials/3d/triposplat.mdx b/zh/tutorials/3d/triposplat.mdx index d456036c7..515269abb 100644 --- a/zh/tutorials/3d/triposplat.mdx +++ b/zh/tutorials/3d/triposplat.mdx @@ -13,10 +13,6 @@ translationBlockHashes: "Model downloads": 44894623 --- - - - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **TripoSplat** 是一款开源模型，能够从单张 2D 图片直接生成 **3D 高斯泼溅（Gaussian splat）** 表示。由 VAST-AI 开发，以开源许可证发布。 @@ -54,83 +50,132 @@ TripoSplat 使用 **前馈架构**，接收单张 RGB 图像并直接预测一 - 加载输入图片（PNG/JPG） - 示例图片：`white-hotel-on-rocky-island.png`（可在模板库中获取） -### TripoSplat（子图） +### TripoSplat: Image to Gaussian Splat (`3d_triposplat_image_to_gaussian_splat`) + +Upload a single 2D image. Generate a high-quality 3D Gaussian splat representation with controllable density and budget for rendering. + +TripoSplat workflow preview + + + + + + Run this workflow instantly on Comfy Cloud + + + Download JSON or search "TripoSplat: Image to Gaussian Splat" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 99 · `white-hotel-on-rocky-island.png` + + + +
+ Input image +
+ +## How it works + +TripoSplat uses a **feed-forward architecture** that takes a single RGB image and directly predicts a set of 3D Gaussian primitives. The pipeline involves: + +1. **Image encoding**: the input image is processed by a vision encoder (DINOv2) +2. **Triplane generation**: features are decoded into a triplane representation +3. **Gaussian prediction**: the triplane is sampled to produce Gaussian parameters (position, scale, rotation, opacity, color) +4. **Rendering**: Gaussians are rendered from arbitrary viewpoints using differentiable splatting + + + This workflow uses a Subgraph node for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. + +## Workflow node guide + +### LoadImage +- Loads your input image (PNG/JPG) +- Sample image: `white-hotel-on-rocky-island.png` (available in Template Library) + +### TripoSplat (subgraph) -主处理子图节点，接收图片并生成 3D 高斯泼溅。暴露的参数： +The main subgraph node processes the image and generates the 3D Gaussian splat. Exposed parameters: -| 参数 | 默认值 | 说明 | +| Parameter | Default | Description | |---|---|---| -| `switch` | — | 启用/禁用子图 | -| `num_gaussians` | — | 生成的高斯基元数量（控制质量/性能） | -| `seed` | — | 随机种子，用于结果复现 | -| `unet_name` | — | TripoSplat 扩散模型检查点 | -| `clip_name` | — | CLIP 视觉编码器模型 | -| `vae_name` | — | 用于编码/解码的 VAE 模型（两个：一个为主 VAE，一个为编码器） | -| `bg_removal_name` | — | 背景去除模型 | +| `switch` | — | Enable/disable the subgraph | +| `num_gaussians` | — | Number of Gaussian primitives to generate (controls quality/performance) | +| `seed` | — | Random seed for reproducibility | +| `unet_name` | — | TripoSplat diffusion model checkpoint | +| `clip_name` | — | CLIP vision encoder model | +| `vae_name` | — | VAE for encoding/decoding (2 entries: one for the main VAE, one for the encoder) | +| `bg_removal_name` | — | Background removal model | ### CreateCameraInfo -- 定义渲染结果的相机轨道 -- 参数：轨道类型、角度、距离、视野等 -- 默认：35° 仰角、距离 30、缩放 2.5 +- Defines the camera orbit for rendering the result +- Parameters: orbit type, angle, distance, field of view, etc. +- Default: orbit at 35° elevation, 30 distance, 2.5 zoom ### RenderSplat -- 从定义的相机角度将高斯泼溅渲染为 2D 图像 -- 参数：输出分辨率（默认 1024×1024）、图像质量设置 +- Renders the Gaussian splat into a 2D image from the defined camera angle +- Parameters: output resolution (default 1024×1024), image quality settings ### SplatToMesh -- 将高斯泼溅转换为网格（可选） -- 参数：网格密度、平滑度、简化程度 +- Converts the Gaussian splat to a mesh (optional) +- Parameters: mesh density, smoothing, simplification ### SaveGLB -- 将结果保存为 GLB 3D 文件 +- Saves the result as a GLB 3D file ### SaveVideo -- 保存渲染 3D 场景的视频 +- Saves a video of the rendered 3D scene ### SplatToFile3D -- 以 SPZ 格式导出高斯泼溅 +- Exports the Gaussian splat in SPZ format ### CreateVideo -- 从渲染帧创建视频 +- Creates a video from rendered frames -## 运行步骤 +## Steps to run -1. **加载图片** — 使用 **LoadImage** 节点加载一张 2D 图片 -2. **运行 TripoSplat 子图** — 模型将生成高斯泼溅表示 -3. **选择输出格式** — 导出为 GLB、SPZ、视频或渲染为网格 -4. **查看结果** — 使用生成的 3D 文件或渲染预览 +1. **Load an image**: use the **LoadImage** node to load a single 2D image +2. **Run the TripoSplat subgraph**: the model will generate a Gaussian splat representation +3. **Choose output format**: export as GLB, SPZ, video, or render to mesh +4. **View results**: use the created 3D file or rendered preview -## 输出选项 +## Output options -| 节点 | 格式 | 用途 | +| Node | Format | Use case | |---|---|---| -| **SaveGLB** | `.glb` | 标准 3D 文件格式，可导入 3D 软件 | -| **SplatToFile3D** | `.spz` | 压缩的高斯泼溅格式，高效存储 | -| **RenderSplat** | 2D 图像 | 从任意角度快速预览结果 | -| **SplatToMesh** | 网格 | 转换为传统网格以供进一步编辑 | +| **SaveGLB** | `.glb` | Standard 3D file format, importable into 3D software | +| **SplatToFile3D** | `.spz` | Compressed Gaussian splat format for efficient storage | +| **RenderSplat** | 2D image | Quick preview of the result from any angle | +| **SplatToMesh** | Mesh | Convert to traditional mesh for further editing | -## 模型下载 +## Model downloads -下载 TripoSplat 模型及所需文件。放入对应的 `models/` 子目录。 +Download the TripoSplat model and required files. Place them in the corresponding `models/` subdirectories. - - triposplat_fp16.safetensors — TripoSplat 扩散模型检查点 + + triposplat_fp16.safetensors: TripoSplat diffusion model checkpoint - - triposplat_vae_decoder_fp16.safetensors — VAE 解码器 + + triposplat_vae_decoder_fp16.safetensors: VAE decoder - flux2-vae.safetensors — Flux.2 VAE，用于潜空间编码 + flux2-vae.safetensors: Flux.2 VAE for latent encoding - dino_v3_vit_h.safetensors — CLIP 视觉编码器（DINOv2） + dino_v3_vit_h.safetensors: CLIP vision encoder (DINOv2) - - birefnet.safetensors — 用于预处理的背景去除模型 + + birefnet.safetensors: Background removal model for preprocessing -### 模型存放位置 +### Model storage location ``` 📂 ComfyUI/ diff --git a/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx b/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx index 266b852e4..8fa89fef8 100644 --- a/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/zh/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "ACE-Step 1.5 ComfyUI Related Resources": 518b9b68 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 ComfyUI 中的 ACE-Step 1.5 @@ -29,27 +28,30 @@ ACE-Step 1.5 是开源音乐生成模型的重大更新，现已在 ComfyUI 中 -## 选项 1：一体化 Checkpoint（推荐） +## Option 1: All-in-One Checkpoint (Recommended) -AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和管理。 +The AIO version packages all models into a single checkpoint file, making it easier to download and manage. -### AIO 工作流 +### ACE-Step 1.5 Music Generation AIO (`audio_ace_step_1_5_checkpoint`) - - 直接在 Comfy Cloud 上运行 AIO 工作流。 - +Input style tags and lyrics to generate a full song. The workflow uses the ACE-Step 1.5 model to produce commercial-grade music in under 10 seconds on consumer hardware. - - 下载一体化 checkpoint 工作流用于本地使用。 + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation AIO" in Template Library + -### AIO 模型下载 +### AIO Model Download - - 一体化 checkpoint 文件（推荐大多数用户使用）。 + + All-in-one checkpoint file (recommended for most users). -**AIO 模型存储位置** +**AIO Model Storage Location** ``` 📂 ComfyUI/ @@ -58,39 +60,44 @@ AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和 │ └── ace_step_1.5_turbo_aio.safetensors ``` -## 选项 2：分离模型文件 +## Option 2: Split Model Files -分离版本允许你单独下载各个模型组件。 +The split version allows you to download individual model components separately. -### 分离模型工作流 +### ACE-Step 1.5 Music Generation Workflow (`audio_ace_step_1_5_split`) - - 直接在 Comfy Cloud 上运行分离模型工作流。 - +Input a text prompt describing the music style and optional lyrics. Generate a full, high-quality audio song in under 10 seconds on consumer hardware. - - 下载分离模型工作流用于本地使用。 + + + Open in Comfy Cloud + + + Download JSON or search "ACE-Step 1.5 Music Generation Workflow" in Template Library + -### 分离模型下载 +### Split Model Downloads - - 扩散模型。 + + + Diffusion model. - - 文本编码器 (0.6B)。 + + Text encoder (0.6B). - - 文本编码器 (1.7B)。 + + Text encoder (1.7B). - - VAE 模型。 + + VAE model. + -**分离模型存储位置** +**Split Models Storage Location** ``` 📂 ComfyUI/ @@ -104,26 +111,26 @@ AIO 版本将所有模型打包成单个 checkpoint 文件，更易于下载和 │ └── ace_1.5_vae.safetensors ``` -## ACE-Step 1.5 在 ComfyUI 中的主要特性 +## ACE-Step 1.5 Key Features in ComfyUI -### 思维链规划 +### Chain-of-Thought Planning -ACE-Step 1.5 模型通过思维链推理综合元数据、歌词和描述来指导扩散过程，从而产生更连贯的长篇作品。 +The ACE-Step 1.5 model synthesizes metadata, lyrics, and captions via Chain-of-Thought reasoning to guide the diffusion process, resulting in more coherent long-form compositions. -### 混合 LM + DiT 架构 +### Hybrid LM + DiT Architecture -ACE-Step 1.5 结合了规划歌曲结构的语言模型和处理音频合成的扩散 Transformer (DiT)，全部在 ComfyUI 中原生运行。 +ACE-Step 1.5 combines a Language Model that plans the song structure with a Diffusion Transformer (DiT) that handles audio synthesis, all running natively in ComfyUI. -## 即将在 ComfyUI 中推出 +## Coming Soon to ComfyUI -这些功能在 ACE-Step 1.5 中可用，但尚未在 ComfyUI 中支持： +These features are available in ACE-Step 1.5 but not yet supported in ComfyUI: -- **翻唱 (Cover)**：将任何歌曲作为输入，配合新的提示词和歌词，模型将以完全不同的风格重新演绎曲目 -- **重绘 (Repaint)**：选择一个片段，仅重新生成该部分，模型会将其拼接回去，同时保持其他部分不变 +- **Cover**: Give the model any song as input along with a new prompt and lyrics, and it will reimagine the track in a completely different style +- **Repaint**: Select a segment, regenerate just that section, and the model stitches it back in while keeping everything else untouched -## ACE-Step 1.5 ComfyUI 相关资源 +## ACE-Step 1.5 ComfyUI Related Resources -- [项目主页](https://ace-step.github.io/) -- [Hugging Face 模型](https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files) -- [GitHub 仓库](https://github.com/ace-step/ACE-Step) -- [博客文章](https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui) +- [Project Page](https://ace-step.github.io/) +- [Hugging Face](https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files) +- [GitHub](https://github.com/ace-step/ACE-Step) +- [Blog Post](https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui) diff --git a/zh/tutorials/audio/ace-step/ace-step-v1.mdx b/zh/tutorials/audio/ace-step/ace-step-v1.mdx index 8e87ef6a6..3455d83f1 100644 --- a/zh/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/zh/tutorials/audio/ace-step/ace-step-v1.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "ACE-Step Related Resources": f74e2db9 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ACE-Step是由中国团队阶跃星辰（StepFun）与ACE Studio联合开发的​​开源音乐生成基础大模型​​，旨在为音乐创作者提供高效、灵活且高质量的音乐生成与编辑工具。 @@ -66,7 +65,6 @@ ACE-Step 作为一个强大的音乐生成基座，提供了丰富的扩展能 下载下面的音频作为输入音频 -

下载示例音频文件用于输入

 @@ -87,7 +85,6 @@ ACE-Step 作为一个强大的音乐生成基座，提供了丰富的扩展能 1. 在 `TextEncodeAceStepAudio` 的 `tags` 中示例工作流中，将原本男声的 `tags` 修改为 `female voice` 来生成女声的音频 2. 在 `TextEncodeAceStepAudio` 的 `lyrics` 中示例工作流中，中对原本的歌词进行了调整修改，具体编辑你可以参考 ACE-Step 项目页面中的示例来了解如何完成修改 - ## ACE-Step 提示词指南 ACE 的提示词目前使用的有两个，一个是 `tags` 一个是 `lyrics`。 @@ -115,7 +112,6 @@ ACE 的提示词目前使用的有两个，一个是 `tags` 一个是 `lyrics` - soft electric drums（软电鼓） - melodic（旋律） - #### 场景类型 结合具体使用场景和氛围，生成符合对应氛围的音乐 @@ -152,7 +148,6 @@ ACE 的提示词目前使用的有两个，一个是 `tags` 一个是 `lyrics` {/* - 歌词编辑： - edit lyrics: 'When I was young' -> 'When you were kid'（编辑歌词示例） */} - ### 歌词（lyrics） #### 歌词结构标签 @@ -169,10 +164,8 @@ ACE 的提示词目前使用的有两个，一个是 `tags` 一个是 `lyrics` - [breakdown] (分解段) - [ad-lib] (即兴段落) - #### 多语言支持 - - ACE-Step V1 是支持多语言的，实际使用的时候 ACE-Step 会获取到对应的不同语言转换后的英文字母，然后进行音乐生成。 - 在 ComfyUI 中我们并没有完全实现全部多语言到英文字母的转换，目前仅实现了[日语平假名和片假名字符](https://github.com/Comfy-Org/ComfyUI/commit/5d3cc85e13833aeb6ef9242cdae243083e30c6fc) 所以如果你需要使用多语言来进行相关的音乐生成，你需要首先将对应的语言转换成英文字母，然后在对应 `lyrics` 开头输入对应语言代码的缩写，比如中文`[zh]` 韩语 `[ko]` 等 @@ -208,7 +201,6 @@ ACE 的提示词目前使用的有两个，一个是 `tags` 一个是 `lyrics` [fr]un amour profond qui ne s'éloignera jamais ``` - 目前 ACE-Step 支持了 19 种语言，但下面十种语言的支持会更好一些： - English - Chinese: [zh] diff --git a/zh/tutorials/audio/stable-audio/stable-audio-1.mdx b/zh/tutorials/audio/stable-audio/stable-audio-1.mdx index 29e124b64..787de81b9 100644 --- a/zh/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/zh/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -15,37 +15,42 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" **相关链接**： - [GitHub: Stability-AI/stable-audio-open-1.0](https://github.com/Stability-AI/stable-audio-open-1.0) -## 工作流 +## Workflow - - 下载 JSON 或在模板库中搜索"Stable Audio 1.0" - +### Stable Audio 1.0: Text to Audio (`audio_stable_audio_example`) - - 在 Comfy Cloud 中打开 - +Generate audio from text prompts using Stable Audio. + +Stable Audio 1.0 text to audio workflow preview -![Stable Audio 1.0 工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_example-1.webp) + + + Open in Comfy Cloud + + + Download JSON or search "Stable Audio 1.0: Text to Audio" in Template Library + + -使用**标准 ComfyUI 节点**，无需自定义节点。加载 Stable Audio 1.0 检查点，CLIP 文本编码器编码提示词，KSampler 在潜空间中降噪，最后 VAE 解码为音频。 +The workflow uses **standard ComfyUI nodes** — no custom nodes required. It loads the Stable Audio 1.0 checkpoint, encodes your prompt via a CLIP text encoder (t5-base), denoises the latent audio with a KSampler, and decodes it to audio through the model's VAE. -**使用方法**： -1. **加载模型** — `CheckpointLoaderSimple` 使用 `stable-audio-open-1.0.safetensors` -2. **写提示词** — 在 `CLIPTextEncode` 节点输入描述（例如"heaven church electronic dance music"） -3. **设置时长** — 调整 `EmptyLatentAudio` 节点的长度值（默认 47.6 秒） -4. 点击**运行**（`Ctrl/Cmd + Enter`）生成音频。文件将保存在 `ComfyUI/output/audio/` +**How to use**: +1. **Load the checkpoint** — The `CheckpointLoaderSimple` node uses `stable-audio-open-1.0.safetensors` +2. **Write a prompt** — Enter your description in the `CLIPTextEncode` node (e.g. "heaven church electronic dance music") +3. **Set duration** — Adjust the `EmptyLatentAudio` node's length value (default 47.6 seconds) +4. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -## 模型下载 +## Model download -加载工作流时，如果模型缺失，ComfyUI 会提示并提供对应下载链接。如需手动设置，请下载以下文件并放在正确目录。 +When loading the workflow, ComfyUI will prompt you with download links for any missing models. To set up manually, download the files below and place them in the correct folders. -### 检查点 +### Checkpoint - - 2.3GB。放入 models/checkpoints/ + + 2.3GB. Place in models/checkpoints/ -放在以下目录： +Place the checkpoint in: ``` 📂 ComfyUI/ @@ -54,13 +59,13 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" │ └── stable-audio-open-1.0.safetensors ``` -### 文本编码器 +### Text encoder - - 提示词处理的文本编码器。放入 models/text_encoders/ + + Text encoder for prompt conditioning. Place in models/text_encoders/ -放在以下目录： +Place the text encoder in: ``` 📂 ComfyUI/ @@ -69,4 +74,4 @@ import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" │ └── t5-base.safetensors ``` -放置完成后，在 ComfyUI 中按快捷键 **R** 刷新节点定义，即可使用最新加载的模型。 +After placing the files, press **R** in ComfyUI to refresh nodes and load the latest models. diff --git a/zh/tutorials/audio/stable-audio/stable-audio-3.mdx b/zh/tutorials/audio/stable-audio/stable-audio-3.mdx index 114dc7b44..43ba10532 100644 --- a/zh/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/zh/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Model download": d729490a --- - import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" @@ -27,65 +26,73 @@ Stable Audio 3 提供三个变体： - [Hugging Face (Comfy-Org/stable-audio-3)](https://huggingface.co/Comfy-Org/stable-audio-3) - [博客：发布公告](https://blog.comfy.org/p/stable-audio-3-day-0-support) -## 可用工作流 +## Available workflows -### Stable Audio 3 Medium +### Stable Audio 3.0 Medium (`audio_stable_audio_3_medium`) - - 下载 JSON 或在模板库中搜索"Stable Audio 3 Medium" - +Input a short text idea, optional duration, seed, and category. Generate stereo audio (music, SFX, or instruments) using Stable Audio 3 with optional AI-driven text expansion. - - 在 Comfy Cloud 中打开 +Stable Audio 3 Medium workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Stable Audio 3.0 Medium" in Template Library + -![Stable Audio 3 Medium 工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium-1.webp) +The **Stable Audio 3 Medium** workflow is a full-featured text-to-audio generation pipeline. You provide a short text idea, optional duration, seed, and category — the workflow expands your prompt using Qwen with a **category-aware reprompt template**, then generates stereo audio via the Stable Audio 3 checkpoint. -**Stable Audio 3 Medium** 工作流是一个完整的文本转音频流水线。你提供一个简短的文本创意、可选时长、种子和类别——工作流会使用 Qwen 配合分类感知的重新提示模板扩展你的提示词，然后通过 Stable Audio 3 检查点生成立体声音频。 +**How to use**: +1. **Text idea** — Enter a short description of the sound, music, or effect you want (e.g. "upbeat electronic dance track with heavy bass") +2. **Duration** — Set the desired clip length in seconds (default varies) +3. **Seed** — Control reproducibility by adjusting the seed value +4. **Category** — Choose a reprompt preset: **Music**, **Instrument**, **SFX**, or **One-shot** +5. **Enable reprompt** — Toggle `use_reprompt` on to let Qwen expand your short idea into a detailed prompt before generation +6. Click **Run** (`Ctrl/Cmd + Enter`) to generate. The audio will be saved to `ComfyUI/output/audio/` -**使用方法**： -1. **文本创意** — 输入你想生成的声音、音乐或效果的文字描述（例如"强劲节拍的电子舞曲"） -2. **时长** — 设置音频片段长度（秒） -3. **种子** — 调整种子值控制可重现性 -4. **类别** — 选择重新提示预设：**音乐（Music）**、**乐器（Instrument）**、**音效（SFX）** 或 **单次音效（One-shot）** -5. **启用重新提示** — 打开 `use_reprompt` 开关，让 Qwen 将你的短创意扩展为详细提示词后再生成 -6. 点击**运行**（`Ctrl/Cmd + Enter`）生成音频。文件将保存在 `ComfyUI/output/audio/` 目录 +### Stable Audio 3.0 Medium Base (`audio_stable_audio_3_medium_base`) -### Stable Audio 3 Medium Base +Input a short text description of a sound, music, or effect. The workflow expands your prompt with Qwen and generates a stereo audio clip from Stable Audio 3. - - 下载 JSON 或在模板库中搜索"Stable Audio 3 Medium Base" - +Stable Audio 3 Medium Base workflow preview - - 在 Comfy Cloud 中打开 + + + Open in Comfy Cloud + + Download JSON or search "Stable Audio 3.0 Medium Base" in Template Library + + -![Stable Audio 3 Medium Base 工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium_base-1.webp) - -简化版本，不包含 Qwen 提示词扩展。接收完整的文本提示词直接传递给模型。当你已经有详细的提示词、希望加快生成速度时使用。 +A simplified version of Stable Audio 3 without Qwen reprompt expansion. It expects a complete text prompt and passes it directly to the model. Use this when you already have a detailed prompt and want faster generation. -**使用方法**： -1. **文本提示词** — 输入详细的音频描述 -2. **时长** — 设置音频片段长度（秒） -3. **种子** — 控制可重现性 -4. 点击**运行**（`Ctrl/Cmd + Enter`）生成音频 +**How to use**: +1. **Text prompt** — Enter a detailed description of the audio you want +2. **Duration** — Set the clip length in seconds +3. **Seed** — Control reproducibility +4. Click **Run** (`Ctrl/Cmd + Enter`) to generate -## 模型下载 +## Model download -加载工作流时，如果模型缺失，ComfyUI 会提示并提供对应下载链接。如需手动设置，请下载以下文件并放在正确目录。 +When loading the workflow, ComfyUI will prompt you with download links for any missing models. To set up manually, download the files below and place them in the correct folders. -### 检查点 +### Checkpoints - - 用于 Medium 工作流。放入 models/checkpoints/ + + + For the Medium workflow. Place in models/checkpoints/ - - 用于 Medium Base 工作流。放入 models/checkpoints/ + + For the Medium Base workflow. Place in models/checkpoints/ + -放在以下目录： +Place checkpoints in: ``` 📂 ComfyUI/ @@ -95,17 +102,19 @@ Stable Audio 3 提供三个变体： │ └── stable_audio_3_medium_base.safetensors ``` -### 文本编码器 +### Text encoders - - 所有 Stable Audio 3 工作流都需要。放入 models/text_encoders/ + + + Required for all Stable Audio 3 workflows. Place in models/text_encoders/ - - Medium 工作流需要（Qwen 重新提示）。放入 models/text_encoders/ + + Required for the Medium workflow (Qwen reprompt). Place in models/text_encoders/ + -放在以下目录： +Place text encoders in: ``` 📂 ComfyUI/ @@ -115,4 +124,4 @@ Stable Audio 3 提供三个变体： │ └── qwen3.5_2b_bf16.safetensors ``` -放置完成后，在 ComfyUI 中按快捷键 **R** 刷新节点定义，即可使用最新加载的模型。 +After placing the files, press **R** in ComfyUI to refresh nodes and load the latest models. diff --git a/zh/tutorials/basic/image-to-image.mdx b/zh/tutorials/basic/image-to-image.mdx index 718f74475..f96ca81f3 100644 --- a/zh/tutorials/basic/image-to-image.mdx +++ b/zh/tutorials/basic/image-to-image.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Try It Yourself": 6b2e3528 --- - import InstallationModels from '/snippets/zh/tutorials/basic/installation-models.mdx' ## 什么是图生图 @@ -74,11 +73,3 @@ import InstallationModels from '/snippets/zh/tutorials/basic/installation-models 对应原理可以参考[文生图](/zh/tutorials/basic/text-to-image)教程中的原理讲解。 - - - - - - - - diff --git a/zh/tutorials/basic/inpaint.mdx b/zh/tutorials/basic/inpaint.mdx index f109e95f1..0c58b5001 100644 --- a/zh/tutorials/basic/inpaint.mdx +++ b/zh/tutorials/basic/inpaint.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "VAE Encoder (for Inpainting) Node": 222c8a2c --- - 本篇将引导了解 AI 绘图中，局部重绘的概念，并在 ComfyUI 中完成局部重绘工作流生成，我们将接触以下内容： - 使用局部重绘工作流完成画面的修改 - 了解并使用 ComfyUI 中遮罩编辑器 diff --git a/zh/tutorials/basic/lora.mdx b/zh/tutorials/basic/lora.mdx index d577558f7..9607b0235 100644 --- a/zh/tutorials/basic/lora.mdx +++ b/zh/tutorials/basic/lora.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Try It Yourself": 9f276881 --- - **LoRA 模型​（Low-Rank Adaptation）** 是一种用于微调大型生成模型（如 Stable Diffusion）的高效技术。 它通过在预训练模型的基础上引入可训练的低秩矩阵，仅调整部分参数，而非重新训练整个模型，从而以较低的计算成本实现特定任务的优化，相对于类似 SD1.5 这样的大模型，LoRA 模型更小，更容易训练。 diff --git a/zh/tutorials/basic/multiple-loras.mdx b/zh/tutorials/basic/multiple-loras.mdx index e837521ac..8506690bf 100644 --- a/zh/tutorials/basic/multiple-loras.mdx +++ b/zh/tutorials/basic/multiple-loras.mdx @@ -20,7 +20,6 @@ translationFrom: tutorials/basic/multiple-loras.mdx ![ComfyUI 中多 LoRA 模型应用示例结果](/images/tutorial/basic/multiple_loras/multiple_loras.png) - ## 相关模型安装 请下载 [dreamshaper_8.safetensors](https://civitai.com/api/download/models/128713?type=Model&format=SafeTensor&size=pruned&fp=fp16) 并保存至 `ComfyUI/models/checkpoints` 目录 diff --git a/zh/tutorials/basic/outpaint.mdx b/zh/tutorials/basic/outpaint.mdx index 1cff771a3..21909d571 100644 --- a/zh/tutorials/basic/outpaint.mdx +++ b/zh/tutorials/basic/outpaint.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "ComfyUI Outpainting Workflow Example Explanation": e06f66e2 --- - import InstallationModels from '/snippets/zh/tutorials/basic/installation-models.mdx' 本篇将引导了解 AI 绘图中扩图的概念，并在 ComfyUI 中完成扩图工作流生成。我们将接触以下内容： @@ -52,7 +51,6 @@ import InstallationModels from '/snippets/zh/tutorials/basic/installation-models ![ComfyUI Outpainting Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/basic/outpaint.png) - ### 扩图工作流使用讲解 ![ComfyUI 扩图工作流示意图](/images/tutorial/basic/outpaint/outpainting_workflow.jpg) diff --git a/zh/tutorials/basic/text-to-image.mdx b/zh/tutorials/basic/text-to-image.mdx index 37e9c8855..22444a034 100644 --- a/zh/tutorials/basic/text-to-image.mdx +++ b/zh/tutorials/basic/text-to-image.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Introduction to SD1.5 Model": cd91e138 --- - import InstallationModels from '/snippets/zh/tutorials/basic/installation-models.mdx' 本篇目的主要带你初步了解 ComfyUI 的文生图的工作流，并初步了解一些 ComfyUI 相关节点的功能和使用。 @@ -233,7 +232,6 @@ KSampler 节点的参数说明如下 **SD1.5(Stable Diffusion 1.5)** 是一个由[Stability AI](https://stability.ai/)开发的AI绘图模型，Stable Diffusion系列的基础版本，基于 **512×512** 分辨率图片训练，所以其对 **512×512** 分辨率图片生成支持较好，体积约为4GB，可以在**消费级显卡（如6GB显存）**上流畅运行。目前 SD1.5 的相关周边生态非常丰富，它支持广泛的插件（如ControlNet、LoRA）和优化工具。 作为AI绘画领域的里程碑模型，SD1.5凭借其开源特性、轻量架构和丰富生态，至今仍是最佳入门选择。尽管后续推出了SDXL/SD3等升级版本，但其在消费级硬件上的性价比仍无可替代。 - ### 基础信息 - **发布时间**：2022年10月 - **核心架构**：基于Latent Diffusion Model (LDM) diff --git a/zh/tutorials/basic/upscale.mdx b/zh/tutorials/basic/upscale.mdx index ce2834368..f296bbab0 100644 --- a/zh/tutorials/basic/upscale.mdx +++ b/zh/tutorials/basic/upscale.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Additional Tips": 5352d8c0 --- - import InstallationModels from '/snippets/zh/tutorials/basic/installation-models.mdx' ## 什么是图像放大 diff --git a/zh/tutorials/controlnet/controlnet.mdx b/zh/tutorials/controlnet/controlnet.mdx index 753c45c72..79b186084 100644 --- a/zh/tutorials/controlnet/controlnet.mdx +++ b/zh/tutorials/controlnet/controlnet.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Start Your Exploration": c5020328 --- - 在 AI 图像生成过程中，要精确控制图像生成并不是一键容易的事情，通常需要通过许多次的图像生成才可能生成满意的图像，但随着 **ControlNet** 的出现，这个问题得到了很好的解决。 ControlNet 是一种基于扩散模型（如 Stable Diffusion）的条件控制生成模型，最早由[Lvmin Zhang](https://lllyasviel.github.io/)与 Maneesh Agrawala 等人于 2023 年提出[Adding Conditional Control to Text-to-Image Diffusion Models](https://arxiv.org/abs/2302.05543) diff --git a/zh/tutorials/controlnet/depth-controlnet.mdx b/zh/tutorials/controlnet/depth-controlnet.mdx index df103196b..3264a8591 100644 --- a/zh/tutorials/controlnet/depth-controlnet.mdx +++ b/zh/tutorials/controlnet/depth-controlnet.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Combining Depth Control with Other Techniques": 9e3bcc3d --- - ## 深度图与 Depth ControlNet 介绍 深度图(Depth Map)是一种特殊的图像，它通过灰度值表示场景中各个物体与观察者或相机的距离。在深度图中，灰度值与距离成反比：​越亮的区域（接近白色）表示距离越近，​越暗的区域（接近黑色）表示距离越远。 @@ -19,7 +18,6 @@ translationBlockHashes: Depth ControlNet 是专门训练用于理解和利用深度图信息的 ControlNet 模型。它能够帮助 AI 正确解读空间关系，使生成的图像符合深度图指定的空间结构，从而实现对三维空间布局的精确控制。 - ### 深度图结合 ControlNet 应用场景 深度图在多种场景中都有比较多的应用： diff --git a/zh/tutorials/controlnet/depth-t2i-adapter.mdx b/zh/tutorials/controlnet/depth-t2i-adapter.mdx index 6c774a563..4e782a85b 100644 --- a/zh/tutorials/controlnet/depth-t2i-adapter.mdx +++ b/zh/tutorials/controlnet/depth-t2i-adapter.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Characteristics of T2I Adapter Usage": d485c810 --- - ## T2I Adapter 介绍 [T2I-Adapter](https://huggingface.co/TencentARC/T2I-Adapter) 是由 ​[腾讯ARC实验室](https://github.com/TencentARC) 开发的轻量级适配器，用于增强文本到图像生成模型（如Stable Diffusion）的结构、颜色和风格控制能力。 diff --git a/zh/tutorials/controlnet/mixing-controlnets.mdx b/zh/tutorials/controlnet/mixing-controlnets.mdx index 05e0f3c74..e4aceaf4a 100644 --- a/zh/tutorials/controlnet/mixing-controlnets.mdx +++ b/zh/tutorials/controlnet/mixing-controlnets.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Multi-dimensional Control Applications for a Single Subject": 5790d753 --- - 在 AI 图像生成中，单一的控制条件往往难以满足复杂场景的需求。混合使用多个 ControlNet 可以同时控制图像的不同区域或不同方面，实现更精确的图像生成控制。 在一些场景下，混合使用 ControlNet 可以利用不同控制条件的特性，来达到更精细的条件控制： diff --git a/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx b/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx index 63c1d973f..bc71e365b 100644 --- a/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx +++ b/zh/tutorials/controlnet/pose-controlnet-2-pass.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Advantages of 2-Pass Image Generation": fba5db79 --- - ## OpenPose 简介 [OpenPose](https://github.com/CMU-Perceptual-Computing-Lab/openpose) 是由卡耐基梅隆大学（CMU）开发的开源实时多人姿态估计系统，是计算机视觉领域的重要技术突破。该系统能够同时检测图像中多个人的： diff --git a/zh/tutorials/flux/flux-1-controlnet.mdx b/zh/tutorials/flux/flux-1-controlnet.mdx index 823ba44bf..10ef9a1b5 100644 --- a/zh/tutorials/flux/flux-1-controlnet.mdx +++ b/zh/tutorials/flux/flux-1-controlnet.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Community Versions of Flux Controlnets": 29728dcf --- - - ![Flux.1 Canny Controlnet](/images/tutorial/flux/flux-1-canny-controlnet.png) ![Flux.1 Depth Controlnet](/images/tutorial/flux/flux-1-depth-controlnet.png) @@ -43,18 +41,36 @@ Metadata 中包含工作流 json 的图片可直接拖入 ComfyUI 或使用菜 - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -## FLUX.1-Canny-dev 完整版工作流 +### Flux.1 Canny Model (`flux_canny_model_example`) + +Generate images guided by edge detection using Flux.1 Canny. + +Flux.1 Canny 工作流预览 - - Download JSON or search "Flux.1 Canny" in Template Library - 在 Comfy Cloud 中打开 + + Download JSON or search "Flux.1 Canny" in Template Library + -### 1. 工作流及相关素材 +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_canny_model_example_input_image.png` + + + +
+ flux_canny_model_example_input_image.png +
+ +**## 1. 工作流及相关素材 请下载下面的工作流图片,并拖入 ComfyUI 以加载工作流 ![ComfyUI 工作流 - ControlNet](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-canny-dev.png) @@ -113,20 +129,34 @@ ComfyUI/ - [ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet) - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -## FLUX.1-Depth-dev-lora 工作流 +### Flux.1 Depth Lora (`flux_depth_lora_example`) + +Generate images guided by depth information using Flux.1 LoRA. + +Flux.1 Depth LoRA 工作流预览 - - - Download JSON or search "Flux.1 Depth LoRA" in Template Library - 在 Comfy Cloud 中打开 + + Download JSON or search "Flux.1 Depth LoRA" in Template Library + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_depth_lora_example_input_image.png` + -LoRA 版本的工作流是在完整版本的基础上，添加了 LoRA 模型,相对于[完整版本的 Flux 工作流](/zh/tutorials/flux/flux-1-text-to-image),增加了对应 LoRA 模型的加载使用节点。 +
+ flux_depth_lora_example_input_image.png +
### 1. 工作流及相关素材 diff --git a/zh/tutorials/flux/flux-1-fill-dev.mdx b/zh/tutorials/flux/flux-1-fill-dev.mdx index 3fbaf598f..accba3bf2 100644 --- a/zh/tutorials/flux/flux-1-fill-dev.mdx +++ b/zh/tutorials/flux/flux-1-fill-dev.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Flux.1 Fill dev Outpainting Workflow": dd827926 --- - - ![Flux.1 fill dev](/images/tutorial/flux/flux-fill-dev-demo.jpeg) ## Flux.1 fill dev 模型介绍 @@ -31,7 +29,6 @@ Flux.1 fill dev 的核心特点： 本文将基于 Flux.1 fill dev 模型来完成 Inpainting 和 Outpainting 的工作流， 如果你不太了解 Inpainting 和 Outpainting 的工作流可以参考 [ComfyUI 布局重绘示例](/zh/tutorials/basic/inpaint) 和 [ComfyUI 扩图示例](/zh/tutorials/basic/outpaint)，部分的相关说明。 - ## Flux.1 Fill dev 工作流模型安装 在开始之前，让我们先完成 Flux.1 Fill dev 模型文件的安装， inpainting 和 outpainting 的工作流中会使用完全相同的模型文件，如果你之前使用过完整版本的 [Flux.1 文生图工作流](/zh/tutorials/flux/flux-1-text-to-image)，那么在这个部分你仅需要下载 **flux1-fill-dev.safetensors** 这个模型文件。 @@ -60,37 +57,41 @@ ComfyUI/ ## Flux.1 Fill dev inpainting 工作流 -### 1. Inpainting 工作流及相关素材 +### Flux.1 Inpaint (`flux_fill_inpaint_example`) + +Fill missing parts of images using Flux.1 Fill Inpainting. + +Flux.1 inpaint 工作流预览 - - Download JSON or search "flux_fill_inpaint" in Template Library - 在 Comfy Cloud 中打开 + + Download JSON or search "Flux.1 Inpaint" in Template Library + +**输入素材** + +Upload this file to the matching `LoadImage` node: + - - - Download JSON or search "flux_fill_outpaint" in Template Library - - - 在 Comfy Cloud 中打开 + + `LoadImage` node 17 · `flux_fill_inpaint_example_input_image.png` - -请下载下面的图片，并拖入 ComfyUI 以加载对应的工作流 -![ComfyUI Flux.1 inpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint.png) +
+ flux_fill_inpaint_example_input_image.png +
-请下载下面的图片，我们将使用它来作为输入图片 -![ComfyUI Flux.1 inpaint input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input.png) +**输出示例** - -对应的图片已经包含 alpha 通道，所以你不需要额外进行蒙版的绘制, 如果你想要自己进行蒙版的绘制请[点击这里](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input_original.png)获取不带蒙版的版本，并参考 [ComfyUI 布局重绘示例](/zh/tutorials/basic/inpaint) 中的 MaskEditor 的使用部分来了解如何在`Load Image`节点中绘制蒙版。 - +
+ 输入图像 + Flux.1 inpaint 输出示例 +
### 2. 参照图片序号检查完成工作流运行 @@ -106,21 +107,32 @@ ComfyUI/ ## Flux.1 Fill dev Outpainting 工作流 -### 1. Outpainting 工作流 +### Flux.1 Outpaint (`flux_fill_outpaint_example`) -请下载下面的图片，并拖入 ComfyUI 以加载对应的工作流 -![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) +Extend images beyond boundaries using Flux.1 outpainting. -请下载下面的图片，我们将使用它来作为输入图片 -![ComfyUI Flux.1 outpaint input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint_input.png) +Flux.1 outpaint 工作流预览 -### 2. 参照图片序号检查完成工作流运行 -![ComfyUI Flux.1 Fill dev Outpainting 工作流](/images/tutorial/flux/flow_diagram_outpaint.jpg) + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "Flux.1 Outpaint" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 17 · `flux_fill_outpaint_example_input_image.png` + + + +
+ flux_fill_outpaint_example_input_image.png +
-1. 确保在`Load Diffusion Model`节点加载了`flux1-fill-dev.safetensors` -2. 确保在`DualCLIPLoader`节点中下面的模型已加载： - - clip_name1: t5xxl_fp16.safetensors - - clip_name2: clip_l.safetensors -3. 确保在`Load VAE`节点中加载了`ae.safetensors` -4. 在`Load Image`节点中上传了文档中提供的输入图片 -5. 点击 `Queue` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 diff --git a/zh/tutorials/flux/flux-1-kontext-dev.mdx b/zh/tutorials/flux/flux-1-kontext-dev.mdx index 73e65d4e2..d9eb19d48 100644 --- a/zh/tutorials/flux/flux-1-kontext-dev.mdx +++ b/zh/tutorials/flux/flux-1-kontext-dev.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Flux.1 Kontext Dev Workflow": ffae9c25 --- - - import PromptTechniques from "/snippets/zh/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -44,7 +42,6 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 目前在 ComfyUI 中，你可以使用所有的这些版本，其中 [Pro 及 Max 版本](/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) 可以通过 API 节点来进行调用，而 Dev 版本开源版本请参考本篇指南中的说明。 - ## 模型下载 @@ -80,20 +77,43 @@ FLUX.1 Kontext 是 Black Forest Labs 推出的突破性多模态图像编辑模 │ └── t5xxl_fp16.safetensors 或者 t5xxl_fp8_e4m3fn_scaled.safetensors ``` -## Flux.1 Kontext Dev 工作流 +### Flux Kontext Dev Image Edit (`flux_kontext_dev_basic`) + +Smart image editing that keeps characters consistent, edits specific parts without affecting others, and preserves original styles. + +Flux Kontext Dev 工作流预览 - - - Download JSON or search "Flux Kontext Dev" in Template Library - 在 Comfy Cloud 中打开 + + Download JSON or search "Flux Kontext Dev" in Template Library + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 190 · `flux_kontext_dev_basic_input_image.jpg` + -这个工作流使用了 `Load Image(from output)` 节点来加载需要编辑的图像，可以让你更方便地获取到编辑后的图像，从而进行多轮次编辑 +
+ flux_kontext_dev_basic_input_image.jpg +
+ +**输出示例** + +
+ 输入图像 + Flux Kontext Dev 输出示例 +
+ +This workflow uses the `Load Image(from output)` node to load the image to be edited, making it more convenient for you to access the edited image for multiple rounds of editing. ### 1. 工作流及输入图片下载 diff --git a/zh/tutorials/flux/flux-1-text-to-image.mdx b/zh/tutorials/flux/flux-1-text-to-image.mdx index d465eddbc..250f7b809 100644 --- a/zh/tutorials/flux/flux-1-text-to-image.mdx +++ b/zh/tutorials/flux/flux-1-text-to-image.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Flux.1 FP8 Checkpoint Version Text-to-Image Example": daf5c58e --- - - ![Flux](/images/tutorial/flux/flux_example.png) Flux 是目前最大的开源AI绘画模型之一，拥有 12B 参数，原始文件大小约为23GB。它由 [Black Forest Labs](https://blackforestlabs.ai/) 开发，该团队由前 Stable Diffusion 团队成员创立。 Flux 以其卓越的画面质量和灵活性而闻名，能够生成高质量、多样化的图像。 @@ -47,37 +45,41 @@ Flux 以其卓越的画面质量和灵活性而闻名，能够生成高质量、 ![Flux Agreement](/images/tutorial/flux/flux_agreement.jpg) -### Flux.1 Dev 完整版本工作流 +### Flux.1 Dev fp8: Text to Image (`flux_dev_checkpoint_example`) -#### 1. 工作流文件 +Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices with limited VRAM, requires only one model file, but image quality is slightly reduced compared to the full version. -请下载下面的图片，并拖入 ComfyUI 中加载工作流。 -![Flux Dev 原始版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_dev_t5fp16.png) +Flux.1 Dev fp8 工作流预览 在 Comfy Cloud 上运行此工作流 - Download JSON or search "Flux.1 Dev" in Template Library + Download JSON or search "Flux.1 Dev fp8" in Template Library -#### 2. 手动安装模型 +**输出示例** + +![Flux.1 Dev fp8 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_checkpoint_example.png) + +#### 1. Workflow File + +#### 2. Manual Model Installation -- `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) 的协议后才能使用浏览器进行下载。 -- 如果你的显存较低，可以尝试使用 [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) 来替换 `t5xxl_fp16.safetensors` 文件。 +- The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) agreement before downloading via browser. +- If your VRAM is low, you can try using [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) to replace the `t5xxl_fp16.safetensors` file. -请下载下面的模型文件： +Please download the following model files: - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) - [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) - -文件保存位置： +Storage location: ``` ComfyUI/ ├── models/ @@ -90,59 +92,43 @@ ComfyUI/ │ └── flux1-dev.safetensors ``` -#### 3. 按步骤检查确保工作流可以正常运行 +#### 3. Steps to Run the Workflow -请参照下面的图片，确保各个模型文件都已经加载完成 +Please refer to the image below to ensure all model files are loaded correctly -![ComfyUI Flux Dev工作流](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) +![ComfyUI Flux Dev Workflow](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) -1. 确保在`DualCLIPLoader`节点中下面的模型已加载： +1. Ensure the `DualCLIPLoader` node has the following models loaded: - clip_name1: t5xxl_fp16.safetensors - clip_name2: clip_l.safetensors -2. 确保在`Load Diffusion Model`节点加载了`flux1-dev.safetensors` -3. 确保在`Load VAE`节点中加载了`ae.safetensors` -4. 点击 `Queue` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 +2. Ensure the `Load Diffusion Model` node has `flux1-dev.safetensors` loaded +3. Make sure the `Load VAE` node has `ae.safetensors` loaded +4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow -得益于 Flux 良好的提示词遵循能力，我们并不需要任何的负向提示词 +Thanks to Flux's excellent prompt following capability, we don't need any negative prompts -### Flux.1 Schnell 完整版本工作流 -#### 1. 工作流文件 - -请下载下面的图片，并拖入 ComfyUI 中加载工作流。 +### Flux.1 Schnell FP8 (`flux_schnell`) -![Flux Schnell 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_t5fp8.png) +Quickly generate images with Flux.1 Schnell fp8 quantized version. Ideal for low-end hardware, requires only 4 steps to generate images. - - - - 在 Comfy Cloud 上运行此工作流 - - - Download JSON or search "Flux.1 Dev FP8" in Template Library - - +Flux.1 Schnell FP8 checkpoint 工作流预览 -Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. 在 Comfy Cloud 上运行此工作流 - Download JSON or search "Flux.1 Schnell FP8" in Template Library + Download JSON or search "Flux.1 Schnell FP8" in Template Library -Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. - - Download JSON or search "Flux.1 Dev FP8" in Template Library - - +Please download the image below and drag it into ComfyUI to load the workflow. -请下载 [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true)并保存至 `ComfyUI/models/Checkpoints/` 目录下。 +Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. -确保对应的 `Load Checkpoint` 节点加载了 `flux1-dev-fp8.safetensors`，即可测试运行。 +Ensure that the corresponding `Load Checkpoint` node loads `flux1-schnell-fp8.safetensors`, and you can try to run the workflow. ### Flux.1 Schnell fp8 Checkpoint 版工作流 diff --git a/zh/tutorials/flux/flux-1-uso.mdx b/zh/tutorials/flux/flux-1-uso.mdx index 68d40307a..c45e51c6b 100644 --- a/zh/tutorials/flux/flux-1-uso.mdx +++ b/zh/tutorials/flux/flux-1-uso.mdx @@ -26,33 +26,41 @@ USO 支持三种主要方法： -### 1. 工作流和输入 +### Flux.1 Dev USO Reference Image Generation (`flux1_dev_uso_reference_image_gen`) -下载下方图像并拖拽到 ComfyUI 中以加载对应的工作流。 +Use reference images to control both style and subject. Keep your character's face while changing artistic style, or apply artistic styles to new scenes. -![工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/bytedance-uso.png) - - -

下载 JSON 工作流

- +Flux.1 Dev USO reference image 工作流预览 - - Download the workflow JSON and drag it into ComfyUI - 在 Comfy Cloud 上运行此工作流 + + Download the workflow JSON and drag it into ComfyUI + -使用下面的图片作为输入 +**输入素材** -![输入图像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/flux/bytedance-uso/input.png) +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 47 · `flux1_dev_uso_reference_image_gen_input_image.png` + + + +
+ flux1_dev_uso_reference_image_gen_input_image.png +
+ +**输出示例** + +
+ 输入图像 + Flux.1 Dev USO 输出示例 +
### 2. 模型链接 @@ -86,7 +94,6 @@ style={{ display: 'inline-block', backgroundColor: '#0078D6', color: '#ffffff', │ │ └── sigclip_vision_patch14_384.safetensors ``` - ### 3. 工作流说明 ![工作流说明](/images/tutorial/flux/flux1_uso_reference_image_gen.jpg) diff --git a/zh/tutorials/flux/flux-2-dev.mdx b/zh/tutorials/flux/flux-2-dev.mdx index ae5eb5839..ae4f42177 100644 --- a/zh/tutorials/flux/flux-2-dev.mdx +++ b/zh/tutorials/flux/flux-2-dev.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Model links": f4c1677b --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' diff --git a/zh/tutorials/flux/flux-2-klein.mdx b/zh/tutorials/flux/flux-2-klein.mdx index b7f715c38..986f3b692 100644 --- a/zh/tutorials/flux/flux-2-klein.mdx +++ b/zh/tutorials/flux/flux-2-klein.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Flux.2 Klein 9B Model Downloads": 0f38ad48 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' diff --git a/zh/tutorials/flux/flux1-krea-dev.mdx b/zh/tutorials/flux/flux1-krea-dev.mdx index 22962640f..c3d4c9ce1 100644 --- a/zh/tutorials/flux/flux1-krea-dev.mdx +++ b/zh/tutorials/flux/flux1-krea-dev.mdx @@ -22,78 +22,81 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **模型许可** 该模型采用 [flux-1-dev-non-commercial-license](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/LICENSE.md) 许可发布 -## Flux.1 Krea Dev ComfyUI 工作流 +### Flux.1 Krea Dev (`flux1_krea_dev`) - +A fine-tuned FLUX model pushing photorealism to the max. -#### 1. 工作流文件 - -下载下面的图片或JSON，并拖入 ComfyUI 以加载对应工作流 -![Flux Krea Dev 工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/krea/flux1_krea_dev.png) +Flux.1 Krea Dev 工作流预览 在 Comfy Cloud 上运行此工作流 - Download JSON or search "Flux.1 Krea Dev" in Template Library + Download JSON or search "Flux.1 Krea Dev" in Template Library +**输出示例** + +![Flux.1 Krea Dev 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux1_krea_dev.png) -#### 2. 模型链接 +#### 1. Workflow Files -**Diffusion model** 下面两个模型选择其中一个版本即可 +#### 2. Manual Model Installation + +Please download the following model files: +**Diffusion model** - [flux1-krea-dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/FLUX.1-Krea-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-krea-dev_fp8_scaled.safetensors) -下面这个版本是原始权重，如果你追求更高质量有足够的显存，可以尝试这个版本 +If you want to pursue higher quality and have enough VRAM, you can try the original model weights - [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) + -- `flux1-dev.safetensors` 文件需要同意 [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) 的协议后才能使用浏览器进行下载。 +The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) agreement before downloading via browser. -如果你使用过 Flux 相关的工作流，下面的模型是相同的，不需要重复下载 +If you have used Flux related workflows before, the following models are the same and don't need to be downloaded again **Text encoders** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) 当你的显存大于 32GB 时推荐使用。 +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. - [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM **VAE** - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -文件保存位置： +File save location: ``` ComfyUI/ ├── models/ │ ├── diffusion_models/ -│ │ └── flux1-krea-dev_fp8_scaled.safetensors 或 flux1-krea-dev.safetensors +│ │ └── flux1-krea-dev_fp8_scaled.safetensors or flux1-krea-dev.safetensors │ ├── text_encoders/ │ │ ├── clip_l.safetensors -│ │ └── t5xxl_fp16.safetensors 或 t5xxl_fp8_e4m3fn.safetensors +│ │ └── t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors │ ├── vae/ │ │ └── ae.safetensors ``` -#### 3. 按步骤检查确保工作流可以正常运行 +#### 3. Step-by-step Verification to Ensure Workflow Runs Properly - 对于低显存用户， 这个模型可能无法在你的设备上顺利运行，你可以等待社区提供 FP8 或 GGUF 版本 + For low VRAM users, this model may not run smoothly on your device, you can wait for the community to provide FP8 or GGUF version. -请参照下面的图片，确保各个模型文件都已经加载完成 +Please refer to the image below to ensure all model files have been loaded correctly -![ComfyUI Flux Krea Dev工作流](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) +![ComfyUI Flux Krea Dev Workflow](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) -1. 确保在`Load Diffusion Model`节点加载了`flux1-krea-dev_fp8_scaled.safetensors` 或 `flux1-krea-dev.safetensors` - - `flux1-krea-dev_fp8_scaled.safetensors` 推荐低显存用户使用 - - `flux1-krea-dev.safetensors` 如果你有足够的显存如 24GB， 你可以尝试这个版本以追求更好的质量 -2. 确保在`DualCLIPLoader`节点中下面的模型已加载： - - clip_name1: t5xxl_fp16.safetensors 或 t5xxl_fp8_e4m3fn.safetensors +1. Ensure that `flux1-krea-dev_fp8_scaled.safetensors` or `flux1-krea-dev.safetensors` is loaded in the `Load Diffusion Model` node + - `flux1-krea-dev_fp8_scaled.safetensors` is recommended for low VRAM users + - `flux1-krea-dev.safetensors` is the original weights, if you have enough VRAM like 24GB you can use it for better quality +2. Ensure the following models are loaded in the `DualCLIPLoader` node: + - clip_name1: t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors - clip_name2: clip_l.safetensors -3. 确保在`Load VAE`节点中加载了`ae.safetensors` -4. 确保 -4. 点击 `Queue` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 \ No newline at end of file +3. Ensure that `ae.safetensors` is loaded in the `Load VAE` node +4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow diff --git a/zh/tutorials/image/anima/anima.mdx b/zh/tutorials/image/anima/anima.mdx index d2691102f..b7245f93c 100644 --- a/zh/tutorials/image/anima/anima.mdx +++ b/zh/tutorials/image/anima/anima.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Limitations": 0adbf360 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Anima** 是由 [CircleStone Labs](https://huggingface.co/circlestone-labs/Anima) 与 Comfy Org 合作推出的开源文生图模型。该模型拥有 **20 亿** 参数，专注于生成高质量的 **动漫和非照片级** 图像——非常适合角色设计、场景绘制和概念艺术。 @@ -36,47 +35,57 @@ Anima 提供两个工作流——基础版适用于标准使用，预览版适 此工作流使用 Subgraph 节点实现模块化处理。查阅 Subgraph 文档了解如何自定义和扩展工作流。 -### Anima Base v1：文生图 +### Anima Base v1: Text to Image (`image_anima_base_v1`) - - 下载 JSON 或在模板库中搜索 "Anima Base v1" - +Input a text prompt describing an anime or artistic illustration. Generate a non-photorealistic image focused on anime concepts, characters, or styles. - +Anima Base v1 text-to-image 工作流预览 + + + 在 Comfy Cloud 中打开 + + Download JSON or search "Anima Base v1" in Template Library + + -#### 快速开始 +**输出示例** -1. 将 ComfyUI 更新至最新版本，或使用 [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) -2. 前往 **模板**，搜索 **Anima Base v1** -3. 选择 **Anima Base v1: Text to Image** 工作流 -4. 下载缺失的模型（参见[模型下载](#anima-模型下载)），输入提示词后点击 **运行** +![Anima Base v1 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_base_v1.png) -#### 示例输出 +#### Get started -Anima Base v1 示例输出 +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) +2. Go to **Template** and search for **Anima Base v1** +3. Select the **Anima Base v1: Text to Image** workflow +4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** -### Anima Preview：动漫文生图 +### Anima Preview: Anime Text-to-Image Generation (`image_anima_preview`) - - 下载 JSON 或在模板库中搜索 "Anima Preview" - +Input a text prompt to generate an anime-style image using the Anima model. Configure settings like steps and CFG scale to control the output. + +Anima Preview text-to-image 工作流预览 - + + 在 Comfy Cloud 中打开 + + Download JSON or search "Anima Preview" in Template Library + + -#### 快速开始 +**输出示例** -1. 将 ComfyUI 更新至最新版本，或使用 [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) -2. 前往 **模板**，搜索 **Anima Preview** -3. 选择 **Anima Anime Text-to-Image Generation** 工作流 -4. 下载缺失的模型（参见[模型下载](#anima-模型下载)），输入提示词后点击 **运行** +![Anima Preview 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_preview.png) -#### 示例输出 +#### Get started -Anima Preview 示例输出 +1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) +2. Go to **Template** and search for **Anima Preview** +3. Select the **Anima Anime Text-to-Image Generation** workflow +4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** ## Anima 模型下载 diff --git a/zh/tutorials/image/boogu/boogu-image-0.1.mdx b/zh/tutorials/image/boogu/boogu-image-0.1.mdx index 7d70429c3..363790595 100644 --- a/zh/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/zh/tutorials/image/boogu/boogu-image-0.1.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Boogu-Image-0.1-Edit image editing workflow": 3c1752bb --- - import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" **Boogu-Image-0.1** 是一个采用 Apache-2.0 许可证的开源统一图像生成与编辑模型系列。它通过整合理解与生成能力的系统，在摄影、文字渲染、风格化和图像编辑任务中均能提供具有竞争力的表现。 diff --git a/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index ad33ee57c..9eab0eb6f 100644 --- a/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/zh/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -49,7 +49,6 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors) - 文件保存位置 ``` 📂 ComfyUI/ diff --git a/zh/tutorials/image/ernie-image/ernie-image.mdx b/zh/tutorials/image/ernie-image/ernie-image.mdx index 51f8c668a..f53b64a78 100644 --- a/zh/tutorials/image/ernie-image/ernie-image.mdx +++ b/zh/tutorials/image/ernie-image/ernie-image.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Examples": 0e2eb115 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **ERNIE-Image** 是百度开源的文生图模型，基于 Apache-2.0 协议发布。该模型基于 **8B** 参数的扩散变换器（DiT）构建，支持精准文字渲染、强大的指令跟随以及结构化视觉生成，可产出高质量图像。 @@ -31,18 +30,28 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [Hugging Face（ERNIE-Image）](https://huggingface.co/Baidu/ERNIE-Image) - [Hugging Face（ComfyUI 支持）](https://huggingface.co/Comfy-Org/ERNIE-Image) - ## ERNIE-Image 文生图工作流 - - 下载 ERNIE-Image 文生图工作流 JSON 文件。 - + + +### Ernie Image: Text to Image (`image_ernie_image`) + +Generate images from text prompts using the ERNIE-Image model. Input a text description to produce detailed, structured visuals with a broad stylistic range. - - 在 Comfy Cloud 上直接运行此工作流。 +ERNIE-Image text-to-image 工作流预览 + + + + 在 Comfy Cloud 中直接运行此工作流 + + + Download the ERNIE-Image text-to-image workflow JSON file + - +**输出示例** + +![ERNIE-Image 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image.png) ### 快速开始 @@ -55,21 +64,20 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 所有重新打包的模型文件均可在 Hugging Face 的 [Comfy-Org/ERNIE-Image](https://huggingface.co/Comfy-Org/ERNIE-Image) 获取。 - + + ERNIE-Image 扩散模型。 - - + ERNIE-Image 文本编码器。 - - + ERNIE-Image 提示词增强器文本编码器。 - - + ERNIE-Image VAE。 + **模型存储位置** @@ -87,33 +95,43 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## ERNIE-Image-Turbo -[ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo) 是经 DMD 和 RL 优化的快速变体，仅需 **8 步** 即可生成图像，而标准模型需要约 50 步。 +[ERNIE-Image-Turbo](https://huggingface.co/Baidu/ERNIE-Image-Turbo) 是经 DMD 和 RL 优化的更快变体，仅需 **8 步**即可生成图像，而标准模型约需 50 步。 - - 下载 ERNIE-Image-Turbo 文生图工作流 JSON 文件。 - +### Ernie Image Turbo: Text To Image (`image_ernie_image_turbo`) - - 在 Comfy Cloud 上直接运行此工作流。 +Generate images from text prompts using the ERNIE-Image turbo model. Input a text description and receive a high-quality image with precise text rendering. + +ERNIE-Image-Turbo text-to-image 工作流预览 + + + + 在 Comfy Cloud 中直接运行此工作流 + + + Download the ERNIE-Image-Turbo text-to-image workflow JSON file + + +**输出示例** + +![ERNIE-Image-Turbo 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ernie_image_turbo.png) ### ERNIE-Image-Turbo 模型下载 - + + ERNIE-Image-Turbo 扩散模型。 - - + ERNIE-Image-Turbo 文本编码器。 - - + ERNIE-Image-Turbo 提示词增强器文本编码器。 - - + ERNIE-Image-Turbo VAE。 + **模型存储位置** diff --git a/zh/tutorials/image/hidream/hidream-e1.mdx b/zh/tutorials/image/hidream/hidream-e1.mdx index 924d4fe4b..67a874c3a 100644 --- a/zh/tutorials/image/hidream/hidream-e1.mdx +++ b/zh/tutorials/image/hidream/hidream-e1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ![HiDream-E1 演示](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -74,16 +72,34 @@ HiDream-E1 是智象未来(HiDream-ai) 正式开源的交互式图像编辑大 ## HiDream E1.1 ComfyUI 原生工作流示例 -E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一百万分辨率，在工作流中使用了 `Scale Image to Total Pixels` 节点来将输入图片动态调整为 1百万像素 +### HiDream E1.1 Image Editing (`hidream_e1_1`) - -这里是在测试使用时对应的显存占用参考： -1. A100 40GB (VRAM 使用率 95%)：第一次生成: 211s，第二次生成: 73s +Edit images with HiDream E1.1. Superior image quality and editing accuracy compared to HiDream-E1-Full. -2. 4090D 24GB (VRAM 使用率 98%) -- 完整版本: Out of memory -- FP8_e4m3fn_fast (VRAM 98%) 第一次生成: 120s， 第二次生成: 91s - +HiDream E1.1 image editing 工作流预览 + + + + 在 Comfy Cloud 中打开 + + + Download JSON or search "HiDream E1.1" in Template Library + + + +**输入素材** + +将此文件上传到匹配的 `LoadImage` 节点： + + + + `LoadImage` node 13 · `hidream_e1_1_input_image.jpg` + + + +
+ hidream_e1_1_input_image.jpg +
### 1. HiDream E1.1 工作流及相关素材 @@ -116,15 +132,44 @@ E1.1 是于 2025年7月16日更新迭代的版本, 这个版本支持动态一 ## HiDream E1 ComfyUI 原生 工作流示例 +### HiDream E1 Image Edit (`hidream_e1_full`) + +Edit images with HiDream E1. Professional natural language image editing model. + +HiDream E1 image editing 工作流预览 + - + 在 Comfy Cloud 中打开 - - Download JSON or search "HiDream E1.1" in Template Library + + Download JSON or search "HiDream E1 Full" in Template Library + + + +**输入素材** + +将此文件上传到匹配的 `LoadImage` 节点： + + + + `LoadImage` node 13 · `hidream_e1_full_input_image.jpg` +
+ hidream_e1_full_input_image.jpg +
+ +**输出示例** + +
+ 输入图像 + HiDream E1 输出示例 +
+ +E1 is a model released on April 28, 2025. + E1 是于 2025 年 4 月 28 日发布的，这个模型只支持 768*768 的分辨率 ### 1. HiDream-e1 工作流及相关素材 diff --git a/zh/tutorials/image/hidream/hidream-i1.mdx b/zh/tutorials/image/hidream/hidream-i1.mdx index 605c94338..01112bc1f 100644 --- a/zh/tutorials/image/hidream/hidream-i1.mdx +++ b/zh/tutorials/image/hidream/hidream-i1.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Other Related Resources": e794d9ef --- - - ![HiDream-I1 演示](https://raw.githubusercontent.com/HiDream-ai/HiDream-I1/main/assets/demo.jpg) HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图模型。该模型拥有17B参数规模，采用 [MIT 许可证](https://github.com/HiDream-ai/HiDream-I1/blob/main/LICENSE) 发布，支持用于个人项目、科学研究以及商用，目前在多项基准测试中该模型表现优异。 @@ -96,134 +94,155 @@ HiDream-I1 是智象未来(HiDream-ai)于2025年4月7日正式开源的文生图 │ └── ... # 将在对应版本的工作流中引导你进行安装 ``` -### HiDream-I1 full 版本工作流 +### HiDream I1 Full (`hidream_i1_full`) + +Generate images with HiDream I1 Full. Complete version with 50 inference steps for highest quality output. + +HiDream I1 Full 工作流预览 - Run this workflow on Comfy Cloud with zero setup + 零配置在 Comfy Cloud 上运行此工作流 - 下载工作流 JSON 文件 + Download the workflow JSON file -#### 1. 模型文件下载 -请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 +**输出示例** + +![HiDream I1 Full 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_full.png) -- FP8 版本：[hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) 需要 27GB 以上的显存 +#### 1. Model File Download -#### 2. 工作流文件下载 +Please select the appropriate version based on your hardware. Click the link and download the corresponding model file to save it to the `ComfyUI/models/diffusion_models/` folder. -请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 -![HiDream-I1 full 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) +- FP8 version: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) requires more than 27GB of VRAM -#### 3. 按步骤完成工作流的运行 +#### 2. Workflow File Download -![HiDream-I1 full 版本步骤图](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) +Please download the image below and drag it into ComfyUI to load the corresponding workflow +![HiDream-I1 Full Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) -按步骤完成工作流的运行 -1. 确保`Load Diffusion Model` 节点中使用的是 `hidream_i1_full_fp8.safetensors` 文件 -2. 确保`QuadrupleCLIPLoader` 中四个对应的 text encoder 被正确加载 +#### 3. Complete the Workflow Step by Step + +![HiDream-I1 Full Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) + +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_full_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. 确保`Load VAE` 节点中使用的是 `ae.safetensors` 文件 -4. 对于 **full** 版本你需要设置 `ModelSamplingSD3` 中的 `shift` 参数为 `3.0` -5. 对于 `Ksampler` 节点，你需要进行以下设置 - - `steps` 设置为 `50` - - `cfg` 设置为 `5.0` - - (可选) `sampler` 设置为 `lcm` - - (可选) `scheduler` 设置为 `normal` -6. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行图片生成 +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **full** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `50` + - Set `cfg` to `5.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation + +### HiDream I1 Dev (`hidream_i1_dev`) -### HiDream-I1 dev 版本工作流 +Generate images with HiDream I1 Dev. Balanced version with 28 inference steps, suitable for medium-range hardware. + +HiDream I1 Dev 工作流预览 - Run this workflow on Comfy Cloud with zero setup + 零配置在 Comfy Cloud 上运行此工作流 - 下载工作流 JSON 文件 + Download the workflow JSON file -#### 1. 模型文件下载 -请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 +**输出示例** + +![HiDream I1 Dev 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_dev.png) -- FP8 版本：[hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) 需要 27GB 以上的显存 +#### 1. Model File Download +Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -#### 2. 工作流文件下载 -请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 +- FP8 version: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) requires more than 27GB of VRAM -![HiDream-I1 dev 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) +#### 2. Workflow File Download +Please download the image below and drag it into ComfyUI to load the corresponding workflow -#### 3. 按步骤完成工作流的运行 +![HiDream-I1 Dev Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) -![HiDream-I1 dev 版本步骤图](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) -按步骤完成工作流的运行 -1. 确保`Load Diffusion Model` 节点中使用的是 `hidream_i1_dev_fp8.safetensors` 文件 -2. 确保`QuadrupleCLIPLoader` 中四个对应的 text encoder 被正确加载 +#### 3. Complete the Workflow Step by Step + +![HiDream-I1 Dev Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_dev_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. 确保`Load VAE` 节点中使用的是 `ae.safetensors` 文件 -4. 对于 **dev** 版本你需要设置 `ModelSamplingSD3` 中的 `shift` 参数为 `6.0` -5. 对于 `Ksampler` 节点，你需要进行以下设置 - - `steps` 设置为 `28` - - (重要) `cfg` 设置为 `1.0` - - (可选) `sampler` 设置为 `lcm` - - (可选) `scheduler` 设置为 `normal` -6. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行图片生成 +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **dev** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `6.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `28` + - (Important) Set `cfg` to `1.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation + +### HiDream I1 Fast (`hidream_i1_fast`) + +Generate images quickly with HiDream I1 Fast. Lightweight version with 16 inference steps, ideal for rapid previews on lower-end hardware. -### HiDream-I1 fast 版本工作流 +HiDream I1 Fast 工作流预览 - - Run this workflow on Comfy Cloud with zero setup + 零配置在 Comfy Cloud 上运行此工作流 - 下载工作流 JSON 文件 + Download the workflow JSON file -- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM - -#### 1. 模型文件下载 -请根据你的硬件情况选择合适的版本，点击链接并下载对应的模型文件保存到 `ComfyUI/models/diffusion_models/` 文件夹下。 +**输出示例** + +![HiDream I1 Fast 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_fast.png) + +#### 1. Model File Download +Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. -- FP8 版本：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) 需要 16GB 以上的显存 -- 完整版本：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) 需要 27GB 以上的显存 +- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM +- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM -#### 2. 工作流文件下载 -请下载下面的图片，并拖入 ComfyUI 中以加载对应的工作流 +#### 2. Workflow File Download +Please download the image below and drag it into ComfyUI to load the corresponding workflow -![HiDream-I1 fast 版本工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) +![HiDream-I1 Fast Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) -#### 3. 按步骤完成工作流的运行 +#### 3. Complete the Workflow Step by Step -![HiDream-I1 fast 版本步骤图](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) +![HiDream-I1 Fast Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) -按步骤完成工作流的运行 -1. 确保`Load Diffusion Model` 节点中使用的是 `hidream_i1_fast_fp8.safetensors` 文件 -2. 确保`QuadrupleCLIPLoader` 中四个对应的 text encoder 被正确加载 +Complete the workflow execution step by step +1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_fast_fp8.safetensors` file +2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. 确保`Load VAE` 节点中使用的是 `ae.safetensors` 文件 -4. 对于 **fast** 版本你需要设置 `ModelSamplingSD3` 中的 `shift` 参数为 `3.0` -5. 对于 `Ksampler` 节点，你需要进行以下设置 - - `steps` 设置为 `16` - - (重要) `cfg` 设置为 `1.0` - - (可选) `sampler` 设置为 `lcm` - - (可选) `scheduler` 设置为 `normal` -6. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行图片生成 +3. Make sure the `Load VAE` node is using the `ae.safetensors` file +4. For the **fast** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` +5. For the `Ksampler` node, you need to make the following settings + - Set `steps` to `16` + - (Important) Set `cfg` to `1.0` + - (Optional) Set `sampler` to `lcm` + - (Optional) Set `scheduler` to `normal` +6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation ## 使用建议 diff --git a/zh/tutorials/image/hidream/hidream-o1.mdx b/zh/tutorials/image/hidream/hidream-o1.mdx index 4ef55b41f..2128c9074 100644 --- a/zh/tutorials/image/hidream/hidream-o1.mdx +++ b/zh/tutorials/image/hidream/hidream-o1.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Additional Notes": 762758eb --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -33,50 +32,24 @@ HiDream-O1-Image 基于 [MIT 协议](https://github.com/HiDream-ai/HiDream-O1-Im ## HiDream-O1-Image Full 工作流 -### 1. 工作流文件下载 +### HiDream O1 Full: Image generation (`image_hidream_o1`) -请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `Image` 找到 "HiDream O1 Full: Image generation" 以加载工作流。 +Input a text prompt and optionally upload reference images. Generate a high-resolution image up to 2048x2048 with text-to-image, editing, or subject-driven personalization. -![HiDream-O1-Image Full 工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) +HiDream O1 Full 工作流预览 - - Download workflow + + + 在 Comfy Cloud 中打开 - - - Open in cloud + + Download JSON or search "HiDream O1 Full" in Template Library + -### 2. 手动下载模型 - -**Checkpoint** — 经过重新打包和量化。所有版本均将最严重的离群值保留在 bf16，并移除了未使用的 deepstack 层： - -- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 -- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8 量化变体 -- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_bf16.safetensors) — bf16 全精度版（文件最大） - -**文本编码器（提示词优化）** — 所有版本通用： +**输出示例** -- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) - -**LoRA（可选）** — Dev 蒸馏也可以作为 LoRA 应用到 Full 模型中，让你可以调节蒸馏强度（由 [Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： - -- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — 全秩版 -- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — 剪枝版 -- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/resolve/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 基于 checkpoint 的替代蒸馏 - -``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 checkpoints/ -│ │ ├── hidream_o1_image_fp8_scaled.safetensors -│ │ ├── hidream_o1_image_mxfp8.safetensors -│ │ └── hidream_o1_image_bf16.safetensors -│ ├── 📂 loras/ -│ │ └── hidream_o1_dev_lora_rank_64_bf16.safetensors -│ └── 📂 text_encoders/ -│ └── gemma4_e4b_it_fp8_scaled.safetensors -``` +![HiDream O1 Full 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) ### 3. 使用工作流 @@ -89,42 +62,41 @@ HiDream-O1-Image 基于 [MIT 协议](https://github.com/HiDream-ai/HiDream-O1-Im ## HiDream-O1-Image Dev 工作流 -### 1. 工作流文件下载 +### HiDream O1 Dev (`image_hidream_o1_dev`) -请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` -> `Image` 找到 "HiDream O1 Dev" 以加载工作流。 +Input a text prompt and optional reference images. Generate a high-resolution image (up to 2048x2048) with support for text-to-image, image editing, and subject-driven personalization. -![HiDream-O1-Image Dev 工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1_dev.png) +HiDream O1 Dev 工作流预览 - - Download workflow + + + 在 Comfy Cloud 中打开 - - - Open in cloud + + Download JSON or search "HiDream O1 Dev" in Template Library + -### 2. 手动下载模型 +**输入素材** -**Checkpoint（Dev 版）** — 经过重新打包和量化。所有版本均将最严重的离群值保留在 bf16，并移除了未使用的 deepstack 层： +Upload this file to the matching `LoadImage` node: -- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8 量化版，在支持的硬件上使用 fp8/mxfp8 矩阵乘法加速安全 MLP 层 -- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8 量化变体 -- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/resolve/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — bf16 全精度版（文件最大） + + + `LoadImage` node 213 · `noir_portrait.png` + + -**文本编码器（提示词优化）** — 所有版本通用： +
+ noir_portrait.png +
-- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/resolve/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) +**输出示例** -``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 checkpoints/ -│ │ ├── hidream_o1_image_dev_fp8_scaled.safetensors -│ │ ├── hidream_o1_image_dev_mxfp8.safetensors -│ │ └── hidream_o1_image_dev_bf16.safetensors -│ └── 📂 text_encoders/ -│ └── gemma4_e4b_it_fp8_scaled.safetensors -``` +
+ 输入图像 + HiDream O1 Dev 输出示例 +
### 3. 使用工作流 diff --git a/zh/tutorials/image/ideogram/ideogram-v4.mdx b/zh/tutorials/image/ideogram/ideogram-v4.mdx index bc4f330ed..a5f5f4b1d 100644 --- a/zh/tutorials/image/ideogram/ideogram-v4.mdx +++ b/zh/tutorials/image/ideogram/ideogram-v4.mdx @@ -10,25 +10,30 @@ translationBlockHashes: "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- - import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; Ideogram 4.0 是 Ideogram 最新推出的文生图模型，已作为开源模型发布，可以在本地完全运行。它具有出色的照片级真实感、精准的文字渲染和风格控制能力。你可以使用自然语言或 **结构化 JSON Prompts** 来实现对布局、颜色和图片内文字的精细控制。 -## Ideogram 4.0 文生图工作流 +### Ideogram v4: Text to Image (`image_ideogram4_t2i`) + +Input a text prompt or structured JSON description. Generate an image with precise layout, color, and style control using Ideogram 4.0. - +Ideogram 4.0 text-to-image 工作流预览 + + + 在 Comfy Cloud 中打开 - - - 下载 JSON 或在模板库中搜索"Ideogram v4: Text to Image" + + Download JSON or search "Ideogram v4: Text to Image" in Template Library + + +**输出示例** -![Ideogram 4.0 生成示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) -*Ideogram 4.0 开源模型生成示例* +![Ideogram 4.0 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) ### Prompt 格式 @@ -46,25 +51,23 @@ Ideogram 4.0 是 Ideogram 最新推出的文生图模型，已作为开源模型 你可以在 Hugging Face 的 [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) 找到所有重新打包的模型文件。 - + + Ideogram 4.0 扩散模型（~13.8 GB）。放入 models/diffusion_models/ - - + Ideogram 4.0 无条件扩散模型（~13.8 GB）。放入 models/diffusion_models/ - - + Ideogram 4.0 文本编码器（~8 GB）。放入 models/text_encoders/ - - + Ideogram 4.0 文本编码器（~2 GB）。放入 models/text_encoders/ - - + Ideogram 4.0 VAE（~335 MB）。放入 models/vae/ + **模型存放路径** diff --git a/zh/tutorials/image/krea/krea-2.mdx b/zh/tutorials/image/krea/krea-2.mdx index e3f110dac..5d7855899 100644 --- a/zh/tutorials/image/krea/krea-2.mdx +++ b/zh/tutorials/image/krea/krea-2.mdx @@ -12,9 +12,6 @@ translationBlockHashes: "Krea-2 Turbo style reference workflow": 5e966a16 --- - - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -46,30 +43,32 @@ Krea 2 提供两个版本： - [GitHub 官方仓库](https://github.com/krea-ai/krea-2) - [技术报告](https://www.krea.ai/blog/krea-2-technical-report) -## Krea-2 Turbo 文生图工作流 +### Krea-2: Text to Image (`image_krea2_turbo_t2i`) + +Generate images from text prompts using Krea 2, a foundation model built for aesthetic quality and creative control. It focuses on rendering expressive, stylistically diverse images. -Krea-2 Turbo 文生图工作流 +Krea-2 Turbo text-to-image 工作流预览 - + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索"Krea-2" + Download JSON or search "Krea-2" in Template Library -工作流分为以下几个部分： +The workflow is organized into a few parts: -1. **Text to Image (Krea-2 Turbo) 子图**：核心生成管线，包含模型加载、提示词处理、采样和 VAE 解码 -2. **ResolutionSelector**：选择输出分辨率。Krea 2 支持 1K 到 2K 的输出，将 megapixels 值设为 2.0 即可获得 2K 分辨率。 -3. **CustomCombo (LoRA 选择器)**：预置的触发词选择器，对应可用的风格 LoRA。如果你下载了额外的 LoRA，可以自定义该选择器并与对应的 LoRA 文件配对使用。 -4. **SaveImage**：保存生成的图像 +1. **Text to Image (Krea-2 Turbo) subgraph**: the core generation pipeline, containing model loading, prompt handling, sampling, and VAE decode +2. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K — set the megapixels value to 2.0 to get 2K resolution. +3. **CustomCombo (LoRA selector)**: a pre-built trigger word selector for the available style LoRAs. If you download additional LoRAs, you can customize this selector and pair them with the corresponding LoRA files accordingly. +4. **SaveImage**: saves the generated image - - 本工作流使用子图节点进行模块化处理。查看子图文档了解如何定制和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ### 一键生成 @@ -145,27 +144,47 @@ Krea 还发布了一组 Krea 2 的风格 LoRA。在 **CustomCombo** 节点中选 │ └── krea2_softwatercolor.safetensors (及其它风格 LoRA) ``` -## Krea-2 Turbo 风格参考工作流 +### Krea-2 Int8: Image Style Reference (`image_krea2_turbo_int8_image_style_reference`) -Krea-2 Turbo 风格参考工作流 +Generate images with the Krea-2 Turbo model while referencing the style of 1–2 uploaded images, using the high-performance Int8 Convrot format for fast inference. + +Krea-2 Turbo style reference 工作流预览 - + 在 Comfy Cloud 中打开 - 下载 JSON 或在模板库中搜索"Krea-2 风格参考" + Download JSON or search "Krea-2 Style Reference" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 69 · `krea2_reference_image.png` -风格参考工作流在 Krea-2 Turbo 流水线的基础上增加了参考图像条件功能。上传一张或多张参考图像，以影响生成输出的美学风格、氛围和视觉方向。 +
+ krea2_reference_image.png +
+ +**输出示例** + +![Krea-2 style reference 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_krea2_turbo_int8_image_style_reference.png) + +The style reference workflow builds on the Krea-2 Turbo pipeline by adding reference image conditioning. Upload one or more reference images to influence the aesthetic style, mood, and visual direction of the generated output. -工作流分为以下几个部分： +The workflow is organized into a few parts: -1. **图像风格参考（Krea-2 Turbo）子图**：带有风格参考支持的核心生成流水线，包含模型加载、参考图像条件、提示词处理和采样 -2. **LoadImage**：上传风格参考图像 -3. **ResolutionSelector**：选择所需的输出分辨率。Krea 2 支持从 1K 到 2K 的输出。 -4. **SaveImage**：保存已生成的图像 +1. **Image Style Reference (Krea-2 Turbo) subgraph**: the core generation pipeline with style reference support, containing model loading, reference image conditioning, prompt handling, and sampling +2. **LoadImage**: upload your style reference images +3. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K. +4. **SaveImage**: saves the generated image ### 使用风格参考 diff --git a/zh/tutorials/image/lens/lens.mdx b/zh/tutorials/image/lens/lens.mdx index 14f673e34..f4bdfee7a 100644 --- a/zh/tutorials/image/lens/lens.mdx +++ b/zh/tutorials/image/lens/lens.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Available models": 5876b860 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Lens** 是 **微软** 开源的文生图模型，采用 MIT 许可证。拥有 **38亿** 参数，采用 **双流 MMDiT** 架构，配合 **GPT-OSS-20B** 文本编码器特征和 **FLUX.2 语义 VAE**，以远少于大模型的训练算力实现了具有竞争力的图像质量。 @@ -37,57 +36,63 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 本工作流使用 Subgraph 节点实现模块化处理。查看 Subgraph 文档了解如何自定义和扩展工作流。 -### Lens +### Lens: Text to Image (`image_lens_t2i`) + +Input a text prompt and select resolution and aspect ratio. Generate a high-quality image using the efficient Lens text-to-image model. + +Lens text-to-image 工作流预览 - 下载 JSON 或在模板库中搜索 "Lens" + Download JSON or search "Lens" in Template Library - {/* TODO: Lens 在 Comfy Cloud 上线后启用 */} - {/**/} + {/* TODO: Enable Cloud template when Lens is available on Comfy Cloud */} + {/**/} {/* 在 Comfy Cloud 中打开*/} {/**/} - +**输出示例** + +![Lens text-to-image 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_t2i.png) -#### 开始使用 + -1. 更新 ComfyUI 到最新版本 - {/* TODO: Cloud 模板上线后添加 Cloud 选项 */} -2. 进入 **模板** 搜索 **Lens** -3. 选择 **Lens** 工作流 -4. 下载缺失的模型（见 [模型下载](#模型下载)），输入提示词，点击 **队列运行** +#### Get started -#### 示例输出 +1. Update ComfyUI to the latest version + {/* TODO: Add Cloud option when template is available */} +2. Go to **Template** and search for **Lens** +3. Select the **Lens** workflow +4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** -Lens 文生图示例输出 +### Lens Turbo: Text to Image (`image_lens_turbo_t2i`) -### Lens Turbo +Input a text prompt and select resolution, aspect ratio, and inference steps. Generate a high-quality image using the Lens text-to-image model. -Lens Turbo 是蒸馏版，只需较少的采样步数即可生成图像，推理速度更快。 +Lens Turbo text-to-image 工作流预览 - 下载 JSON 或在模板库中搜索 "Lens Turbo" + Download JSON or search "Lens Turbo" in Template Library - {/* TODO: Lens Turbo 在 Comfy Cloud 上线后启用 */} - {/**/} + {/* TODO: Enable Cloud template when Lens Turbo is available on Comfy Cloud */} + {/**/} {/* 在 Comfy Cloud 中打开*/} {/**/} -#### 开始使用 +**输出示例** -1. 更新 ComfyUI 到最新版本 - {/* TODO: Cloud 模板上线后添加 Cloud 选项 */} -2. 进入 **模板** 搜索 **Lens Turbo** -3. 选择 **Lens Turbo** 工作流 -4. 下载缺失的模型（见 [模型下载](#模型下载)），输入提示词，点击 **队列运行** +![Lens Turbo text-to-image 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_turbo_t2i.png) -#### 示例输出 +#### Get started -Lens Turbo 文生图示例输出 +1. Update ComfyUI to the latest version + {/* TODO: Add Cloud option when template is available */} +2. Go to **Template** and search for **Lens Turbo** +3. Select the **Lens Turbo** workflow +4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** ## 模型下载 diff --git a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index d1a8a5156..6b61924e4 100644 --- a/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/zh/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Prompt format": 5b819c6c --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **NewBie-image-Exp0.1** 是由 NewBieAI Lab 开发的 35 亿参数 DiT 模型，专为动漫风格文生图设计。基于 Next-DiT 架构构建，能够生成细节丰富、视觉效果出色的动漫图像。 @@ -29,19 +27,24 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [Hugging Face](https://huggingface.co/NewBie-AI/NewBie-image-Exp0.1) - [入门指南](https://ai.feishu.cn/wiki/P3sgwUUjWih8ZWkpr0WcwXSMnTb) -## NewBie-image 文生图工作流 +### NewBie Exp0.1: Anime Generation (`image_newbieimage_exp0_1-t2i`) + +Generate detailed anime-style images with NewBie Exp0.1's Next-DiT architecture. Supports XML structured prompts for better multi-character scenes and attribute binding. + +NewBie-image text-to-image 工作流预览 - - - Download JSON or search "NewBie-image" in Template Library + + 在 Comfy Cloud 中打开 - - Open in cloud + + Download JSON or search "NewBie-image" in Template Library - +**输出示例** + +![NewBie-image 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_newbieimage_exp0_1-t2i.png) diff --git a/zh/tutorials/image/omnigen/omnigen2.mdx b/zh/tutorials/image/omnigen/omnigen2.mdx index fd70c11b8..202fad0bb 100644 --- a/zh/tutorials/image/omnigen/omnigen2.mdx +++ b/zh/tutorials/image/omnigen/omnigen2.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "ComfyUI OmniGen2 Image Editing Workflow": 1e06072e --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## 关于 OmniGen2 @@ -51,7 +49,6 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 **Text Encoders）** - [qwen_2.5_vl_fp16.safetensors](https://huggingface.co/Comfy-Org/Omnigen2_ComfyUI_repackaged/blob/main/split_files/text_encoders/qwen_2.5_vl_fp16.safetensors) - 文件保存位置： ``` 📂 ComfyUI/ @@ -66,15 +63,24 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 ## ComfyUI OmniGen2 文生图工作流 -### 1. 工作流文件下载 +### OmniGen2: Text to Image (`image_omnigen2_t2i`) + +Generate high-quality images from text prompts using OmniGen2's unified 7B multimodal model with dual-path architecture. - +OmniGen2 text-to-image 工作流预览 + + - Open and run this workflow directly in Comfy Cloud. + 在 Comfy Cloud 中直接打开并运行此工作流 + + + Download JSON or search "OmniGen2" in Template Library -![文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_t2i.png) +**输出示例** + +![OmniGen2 text-to-image 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_omnigen2_t2i.png) ### 2. 按步骤完成工作流运行 @@ -96,18 +102,41 @@ OmniGen2 是一个强大且高效的统一多模态生成模型，总参数量 OmniGen2 有丰富的图像编辑能力，并且支持为图像添加文本 -### 1. 工作流文件下载 +### OmniGen2 Image Edit (`image_omnigen2_image_edit`) + +Edit images with natural language instructions using OmniGen2's advanced image editing capabilities and text rendering support. - +OmniGen2 image edit 工作流预览 + + - Open and run this workflow directly in Comfy Cloud. + 在 Comfy Cloud 中直接打开并运行此工作流 + + + Download JSON or search "OmniGen2 Image Edit" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 16 · `image_omnigen2_image_edit_input_image.png` -![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/image_omnigen2_image_edit.png) -下载下面的图片，我们将使用它作为输入图片。 -![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/image/omnigen2/input_fairy.png) +
+ image_omnigen2_image_edit_input_image.png +
+ +**输出示例** +
+ 输入图像 + OmniGen2 image edit 输出示例 +
### 2. 按步骤完成工作流运行 diff --git a/zh/tutorials/image/ovis/ovis-image.mdx b/zh/tutorials/image/ovis/ovis-image.mdx index 9d4d06823..5ffde4c19 100644 --- a/zh/tutorials/image/ovis/ovis-image.mdx +++ b/zh/tutorials/image/ovis/ovis-image.mdx @@ -20,19 +20,20 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [GitHub](https://github.com/AIDC-AI/Ovis-Image) - [Hugging Face](https://huggingface.co/AIDC-AI/Ovis-Image-7B) -## Ovis-Image 文生图工作流 +### Ovis-Image Text to Image (`image_ovis_text_to_image`) + +Ovis-Image is a 7B text-to-image model specifically optimized for high-quality text rendering in generated images. Designed to operate efficiently under stringent computational constraints. + +Ovis-Image text-to-image 工作流预览 - - + 在 Comfy Cloud 中打开 - Download JSON or search "Ovis image" in Template Library + Download JSON or search "Ovis image" in Template Library - - diff --git a/zh/tutorials/image/pixeldit/pixeldit.mdx b/zh/tutorials/image/pixeldit/pixeldit.mdx index cb3834a62..17ce2bda6 100644 --- a/zh/tutorials/image/pixeldit/pixeldit.mdx +++ b/zh/tutorials/image/pixeldit/pixeldit.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Model downloads": e4bafb0a --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **PixelDiT** 是 NVIDIA 开发的像素空间扩散变换器，用于 **1024px** 文本到图像生成。与传统在潜空间操作的扩散模型不同，PixelDiT 使用双层 DiT 架构直接在像素空间生成图像——结合了 patch 级 DiT 和像素级 DiT，并通过 MM-DiT 融合实现文本和图像 token 之间的联合注意力。 diff --git a/zh/tutorials/image/qwen/qwen-image-2512.mdx b/zh/tutorials/image/qwen/qwen-image-2512.mdx index 927c96793..2628e96ec 100644 --- a/zh/tutorials/image/qwen/qwen-image-2512.mdx +++ b/zh/tutorials/image/qwen/qwen-image-2512.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Qwen-Image-2512 ComfyUI Native Workflow Example": 50fbe75c --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Qwen-Image-2512** 是 Qwen-Image 文生图基础模型的 12 月更新版本。与 8 月发布的基础 Qwen-Image 模型相比，Qwen-Image-2512 在图像质量和真实感方面有显著提升。 diff --git a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx index 11739c3be..8197ff4b8 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit-2511.mdx @@ -48,7 +48,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [qwen_image_vae.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/blob/main/split_files/vae/qwen_image_vae.safetensors) - ### 2. 模型下载 **Text Encoders（文本编码器）** @@ -82,4 +81,3 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' │ └── qwen_image_vae.safetensors ``` - diff --git a/zh/tutorials/image/qwen/qwen-image-edit.mdx b/zh/tutorials/image/qwen/qwen-image-edit.mdx index 0955814ab..542892df1 100644 --- a/zh/tutorials/image/qwen/qwen-image-edit.mdx +++ b/zh/tutorials/image/qwen/qwen-image-edit.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Qwen-Image-Edit ComfyUI Native Workflow Example": 6703460f --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Qwen-Image-Edit** 是 Qwen-Image 的图像编辑版本。它基于20B的Qwen-Image模型进一步训练，成功将Qwen-Image的文本渲染特色能力拓展到编辑任务上，以支持精准的文字编辑。此外，Qwen-Image-Edit将输入图像同时输入到Qwen2.5-VL（获取视觉语义控制）和VAE Encoder（获得视觉外观控制），以同时获得语义/外观双重编辑能力。 @@ -43,23 +41,41 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' -### 1. 工作流文件 +### Qwen Image Edit (`image_qwen_image_edit`) + +Edit images with precise bilingual text editing and dual semantic/appearance editing capabilities using Qwen-Image-Edit's 20B MMDiT model. -更新 ComfyUI 后你可以从模板中找到工作流文件，或者将下面的工作流拖入 ComfyUI 中加载 -![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/qwen_image_edit.png) +Qwen-Image-Edit 工作流预览 + + 零配置在 Cloud GPU 上运行此工作流 + - Download JSON or search "image_qwen_image_edit" in Template Library + Download JSON or search "Qwen Image Edit" in Template Library - - Run this workflow on Cloud GPUs with zero setup + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 78 · `image_qwen_image_edit_input_image.png` +
+ image_qwen_image_edit_input_image.png +
+ +**输出示例** -下载下面的图片作为输入 -![Qwen-image 文生图工作流](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-edit/input.png) +
+ 输入图像 + Qwen-Image-Edit 输出示例 +
### 2. 模型下载 diff --git a/zh/tutorials/image/qwen/qwen-image-layered.mdx b/zh/tutorials/image/qwen/qwen-image-layered.mdx index ae263a941..c406c92a1 100644 --- a/zh/tutorials/image/qwen/qwen-image-layered.mdx +++ b/zh/tutorials/image/qwen/qwen-image-layered.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Workflow settings": 098636f1 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Qwen-Image-Layered** 是阿里巴巴通义千问团队开发的模型，能够将图像分解为多个 RGBA 图层。这种分层表示解锁了固有的可编辑性：每个图层都可以独立操作而不影响其他内容。 @@ -41,7 +39,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' | | - ## 模型下载 diff --git a/zh/tutorials/image/qwen/qwen-image.mdx b/zh/tutorials/image/qwen/qwen-image.mdx index 9e9c600e9..561286b61 100644 --- a/zh/tutorials/image/qwen/qwen-image.mdx +++ b/zh/tutorials/image/qwen/qwen-image.mdx @@ -13,11 +13,8 @@ translationBlockHashes: "Qwen Image Union ControlNet LoRA Workflow": a08d8e37 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **Qwen-Image** 是阿里巴巴通义千问团队发布的首个图像生成基础模型，这是一个拥有 20B 参数的 MMDiT（多模态扩散变换器）模型，基于 Apache 2.0 许可证开源。该模型在**复杂文本渲染**和**精确图像编辑**方面取得了显著进展，无论是英语还是中文等多种语言都能实现高保真输出。 **模型亮点**： @@ -46,7 +43,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' allowFullScreen > - **Qwen-Image ControlNet in ComfyUI - DiffSynth** +### Qwen-Image: Text to Image (`image_qwen_image`) -## Qwen-Image 原生工作流示例 +Generate images with exceptional multilingual text rendering and editing capabilities using Qwen-Image's 20B MMDiT model. - +Qwen-Image text-to-image 工作流预览 -| - + + 在 Comfy Cloud 中打开 - + + Download JSON or search "Qwen-Image" in Template Library - -在本篇文档所附工作流中使用的不同模型有三种 -1. Qwen-Image 原版模型 fp8_e4m3fn -2. 8步加速版： Qwen-Image 原版模型 fp8_e4m3fn 使用 lightx2v 8步 LoRA, -3. 蒸馏版:Qwen-Image 蒸馏版模型 fp8_e4m3fn +**输出示例** -**显存使用参考** +![Qwen-Image 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_qwen_image.png) + +There are three different models used in the workflow attached to this document: +1. Qwen-Image original model fp8_e4m3fn +2. 8-step accelerated version: Qwen-Image original model fp8_e4m3fn with lightx2v 8-step LoRA +3. Distilled version: Qwen-Image distilled model fp8_e4m3fn + +**VRAM Usage Reference** GPU: RTX4090D 24GB -| 使用模型 | VRAM Usage | 首次生成 | 第二次生成 | -| --------------------------------- | ---------- | -------- | ---------- | -| fp8_e4m3fn | 86% | ≈ 94s | ≈ 71s | -| fp8_e4m3fn 使用 lightx2v 8步 LoRA | 86% | ≈ 55s | ≈ 34s | -| 蒸馏版 fp8_e4m3fn | 86% | ≈ 69s | ≈ 36s | +| Model Used | VRAM Usage | First Generation | Second Generation | +| --------------------------------------- | ---------- | --------------- | ---------------- | +| fp8_e4m3fn | 86% | ≈ 94s | ≈ 71s | +| fp8_e4m3fn with lightx2v 8-step LoRA | 86% | ≈ 55s | ≈ 34s | +| Distilled fp8_e4m3fn | 86% | ≈ 69s | ≈ 36s | ### 1. 工作流文件 @@ -100,7 +101,6 @@ GPU: RTX4090D 24GB - Qwen-Image_fp8 (20.4 GB) - 蒸馏版本 (非官方，仅需 15 步) - 所有模型均可在 [Huggingface](https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main) 或者 [魔搭](https://modelscope.cn/models/Comfy-Org/Qwen-Image_ComfyUI/files) 找到 **Diffusion model** @@ -162,18 +162,36 @@ Qwen_image_distill 蒸馏版模型和 lightx2v 的 8 步加速 LoRA 似乎并不兼容，你可以测试具体的组合参数来验证组合使用的方式是否可行 -## Qwen Image InstantX ControlNet 工作流 +### Qwen-Image InstantX Union ControlNet (`image_qwen_image_instantx_controlnet`) + +Generate images with Qwen-Image InstantX ControlNet, supporting canny, soft edge, depth, and pose. -这是一个 ControlNet 模型 +Qwen-Image InstantX ControlNet 工作流预览 + 在 Comfy Cloud 中打开 - + + Download JSON or search "Qwen-Image InstantX ControlNet" in Template Library -### 1. 工作流及输入图片 +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 71 · `image_qwen_image_instantx_controlnet_input_image.jpg` + + + +
+ image_qwen_image_instantx_controlnet_input_image.jpg +
+ +**输出示例. 工作流及输入图片 下载下面的图片并拖入 ComfyUI 以加载工作流 ![workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/qwen/qwen-image-instantx-controlnet/image_qwen_image_instantx_controlnet.png) @@ -223,20 +241,34 @@ ComfyUI/ 3. 这里是一个子图，这里是 ComfyUI 支持的 lotus Depth 模型，你可以在模板中找到 Lotus Depth 或者编辑对应子图了解对应工作流，请确保所有模型都正确加载 4. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 -## Qwen Image ControlNet DiffSynth-ControlNets Model Patches 工作流 +### Qwen-Image ControlNet Model Patch (`image_qwen_image_controlnet_patch`) + +Control image generation using Qwen-Image ControlNet models. Supports canny, depth, and inpainting controls through model patching. + +Qwen-Image ControlNet model patch 工作流预览 + 在 Comfy Cloud 中打开 - + + Download JSON or search "Qwen-Image ControlNet Patch" in Template Library -这个模型实际上并不是一个 controlnet，而是一个 Model patch， 支持 canny、depth、inpaint 三种不同的控制模式 +**输入素材** + +Upload this file to the matching `LoadImage` node: -原始模型地址：[DiffSynth-Studio/Qwen-Image ControlNet](https://www.modelscope.cn/collections/Qwen-Image-ControlNet-6157b44e89d444) -Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/tree/main/split_files/model_patches) + + + `LoadImage` node 71 · `image_qwen_image_controlnet_patch_input_image.png` + + +
+ image_qwen_image_controlnet_patch_input_image.png +
### 1. 工作流及输入图片 @@ -255,7 +287,6 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http - [qwen_image_depth_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_depth_diffsynth_controlnet.safetensors) - [qwen_image_inpaint_diffsynth_controlnet.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/model_patches/qwen_image_inpaint_diffsynth_controlnet.safetensors) - ### 3. 工作流使用说明 目前 diffsynth 有三个 patch 的模型： Canny、Detph、Inpaint 三个模型 @@ -293,17 +324,44 @@ Comfy Org rehost 地址： [Qwen-Image-DiffSynth-ControlNets/model_patches](http 5. 如需要可以修改 `QwenImageDiffsynthControlnet` 节点的 `strength` 强度来控制对应的控制强度 6. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来运行工作流 -## Qwen Image union ControlNet LoRA 工作流 +### Qwen-Image Union Control (`image_qwen_image_union_control_lora`) + +Generate images with precise structural control using Qwen-Image's unified ControlNet LoRA. Supports multiple control types including canny, depth, lineart, softedge, normal, and openpose. + +Qwen-Image Union Control 工作流预览 + 在 Comfy Cloud 中打开 - + + Download JSON or search "Qwen-Image Union Control" in Template Library -原始模型地址：[DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) -Comfy Org reshot 地址: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): 图像结构控制lora 支持 canny、depth、post、lineart、softedge、normal、openpose +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 73 · `image_qwen_image_union_control_lora_input_image.png` + + + +
+ image_qwen_image_union_control_lora_input_image.png +
+ +**输出示例** + +
+ 输入图像 + Qwen-Image Union Control 输出示例 +
+ +Original model address: [DiffSynth-Studio/Qwen-Image-In-Context-Control-Union](https://www.modelscope.cn/models/DiffSynth-Studio/Qwen-Image-In-Context-Control-Union/) +Comfy Org rehost address: [qwen_image_union_diffsynth_lora.safetensors](https://huggingface.co/Comfy-Org/Qwen-Image-DiffSynth-ControlNets/blob/main/split_files/loras/qwen_image_union_diffsynth_lora.safetensors): Image structure control LoRA supporting canny, depth, pose, lineart, softedge, normal, openpose ### 1. 工作流及输入图片 diff --git a/zh/tutorials/image/z-image/z-image-turbo.mdx b/zh/tutorials/image/z-image/z-image-turbo.mdx index c6cd9c1d5..4a65aea6c 100644 --- a/zh/tutorials/image/z-image/z-image-turbo.mdx +++ b/zh/tutorials/image/z-image/z-image-turbo.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Z-Image-Turbo Fun Union ControlNet workflow": 34191a15 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -33,17 +31,28 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [GitHub](https://github.com/Tongyi-MAI/Z-Image) - [Hugging Face](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo) + + ## Z-Image-Turbo 文生图工作流 +### Z-Image-Turbo: Text to Image (`image_z_image_turbo`) + +An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer, supports English & Chinese. + +Z-Image-Turbo text-to-image 工作流预览 + - - 下载 Z-Image-Turbo 文生图工作流 JSON 文件。 + + 在 Comfy Cloud 中直接运行此工作流 - - 在 ComfyUI Cloud 上直接运行此工作流。 + + Download the Z-Image-Turbo text-to-image workflow JSON file - + +**输出示例** + +![Z-Image-Turbo 输出示例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_z_image_turbo.png) ### Z-Image-Turbo 模型下载 @@ -73,27 +82,32 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' ## Z-Image-Turbo Fun Union ControlNet 工作流 -此工作流使用 Z-Image-Turbo Fun Union ControlNet 模型生成带有 ControlNet 引导的图像。它对参考图像应用 Canny 边缘检测，并使用 ControlNet 来引导生成过程。 +### Z-Image-Turbo Fun Union ControlNet (`image_z_image_turbo_fun_union_controlnet`) + +Multi-control ControlNet supporting Canny, HED, Depth, Pose, and MLSD for Z-Image-Turbo. - - 下载 Z-Image-Turbo Fun Union ControlNet 工作流 JSON 文件。 +Z-Image-Turbo Fun Union ControlNet 工作流预览 + + + + 在 Comfy Cloud 中直接运行此工作流 + + + Download the Z-Image-Turbo Fun Union ControlNet workflow JSON file -### ControlNet 所需的额外模型 + + +**输入素材** - - Z-Image-Turbo ControlNet 模型补丁。 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 58 · `image_z_image_turbo_fun_union_controlnet_input_image.png` -**模型存储位置** + + +
+ image_z_image_turbo_fun_union_controlnet_input_image.png +
-``` -📂 ComfyUI/ -├── 📂 models/ -│ ├── 📂 text_encoders/ -│ │ └── qwen_3_4b.safetensors -│ ├── 📂 diffusion_models/ -│ │ └── z_image_turbo_bf16.safetensors -│ ├── 📂 vae/ -│ │ └── ae.safetensors -│ └── 📂 model_patches/ -│ └── Z-Image-Turbo-Fun-Controlnet-Union.safetensors -``` diff --git a/zh/tutorials/llm/gemma4/gemma4.mdx b/zh/tutorials/llm/gemma4/gemma4.mdx index c953ec9f4..94afb270e 100644 --- a/zh/tutorials/llm/gemma4/gemma4.mdx +++ b/zh/tutorials/llm/gemma4/gemma4.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Model Download": 9f2919ea --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -35,58 +34,79 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - [Google AI for Developers](https://ai.google.dev/gemma) - [ComfyUI 源码 (nodes_textgen.py)](https://github.com/Comfy-Org/ComfyUI/blob/master/comfy_extras/nodes_textgen.py) -## 可用工作流 +## Available workflow -### Gemma 4：文本生成 +### Gemma4: Text Generation (`llm_gemma4_text_gen`) - - 下载 JSON 或在模板库中搜索 "Gemma 4 Text Generation" - +Input your text prompt and optionally an image, audio, or video. Generate text output with configurable reasoning, coding, and multilingual support. + +Gemma 4 text generation workflow preview - - 在 Comfy Cloud 中打开 + + + Open in Comfy Cloud + + Download JSON or search "Gemma4: Text Generation" in Template Library + + + +**输入素材** -![Gemma 4 文本生成工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_gemma4_text_gen-1.webp) +Upload these optional files to the matching nodes: + + + + `LoadImage` node 2 · `the_lily_veil.png` + + + `LoadAudio` node 5 · `voice_demo.mp3` + + + `LoadVideo` node 6 · `video_wan_vace_inpainting_input_video.mp4` + + -该工作流展示了 Gemma 4 的核心**文本生成**能力。它可以在文本提示词之外，额外接受可选的图像、音频或视频作为上下文输入，并生成自然语言输出——支持推理、编程和多语言提示。 +This workflow demonstrates the core **text generation** capabilities of Gemma 4. It accepts an optional image, audio file, or video as additional context alongside your text prompt, and generates natural language output — with support for reasoning, coding, and multilingual prompts. -**输入**： -- **文本提示词** — 你的问题或指令 -- **图像**（可选）— 用于视觉理解任务（OCR、目标检测、图表阅读等） -- **音频**（可选）— 用于语音识别或转录 -- **视频**（可选）— 用于视频帧理解（内部按 1 FPS 采样） +**Inputs**: +- **Text prompt** — your question or instruction +- **Image** (optional) — for visual understanding tasks (OCR, object detection, chart reading, etc.) +- **Audio** (optional) — for speech recognition or transcription +- **Video** (optional) — for video understanding across frames (subsampled to 1 FPS internally) -**关键控制参数**： -- **Max length** — 生成的最大 token 数（默认 256） -- **Sampling mode** — 开关采样，调节 temperature、top-k、top-p、重复惩罚和随机种子 -- **Thinking mode** — 启用逐步推理，在最终答案前展示思考过程 -- **Use default template** — 使用模型内置的系统提示词模板 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Thinking mode** — enable step-by-step reasoning before the final answer +- **Use default template** — apply the built-in system prompt for the model -**输出**： -- **Generated text** — 模型生成的文本响应 +**Output**: +- **Generated text** — the model's response as a plain text string - - 该工作流使用 Subgraph 节点实现模块化处理。查看 Subgraph 文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 模型下载 +## Model Download -Gemma 4 模型在 ComfyUI 中以文本编码器（text encoder）形式加载。下载对应的模型文件并放入正确的目录： +Gemma 4 models are loaded as text encoders in ComfyUI. Download the relevant model file and place it in the correct directory: - - 快速轻量，推荐消费级 GPU 使用。 + + + Fast, lightweight. Recommended for consumer GPUs. - - 性能均衡，工作流默认使用此模型。 + + Balanced performance. The default model in the workflow. - - 浏览所有 Gemma 4 模型权重。 + + Browse all Gemma 4 model weights. + -将下载的 `.safetensors` 文件放入以下目录： +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ diff --git a/zh/tutorials/llm/qwen/qwen3.mdx b/zh/tutorials/llm/qwen/qwen3.mdx index 8e0bfa5bc..89ec2ee67 100644 --- a/zh/tutorials/llm/qwen/qwen3.mdx +++ b/zh/tutorials/llm/qwen/qwen3.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Model Download": d1d0cfd2 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -25,7 +24,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **ComfyUI 原生支持** — 使用内置 `TextGenerate` 节点，无需自定义节点 - **轻量级** — 与 Qwen3.5 共用相同的文本编码器格式，提供 2B/4B/9B 三种版本适应不同硬件 - ## 使用场景 Qwen 3.0 非常适合在 ComfyUI 工作流中需要结构化文本生成和智能推理的任务： @@ -35,59 +33,64 @@ Qwen 3.0 非常适合在 ComfyUI 工作流中需要结构化文本生成和智 - **文本分析** — 从文本输入中提取信息、分类内容或生成结构化报告。 - **链式推理** — 启用思考模式处理复杂多步任务，在生成最终输出前进行中间推理。 -## 可用工作流 +## Available workflow -### Qwen 3.0：文本生成 +### Qwen3.0: Text Generation (`llm_qwen3_text_gen`) - - 下载 JSON 或在模板库中搜索 "Qwen 3.0 Text Generation" - +Input a text prompt to generate detailed, reasoned responses using the Qwen3-4B-Thinking model. - - 在 Comfy Cloud 中打开 - +Qwen 3.0 text generation workflow preview -![Qwen 3.0 文本生成工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_text_gen-1.webp) + + + Open in Comfy Cloud + + + Download JSON or search "Qwen3.0: Text Generation" in Template Library + + -本工作流展示了 Qwen 3.0 的核心**文本生成**能力。它接收文本提示词，并利用模型内置的推理能力生成详细的结构化回复。 +This workflow demonstrates the core **text generation** capabilities of Qwen 3.0. It accepts a text prompt and generates detailed, structured responses using the model's built-in reasoning capabilities. -**输入**： -- **文本提示词** — 你的问题、指令或任务描述 +**Inputs**: +- **Text prompt** — your question, instruction, or task description -**关键控制**： -- **Max length** — 生成的最大 token 数量（默认 256） -- **Sampling mode** — 开关采样，并调整 temperature、top-k、top-p、重复惩罚和随机种子 -- **Thinking mode** — 在最终答案前启用逐步推理 -- **Use default template** — 使用模型内置系统提示词模板 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Thinking mode** — enable step-by-step reasoning before the final answer +- **Use default template** — apply the built-in system prompt for the model -**输出**： -- **Generated text** — 模型的纯文本回复 +**Output**: +- **Generated text** — the model's response as a plain text string - - 本工作流使用了 Subgraph 节点进行模块化处理。查看 Subgraph 文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 模型下载 +## Model Download -Qwen 3.0 模型以文本编码器的形式加载到 ComfyUI 中，模型文件与 Qwen3.5 共用。根据你的硬件选择合适的版本： +Qwen 3.0 models are loaded as text encoders in ComfyUI. The model files are shared with Qwen3.5 — download the variant that best fits your hardware: - - 轻量版，约 4.5 GB。适合低显存环境和快速下载。 + + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - - 大小和质量均衡。推荐大多数消费级 GPU。 + + Balanced size and quality. Recommended for most consumer GPUs. - - 最大版本，约 19 GB。输出质量更高，需要更多显存。 + + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + -将下载的 `.safetensors` 文件放入： +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ -│ └── qwen3.5_4b_bf16.safetensors # 或 2b / 9b 版本 +│ └── qwen3.5_4b_bf16.safetensors # or 2b / 9b variant ``` diff --git a/zh/tutorials/llm/qwen/qwen3_5.mdx b/zh/tutorials/llm/qwen/qwen3_5.mdx index 1e14b68d0..44927f0b9 100644 --- a/zh/tutorials/llm/qwen/qwen3_5.mdx +++ b/zh/tutorials/llm/qwen/qwen3_5.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Model Download": 84a7c321 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -26,7 +25,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - **ComfyUI 原生支持** — 使用内置 `TextGenerate` 节点，无需自定义节点 - **轻量级** — 4B 参数模型，适合消费级 GPU - ## 使用场景 Qwen3.5 在需要将视觉理解与文本生成结合的 ComfyUI 场景中表现出色： @@ -37,59 +35,78 @@ Qwen3.5 在需要将视觉理解与文本生成结合的 ComfyUI 场景中表现 - **视觉问答** — 询问关于图像内容的问题（"这张图里有什么物体？"、"背景是什么颜色？"），获得结构化的文本答案。 - **文字读取** — 使用合适的 prompt，模型可能会尝试读取图片中的可见文字或标签，但可靠性取决于文字渲染的质量和清晰度。 -## 可用工作流 +## Available workflow + +### Qwen3.5: Text Generation (`llm_qwen3_5_text_gen`) + +Use the Qwen3.5 model to analyze an input image and generate descriptive text prompts. This workflow performs image captioning and reverse prompt engineering. -### Qwen3.5: 文本生成 +Qwen3.5 text generation workflow preview - - 下载 JSON 或在模板库中搜索 "Qwen3.5 Text Generation" + + + Open in Comfy Cloud + + Download JSON or search "Qwen3.5: Text Generation" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadImage` node: - - 在 Comfy Cloud 中打开 + + + `LoadImage` node 2 · `man_with_red_hat.png` + -![Qwen3.5 文本生成工作流](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/llm_qwen3_5_text_gen-1.webp) +
+ Input image +
-此工作流展示了 Qwen3.5 的**文本生成和图像理解**能力。接受文本提示词和可选图像输入，根据输入生成描述性文本或结构化分析。 +This workflow demonstrates the **text generation and image understanding** capabilities of Qwen3.5. It accepts a text prompt and an optional image, and generates descriptive text or structured analysis based on the input. -**输入**： -- **文本提示词** — 你的问题、指令或任务描述 -- **图像**（可选） — 用于视觉理解任务（图像描述、反向提示词工程、提示词优化等） +**Inputs**: +- **Text prompt** — your question, instruction, or task description +- **Image** (optional) — for visual understanding tasks (image captioning, reverse prompt engineering, prompt optimization, etc.) -**关键控制参数**： -- **最大长度** — 生成的最大 token 数（默认 256） -- **采样模式** — 开关采样，调节温度、top-k、top-p、重复惩罚和随机种子 -- **使用默认模板** — 应用模型内置系统提示词 +**Key controls**: +- **Max length** — maximum number of tokens to generate (default 256) +- **Sampling mode** — toggle sampling on/off and adjust temperature, top-k, top-p, repetition penalty, and seed +- **Use default template** — apply the built-in system prompt for the model -**输出**： -- **生成的文本** — 模型回复的纯文本字符串 +**Output**: +- **Generated text** — the model's response as a plain text string - - 本工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 模型下载 +## Model Download -Qwen3.5 模型以文本编码器的形式加载到 ComfyUI 中。根据你的硬件选择合适的版本： +Qwen3.5 models are loaded as text encoders in ComfyUI. Choose the variant that best suits your hardware: - - 轻量版，约 4.5 GB。适合低显存环境和快速下载。 + + + Lightweight, ~4.5 GB. Best for low VRAM setups and fast downloads. - - 大小和质量均衡。推荐大多数消费级 GPU。 + + Balanced size and quality. Recommended for most consumer GPUs. - - 最大版本，约 19 GB。输出质量更高，需要更多显存。 + + Largest variant, ~19 GB. Higher quality output, requires more VRAM. + -将下载的 `.safetensors` 文件放入： +Place the downloaded `.safetensors` file in: ``` 📂 ComfyUI/ ├── 📂 models/ │ └── 📂 text_encoders/ -│ └── qwen3.5_4b_bf16.safetensors # 或 2b / 9b 版本 +│ └── qwen3.5_4b_bf16.safetensors # or 2b / 9b variant ``` diff --git a/zh/tutorials/partner-nodes/anthropic/claude.mdx b/zh/tutorials/partner-nodes/anthropic/claude.mdx index 9844f4354..fe98b6a66 100644 --- a/zh/tutorials/partner-nodes/anthropic/claude.mdx +++ b/zh/tutorials/partner-nodes/anthropic/claude.mdx @@ -32,8 +32,6 @@ Anthropic Claude 是 Anthropic 推出的强大 AI 模型系列，以其出色的 Anthropic Claude Chat workflow preview -Anthropic Claude Chat workflow preview - diff --git a/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx b/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx index ba858cab5..f9074576f 100644 --- a/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx +++ b/zh/tutorials/partner-nodes/beeble/beeble-switchx.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Beeble SwitchX: Video Edit": 166c2e59 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx b/zh/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx index 3fe054421..919242355 100644 --- a/zh/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx +++ b/zh/tutorials/partner-nodes/black-forest-labs/flux-1-1-pro-ultra-image.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Flux 1.1[pro] Image-to-Image Tutorial": 6b3cfbea --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx b/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx index f61a321c8..c31f98ab3 100644 --- a/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx +++ b/zh/tutorials/partner-nodes/black-forest-labs/flux-1-kontext.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "Flux.1 Kontext Max Image Partner Nodes Workflow": df399d73 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import PromptTechniques from "/snippets/zh/tutorials/flux/prompt-techniques.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -108,5 +107,4 @@ Kontext 的核心优势在于其出色的上下文理解能力和角色一致性 3. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行图像的编辑。 4. 等待 API 返回结果后，你可在 `Save Image` 节点中查看编辑后的图像，对应的图像也会被保存至 `ComfyUI/output/` 目录下。 - diff --git a/zh/tutorials/partner-nodes/bria/background-removal.mdx b/zh/tutorials/partner-nodes/bria/background-removal.mdx index ba2e260b0..4b94a7c10 100644 --- a/zh/tutorials/partner-nodes/bria/background-removal.mdx +++ b/zh/tutorials/partner-nodes/bria/background-removal.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Video Background Processing": aab02df5 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/bria/fibo.mdx b/zh/tutorials/partner-nodes/bria/fibo.mdx index 0ca77f623..60e3f4215 100644 --- a/zh/tutorials/partner-nodes/bria/fibo.mdx +++ b/zh/tutorials/partner-nodes/bria/fibo.mdx @@ -52,4 +52,3 @@ FIBO Edit 擅长各种编辑任务： - **文字调整**：调整图像中的文字 - **艺术风格转换**：以不同的艺术风格重新想象您的图像 - diff --git a/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx b/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx index 9f4f45751..107fa8b9d 100644 --- a/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seed-audio-1-0.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 1ae5cf45 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx b/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx index 45d623c7f..deca1d21d 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedance-2-0-real-human.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Available workflows": 16803a18 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -102,14 +100,13 @@ ComfyUI 提供两套预置的 Seedance 2.0 真人支持模板。两者都会将 Seedance 2.0 Real Human R2V workflow preview -Seedance 2.0 Real Human R2V workflow preview - Try the Seedance 2.0 Real Human Reference-to-Video workflow instantly on Comfy Cloud. - + 获取 Seedance 2.0 真人支持参考图生成视频工作流文件。 +
@@ -130,7 +127,8 @@ Download this sample input image to try the workflow: Try the Seedance 2.0 Real Human First-Last-Frame-to-Video workflow instantly on Comfy Cloud. - + 获取 Seedance 2.0 真人支持首尾帧生成视频工作流文件。 +
diff --git a/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx b/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx index f06b467b5..c7255e525 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedance-2-0.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Using real-person and AI-generated portraits in ComfyUI for Seedance 2.0": d9966fae --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx b/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx index eceaf0e62..92339232d 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedream-5-lite.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Get started": d14874f4 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -68,8 +66,6 @@ Seedream 5.0 lite 是 BytePlus 最新的图像生成模型。它是 Seedream 系 Seedream 5.0 Lite Text-to-Image workflow preview -Seedream 5.0 Lite Text-to-Image workflow preview - diff --git a/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx b/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx index a6dd9b74f..8af271bf8 100644 --- a/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx +++ b/zh/tutorials/partner-nodes/bytedance/seedream-5-pro.mdx @@ -48,7 +48,6 @@ Seedream 5.0 Pro 是字节跳动推出的专业级图像生成模型，作为 Se Seedream 5.0 Pro Image Edit workflow preview -Seedream 5.0 Pro Image Edit workflow preview 在 Comfy Cloud 中打开 diff --git a/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx b/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx index d88d8442f..d5b1f9b03 100644 --- a/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx +++ b/zh/tutorials/partner-nodes/google/gemini-omni-flash.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Get started": 64517938 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/google/nano-banana-2-lite.mdx b/zh/tutorials/partner-nodes/google/nano-banana-2-lite.mdx index 795af68ee..5d8f9834d 100644 --- a/zh/tutorials/partner-nodes/google/nano-banana-2-lite.mdx +++ b/zh/tutorials/partner-nodes/google/nano-banana-2-lite.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 5a827068 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/google/nano-banana-2.mdx b/zh/tutorials/partner-nodes/google/nano-banana-2.mdx index 02696c3fd..a4ebdc17b 100644 --- a/zh/tutorials/partner-nodes/google/nano-banana-2.mdx +++ b/zh/tutorials/partner-nodes/google/nano-banana-2.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Get started": f6189d9e --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/google/nano-banana-pro.mdx b/zh/tutorials/partner-nodes/google/nano-banana-pro.mdx index b8237d03f..29d0466fd 100644 --- a/zh/tutorials/partner-nodes/google/nano-banana-pro.mdx +++ b/zh/tutorials/partner-nodes/google/nano-banana-pro.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Get started": 38fdf056 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx index e5d9c9455..38fc26427 100644 --- a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx +++ b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-0.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "HappyHorse 1.0 video edit": 5b531078 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx index 4b44c86c3..98f466f27 100644 --- a/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx +++ b/zh/tutorials/partner-nodes/happyhorse/happyhorse1-1.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Getting started": 27bbb438 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx b/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx index 92dcdbb8c..865405ce3 100644 --- a/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx +++ b/zh/tutorials/partner-nodes/hunyuan3d/hunyuan3d-3-0.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Advanced features": 6b37a964 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx b/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx index de78da338..f54c814b0 100644 --- a/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx +++ b/zh/tutorials/partner-nodes/hunyuan3d/model-generation.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Multi-view-to-3D workflow": ff251560 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx b/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx index 228395c3d..44b07ee7e 100644 --- a/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx +++ b/zh/tutorials/partner-nodes/ideogram/ideogram-v4.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx index 1fbb77a57..7c5b4ddfc 100644 --- a/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx +++ b/zh/tutorials/partner-nodes/kling/kling-motion-control.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Tips for better results": b6fdceb5 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -52,34 +50,33 @@ Kling 2.6 Motion Control 是由快手开发的专门多模态模型，能够实 ## Kling 2.6 Motion Control 工作流 - - Kling 2.6 Motion Control workflow preview - Kling 2.6 Motion Control workflow preview - - Run the Kling 2.6 Motion Control workflow on Comfy Cloud. + + + 在 Comfy Cloud 中打开 - - - Download the workflow JSON file for local use. + + 下载 JSON，或在模板库中搜索「Kling2.6: Motion Control」 + -
-Input materials +**输入素材** -Download these sample input files to try the workflow: +将以下文件上传到对应的 `LoadImage` 和 `LoadVideo` 节点： - - Download sample reference image + + `LoadImage` 节点 11 · `streetwear_fox.jpg` - - Download sample motion reference video + + `LoadVideo` 节点 2 · `street_dancer.mp4` -
- + +
+ streetwear_fox.jpg +
## 输入要求 diff --git a/zh/tutorials/partner-nodes/krea2/krea2-t2i.mdx b/zh/tutorials/partner-nodes/krea2/krea2-t2i.mdx index 99214cba7..ee55a68b0 100644 --- a/zh/tutorials/partner-nodes/krea2/krea2-t2i.mdx +++ b/zh/tutorials/partner-nodes/krea2/krea2-t2i.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Additional Notes": 7099b5ae --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx b/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx index 40db73bd1..52b97490e 100644 --- a/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx +++ b/zh/tutorials/partner-nodes/luma/luma-uni-1.mdx @@ -22,8 +22,6 @@ translationBlockHashes: "Key takeaway": 77d51b1d --- - - 在 **ComfyUI** 里，Luma **Uni-1** 以 **合作伙伴 API 节点**形式接入：**Create** 路线根据提示词生成新图（可挂多张参考图）；**Modify** 路线以已有图像为输入做定向编辑。操作上与普通工作流相同——用 **加载图像** / **保存图像** 与 Luma 节点串联，在节点上填写提示词、种子、宽高比与参考图槽位，再在本地排队运行或通过 **Comfy Cloud** 打开下方模板。 Luma 将 Uni-1 表述为非扩散、自回归类的模型，会在成图前对提示词做推理；在画布侧更需要关注的是 **Create / Modify** 选型、参考图角色写清楚，以及用种子做可控迭代。 @@ -77,8 +75,6 @@ Uni-1 在广泛的任务中表现优异： Luma Uni-1 Image Create workflow preview -Luma Uni-1 Image Create workflow preview - diff --git a/zh/tutorials/partner-nodes/meshy/meshy-6.mdx b/zh/tutorials/partner-nodes/meshy/meshy-6.mdx index 5bdd265b8..d7c459b30 100644 --- a/zh/tutorials/partner-nodes/meshy/meshy-6.mdx +++ b/zh/tutorials/partner-nodes/meshy/meshy-6.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Multi-view to Model Workflow": 0cf6bb73 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx b/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx index 7d82f0d0e..0fa08ea14 100644 --- a/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx +++ b/zh/tutorials/partner-nodes/moonvalley/moonvalley-video-generation.mdx @@ -12,12 +12,10 @@ translationBlockHashes: "Moonvalley Video-to-Video Workflow": 5de07f68 --- - **服务不可用**：Moonvalley API 服务已不再提供，相关节点已弃用，可能无法正常使用。 - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -58,7 +56,6 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成

下载 Json 格式工作流文件

- ### 2. 按步骤完成工作流的运行 ![文本生视频作流](/images/tutorial/api_nodes/moonvalley/api_moonvalley_text_to_video.jpg) @@ -69,7 +66,6 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 4. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行视频的生成 5. 等待 API 返回结果后，你可在 `Save Video` 节点中查看生成的视频，对应的视频也会被保存至 `ComfyUI/output/` 目录下 - ## Moonvalley 图生视频工作流 ### 1. 工作流文件下载 @@ -99,7 +95,6 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 5. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行视频的生成 6. 等待 API 返回结果后，你可在 `Save Video` 节点中查看生成的视频，对应的视频也会被保存至 `ComfyUI/output/` 目录下 - ## Moonvalley 视频转视频工作流 `Moonvalley Marey Video to Video` 节点将允许你输入一段参考视频来进行视频的重绘，你可以参考视频画面动态或者角色姿态动作来进行视频的绘制生成。 @@ -124,7 +119,6 @@ Moonvalley Marey Realism v1.5 是专为影视级创作打造的 AI 视频生成 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/moonvalley/api_moonvalley_video_to_video_input.mp4" > - ### 2. 按步骤完成工作流的运行 ![视频转绘工作流](/images/tutorial/api_nodes/moonvalley/api_moonvalley_video_to_video.jpg) diff --git a/zh/tutorials/partner-nodes/openai/dall-e-2.mdx b/zh/tutorials/partner-nodes/openai/dall-e-2.mdx index 9822d2418..9f9674b6f 100644 --- a/zh/tutorials/partner-nodes/openai/dall-e-2.mdx +++ b/zh/tutorials/partner-nodes/openai/dall-e-2.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/zh/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -106,7 +105,6 @@ DALL·E 2 支持图像编辑功能，允许您使用蒙版指定要替换的区 6. 编辑 `prompt` 节点的提示词 7. 运行工作流 - **注意事项** - 如果您想使用图像编辑功能，必须同时提供图像和蒙版（缺一不可） diff --git a/zh/tutorials/partner-nodes/openai/dall-e-3.mdx b/zh/tutorials/partner-nodes/openai/dall-e-3.mdx index c1c6c920b..160da3788 100644 --- a/zh/tutorials/partner-nodes/openai/dall-e-3.mdx +++ b/zh/tutorials/partner-nodes/openai/dall-e-3.mdx @@ -12,14 +12,12 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/zh/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; OpenAI DALL·E 3 是 ComfyUI 合作伙伴节点系列中的一员，它允许用户通过 OpenAI 的 **DALL·E 3** 模型生成图像。此节点支持文本到图像的生成功能。 - ![OpenAI DALL·E 2 节点截图](/images/comfy_core/api_nodes/openai-dall-e-3.jpg) ## 节点概述 @@ -56,7 +54,6 @@ DALL·E 3 是 OpenAI 的最新图像生成模型，能够根据文本提示创 ![ComfyUI openai-dall-e-3 工作流](/images/tutorial/api_nodes/openai/openai-dall-e-3/text2image.jpg) - 1. 在 ComfyUI 中添加 **OpenAI DALL·E 3** 节点 2. 在提示文本框中输入您想要生成的图像描述 3. 根据需要调整可选参数（质量、风格、尺寸等） diff --git a/zh/tutorials/partner-nodes/openai/gpt-image-1.mdx b/zh/tutorials/partner-nodes/openai/gpt-image-1.mdx index 13104b376..207f3b172 100644 --- a/zh/tutorials/partner-nodes/openai/gpt-image-1.mdx +++ b/zh/tutorials/partner-nodes/openai/gpt-image-1.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Faq from "/snippets/zh/tutorials/partner-nodes/faq.mdx"; import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -70,7 +69,6 @@ OpenAI GPT-Image-1 是 ComfyUI 合作伙伴节点系列中的一员，它允许 你只需要加载 `OpenAI GPT-Image-1` 节点，在 `prompt` 节点中输入你想要生成的图像的描述，连接一个 `保存图像（Save Image）` 节点，然后运行工作流即可。 - ### 图生图（Image to Image）示例 下面的图片包含了一个简单的图生图工作流，请下载对应的图像，并拖入 ComfyUI 以加载对应的工作流 @@ -79,7 +77,6 @@ OpenAI GPT-Image-1 是 ComfyUI 合作伙伴节点系列中的一员，它允许 我们将使用下面的图片作为输入： ![ComfyUI openai-gpt-image-1 工作流 input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/api_nodes/GPT-Image-1/input.webp) - 这个工作流中，我们使用 `OpenAI GPT-Image-1` 节点生成图像，并使用 `加载图像（Load Image）` 节点加载输入的图像，然后连接到 `OpenAI GPT-Image-1` 节点的 `image` 输入中。 ![ComfyUI openai-gpt-image-1 工作流示例](/images/tutorial/api_nodes/openai/gpt-image-1/image2image.jpg) diff --git a/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx b/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx index 33a654fb1..8539b10e7 100644 --- a/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx +++ b/zh/tutorials/partner-nodes/openai/gpt-image-2.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Hybrid Pipelines": 9c6479f0 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/openrouter/llm.mdx b/zh/tutorials/partner-nodes/openrouter/llm.mdx index a49f6a177..ac42226f9 100644 --- a/zh/tutorials/partner-nodes/openrouter/llm.mdx +++ b/zh/tutorials/partner-nodes/openrouter/llm.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Example workflow (`api_openrouter_llm`)": f1dbfe6a --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/overview.mdx b/zh/tutorials/partner-nodes/overview.mdx index 790110de9..371fb1923 100644 --- a/zh/tutorials/partner-nodes/overview.mdx +++ b/zh/tutorials/partner-nodes/overview.mdx @@ -17,7 +17,6 @@ translationBlockHashes: "FAQs": 2b8e1938 --- - import Requirements from "/snippets/zh/tutorials/partner-nodes/requirements.mdx"; import Faq from "/snippets/zh/tutorials/partner-nodes/faq.mdx"; @@ -39,7 +38,6 @@ Partner Nodes 是一组特殊的节点，它们能够连接到外部 API 服务 ![Select Comfy API Key Login](/images/interface/setting/user/user-login-api-1.jpg) - ## 使用 ComfyUI 账户 API Key 集成来调用付费合作节点 目前我们支持通过 ComfyUI 账户 API Key 集成来访问我们的服务来调用付费合作节点，请参考 API Key 集成章节了解如何使用 API Key 集成来调用付费合作节点。 diff --git a/zh/tutorials/partner-nodes/pricing.mdx b/zh/tutorials/partner-nodes/pricing.mdx index c7586e088..0a9e4fde2 100644 --- a/zh/tutorials/partner-nodes/pricing.mdx +++ b/zh/tutorials/partner-nodes/pricing.mdx @@ -45,7 +45,6 @@ translationBlockHashes: "Cloud GPU": 103c55e6 --- - 下表列出了当前合作伙伴节点的定价。所有价格均以积分计。 ## Anthropic diff --git a/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx b/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx index dfbe0da9a..c6f80d6c5 100644 --- a/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx +++ b/zh/tutorials/partner-nodes/recraft/recraft-v4.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Additional notes": d5f3f109 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/reve/reve-image.mdx b/zh/tutorials/partner-nodes/reve/reve-image.mdx index c0f9cae8f..33e11648a 100644 --- a/zh/tutorials/partner-nodes/reve/reve-image.mdx +++ b/zh/tutorials/partner-nodes/reve/reve-image.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Available nodes": 5927bf70 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/rodin/model-generation.mdx b/zh/tutorials/partner-nodes/rodin/model-generation.mdx index fe41bfdcc..ec29316f5 100644 --- a/zh/tutorials/partner-nodes/rodin/model-generation.mdx +++ b/zh/tutorials/partner-nodes/rodin/model-generation.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Other Related Nodes": c885ce31 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/runway/image-generation.mdx b/zh/tutorials/partner-nodes/runway/image-generation.mdx index 3cf646552..3a917ff3e 100644 --- a/zh/tutorials/partner-nodes/runway/image-generation.mdx +++ b/zh/tutorials/partner-nodes/runway/image-generation.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Runway Image Reference-to-Image Workflow": cc84e030 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/runway/video-generation.mdx b/zh/tutorials/partner-nodes/runway/video-generation.mdx index 7000f87c8..7ff0a7a54 100644 --- a/zh/tutorials/partner-nodes/runway/video-generation.mdx +++ b/zh/tutorials/partner-nodes/runway/video-generation.mdx @@ -11,7 +11,6 @@ translationBlockHashes: "First-Last Frame Video Generation Workflow": b7f1102b --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -88,7 +87,6 @@ Runway 是一家专注于生成式 AI 的科技公司，提供强大的视频生 ## 首尾帧视频生成工作流 - ### 1. 工作流文件下载 下面的视频的`metadata`中已经包含工作流信息，请下载并拖入 ComfyUI 中加载对应工作流。 diff --git a/zh/tutorials/partner-nodes/sonilo/video-to-music.mdx b/zh/tutorials/partner-nodes/sonilo/video-to-music.mdx index f32e0d0b1..204b25320 100644 --- a/zh/tutorials/partner-nodes/sonilo/video-to-music.mdx +++ b/zh/tutorials/partner-nodes/sonilo/video-to-music.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Use cases": 90983ea2 --- - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; diff --git a/zh/tutorials/partner-nodes/tripo/model-generation.mdx b/zh/tutorials/partner-nodes/tripo/model-generation.mdx index 6be569bf0..673b76990 100644 --- a/zh/tutorials/partner-nodes/tripo/model-generation.mdx +++ b/zh/tutorials/partner-nodes/tripo/model-generation.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Subsequent Task Processing for the Same Task": d24582e4 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -73,7 +71,6 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 - 下载下面的图片作为输入图片 ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/image_to_model/panda.jpg) @@ -90,7 +87,6 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 3. 点击 `Run` 按钮，或者使用快捷键 `Ctrl(cmd) + Enter(回车)` 来执行模型的生成，工作流完成后对应的模型会自动保存至 `ComfyUI/output/` 目录下 4. 模型下载请参考文生图部分的说明 - ## 多视图模型生成工作流 ### 1. 工作流文件下载 @@ -103,7 +99,6 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 - 下载下面的图片作为输入图片 ![前视图](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/api_nodes/tripo/multiview_to_image/front.jpg) @@ -123,7 +118,6 @@ Tripo AI 是一家专注于生成式 AI 3D 建模的公司，它提供用户友 4. 其它视图输入可以参考步骤图中的示意将对应节点的模式设置为 `总是（always）` 来启用 5. 模型下载请参考文生图部分的说明 - ## 对应任务的后续任务处理 Tripo 的对应节点提供了对于同一任务的后续处理，只需要在相关节点中输入对应的`model_task_id` 即可,我们在相关模板中也已提供了对应的节点，你也可以按需通过修改对应节点模式来启用 diff --git a/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx b/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx index e864acb72..4af7b2feb 100644 --- a/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx +++ b/zh/tutorials/partner-nodes/tripo/tripo-3-1.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Available Workflows": 5ec1d9c5 --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -103,4 +101,3 @@ Generate a high-detail 3D model from multiple view images using Tripo 3.1. | 材质输出 | PBR 就绪 | 标准材质贴图 | | 优化支持 | 支持 | 支持（Refine Draft 仅限 v1.4） | - diff --git a/zh/tutorials/partner-nodes/wan/wan2-7.mdx b/zh/tutorials/partner-nodes/wan/wan2-7.mdx index 97f48298b..48fabf4e4 100644 --- a/zh/tutorials/partner-nodes/wan/wan2-7.mdx +++ b/zh/tutorials/partner-nodes/wan/wan2-7.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Wan2.7 video edit": 0d2511bb --- - - import ReqHint from "/snippets/zh/tutorials/partner-nodes/req-hint.mdx"; import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; @@ -47,72 +45,56 @@ Wan2.7 是阿里巴巴最新的视频生成模型，现已通过合作伙伴节 Wan2.7 I2V workflow preview -Wan2.7 I2V workflow preview - - - - - - - -Wan2.7 T2V workflow preview - - - - - - - -Wan2.7 R2V workflow preview - - - - - - - -Wan2.7 Video Edit workflow preview - - - - - + + 获取 Wan2.7 图生视频工作流文件。 + + + 在 Comfy Cloud 中立即试用图生视频工作流。 + ## Wan2.7 文生视频 从纯文本提示词生成视频。可选加入音频输入和多镜头叙事，实现更丰富的故事讲述。 +Wan2.7 T2V workflow preview + - + 获取 Wan2.7 文生视频工作流文件。 - - 在 Comfy Cloud 上即刻体验文生视频工作流。 + + 在 Comfy Cloud 中立即试用文生视频工作流。 + ## Wan2.7 参考生成视频 使用人物视觉外观的参考图像，并可选配声线参考。支持最多 5 个真人输入，实现多角色交互场景。 +Wan2.7 R2V workflow preview + - + 获取 Wan2.7 参考生成视频工作流文件。 - - 在 Comfy Cloud 上即刻体验参考生成视频工作流。 + + 在 Comfy Cloud 中立即试用参考生成视频工作流。 + ## Wan2.7 视频编辑 使用文本提示词、参考图像或风格迁移来编辑或复制现有视频。 +Wan2.7 Video Edit workflow preview + - + 获取 Wan2.7 视频编辑工作流文件。 - - 在 Comfy Cloud 上即刻体验视频编辑工作流。 + + 在 Comfy Cloud 中立即试用视频编辑工作流。 \ No newline at end of file diff --git a/zh/tutorials/utility/depth-anything-3.mdx b/zh/tutorials/utility/depth-anything-3.mdx index 26666a4db..b56601c10 100644 --- a/zh/tutorials/utility/depth-anything-3.mdx +++ b/zh/tutorials/utility/depth-anything-3.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Community Resources": 8c5d05ab --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' # ComfyUI Depth Anything 3 简介 @@ -54,80 +53,103 @@ ComfyUI/ │ │ └── depth_anything_3_metric_large.safetensors ``` -## 示例工作流 +## Example Workflows ---- +### Depth Anything 3: Image Depth Estimation (`utility_depth_anything3_image_depth_estimation`) -## 1. 图像深度估计 +Upload one image and generate a depth map using Depth Anything 3. View a side-by-side comparison of the original image and depth output. -**功能说明：** 上传一张图像，使用 **Image Depth Estimation (Depth Anything 3)** 生成深度图。结果在 **Depth Preview** 中显示，提供原始图像与深度输出的并排对比视图。 +Depth Anything 3 image depth estimation workflow preview + + Open in Comfy Cloud + - 下载 JSON 或在模板库中搜索 "Depth Anything 3" + Download JSON or search "Depth Anything 3: Image Depth Estimation" in Template Library - - 获取此工作流的示例输入图片 + + +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 85 · `retro_futuristic_home.png` -
- 图像深度估计输出 - 图像深度估计对比 +
+ Input image
-### 运行步骤 +### Steps to Run -1. **LoadImage** — 加载输入图像 -2. **LoadDA3Model** — 选择 Depth Anything 3 变体 -3. **运行** — 点击 Queue 或使用 `Cmd+Enter` -4. 工作流输出深度图和并排比较结果 +1. **LoadImage** — load your input image +2. **LoadDA3Model** — select a Depth Anything 3 variant +3. **Run** — click Queue or use `Cmd+Enter` +4. The workflow outputs a depth map and side-by-side comparison - - 此工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. ---- +### Depth Anything 3: Video Depth Estimation (`utility_depth_anything3_video_depth_estimation`) -## 2. 视频深度估计 +Upload a video to generate a per-frame depth sequence. Inside the subgraph, **GetVideoComponents** splits the input video into frames, **LoadDA3Model** loads the model, and **SetVideoComponents** reassembles the depth frames back into a video output. -**功能说明：** 上传一个视频，运行 **Video Depth Estimation (Depth Anything 3)** 生成逐帧深度序列。在子图内部，**GetVideoComponents** 将输入视频拆分为帧，**LoadDA3Model** 加载模型，**SetVideoComponents** 将深度帧重新组合为视频输出。 +Depth Anything 3 video depth estimation workflow preview + + Open in Comfy Cloud + - 下载 JSON 或在模板库中搜索 "Depth Anything 3" + Download JSON or search "Depth Anything 3: Video Depth Estimation" in Template Library - - 在 Comfy Cloud 中打开 + + +**输入素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 87 · `empty_room_assembly.mp4` -![视频深度估计预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_depth_anything3_video_depth_estimation-1.webp) +
+ +
-### 运行步骤 +### Steps to Run -1. **LoadVideo** — 加载输入视频 -2. **选择模型** — 在 **Small**、**Base**、**Mono-Large** 或 **Metric-Large** 中选择 -3. **运行** — 点击 Queue 或使用 `Cmd+Enter` -4. 工作流输出逐帧深度图视频 +1. **LoadVideo** — load your input video +2. **Select Model** — choose between **Small**, **Base**, **Mono-Large**, or **Metric-Large** +3. **Run** — click Queue or use `Cmd+Enter` +4. The workflow outputs a video with per-frame depth maps - - 此工作流使用子图节点进行模块化处理。查看子图文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 模型变体 +## Model Variants -| 变体 | head_type | 天空检测 | 置信度 | 相机解码 | 最佳用途 | -|------|-----------|:-------:|:------:|:--------:|----------| -| **Small** | dualdpt | ❌ | ✅ | ✅ | 快速推理、移动/边缘设备 | -| **Base** | dualdpt | ❌ | ✅ | ✅ | 均衡性能 | -| **Mono-Large** | dpt | ✅ | ❌ | ❌ | 带天空检测的单目深度 | -| **Metric-Large** | dpt | ✅ | ❌ | ❌ | 物理度量深度（米级输出） | +| Variant | head_type | has_sky | has_confidence | camera_decoder | Best for | +|---------|-----------|:-------:|:--------------:|:--------------:|----------| +| **Small** | dualdpt | ❌ | ✅ | ✅ | Fast inference, mobile/edge | +| **Base** | dualdpt | ❌ | ✅ | ✅ | Balanced performance | +| **Mono-Large** | dpt | ✅ | ❌ | ❌ | Monocular depth with sky detection | +| **Metric-Large** | dpt | ✅ | ❌ | ❌ | Physical metric depth in metres | -- **Small** 和 **Base** 使用 `dualdpt` 头类型，支持置信度估计和相机解码器，适用于多视角应用。 -- **Mono-Large** 和 **Metric-Large** 使用 `dpt` 头类型，支持天空检测。Metric-Large 输出原始米级深度。 +- **Small** and **Base** use the `dualdpt` head type with confidence estimation and camera decoder support for multi-view applications. +- **Mono-Large** and **Metric-Large** use the `dpt` head type with sky detection. Metric-Large outputs raw depth in metres. -## 社区资源 +## Community Resources -- [Depth Anything 3 GitHub (ByteDance-Seed)](https://github.com/ByteDance-Seed/Depth-Anything-3) — 研究论文和代码 -- [Comfy-Org/Depth-Anything-3](https://huggingface.co/Comfy-Org/Depth-Anything-3) — 官方 ComfyUI 模型权重 +- [Depth Anything 3 GitHub (ByteDance-Seed)](https://github.com/ByteDance-Seed/Depth-Anything-3) — Research paper and code +- [Comfy-Org/Depth-Anything-3](https://huggingface.co/Comfy-Org/Depth-Anything-3) — Official ComfyUI model weights diff --git a/zh/tutorials/utility/face-detection/mediapipe.mdx b/zh/tutorials/utility/face-detection/mediapipe.mdx index 7a7289a5d..6fea6abca 100644 --- a/zh/tutorials/utility/face-detection/mediapipe.mdx +++ b/zh/tutorials/utility/face-detection/mediapipe.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Community Resources": ec7aa0f5 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -31,33 +30,48 @@ MediaPipe Face Detection 已原生集成到 ComfyUI（PR [#14009](https://github > **范围说明：** 仅人脸检测 — BlazeFace + FaceMesh v2 + ARKit blendshape。不包括手部、姿态或整体检测。 -## MediaPipe Face Detection 工作流 +## MediaPipe Face Detection Workflow -### 1. 下载工作流 +### Mediapipe: Image Face Detection (`utility_face_detection_mediapipe`) -将 ComfyUI 更新到最新版本，然后进入 `Workflow` → `Browse Templates`，在 Utility 分类下找到 "Mediapipe: Image Face Detection"。 +Input an image and detect up to 6 facial landmarks per face, enabling ultrafast multi-face detection. - - 下载工作流 - +Mediapipe image face detection workflow preview - - 在云端打开 + + + Open in Comfy Cloud + + Download JSON or search "Mediapipe: Image Face Detection" in Template Library + + + +**输入素材** - - 获取此工作流的示例输入图片 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 5 · `soft_neon_girl.png` + + +
+ Input image +
+ +### 1. Download the Workflow -![MediaPipe Face Detection 预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/utility_face_detection_mediapipe-1.webp) +Update your ComfyUI to the latest version, then go to `Workflow` → `Browse Templates` and find "Mediapipe: Image Face Detection" under the Utility category. -### 2. 下载模型 +### 2. Download the Model -MediaPipe Face Detection 模型托管在 [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe) 上。 +The MediaPipe Face Detection model is hosted on the [Comfy-Org MediaPipe model repository](https://huggingface.co/Comfy-Org/mediapipe). -- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/resolve/main/detection/mediapipe_face_fp32.safetensors) +- [mediapipe_face_fp32.safetensors](https://huggingface.co/Comfy-Org/mediapipe/blob/main/detection/mediapipe_face_fp32.safetensors) -将其放置在以下目录结构中： +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -66,54 +80,54 @@ MediaPipe Face Detection 模型托管在 [Comfy-Org MediaPipe model repository]( └── mediapipe_face_fp32.safetensors ``` -### 3. 使用工作流 +### 3. Using the Workflow -本工作流使用一个 **subgraph**（子图）节点来协调人脸检测、可视化和遮罩生成。子图暴露了以下控制参数： +This workflow uses a **subgraph** node that orchestrates face detection, visualization, and mask generation. The subgraph exposes the following controls: - - 本工作流使用 Subgraph 节点进行模块化处理。查看 Subgraph 文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -**子图输入：** +**Subgraph inputs:** -| 输入 | 描述 | -|------|------| -| **image** | 要分析的输入图片批次 | -| **face_landmarker** | 可选。留空则使用内置模型加载器。连接外部 `FACE_DETECTION_MODEL` 输出可覆盖 | +| Input | Description | +|-------|-------------| +| **image** | Input image batch to analyze | +| **face_landmarker** | Optional. Leave empty to use the built-in model loader. Connect an external `FACE_DETECTION_MODEL` output to override | -**子图参数：** +**Subgraph parameters:** -| 参数 | 默认值 | 描述 | -|------|:------:|------| -| **model_name** | `mediapipe_face_fp32.safetensors` | `ComfyUI/models/detection/` 目录下的模型文件。如缺失则下载上述模型 | -| **detector_variant** | `short` | **short** — 针对近距离/大脸优化（约 2 米范围）。**full** — 覆盖更小/更远的人脸（约 5 米），速度较慢。**both** — 同时运行两个检测器，保留每帧检测到更多人脸的结果（约 2 倍开销） | -| **num_faces** | `1` | 每帧最多返回的人脸数。`0` = 不限制（返回所有检测结果）。范围：0–16 | -| **custom_face_oval** | `false` | 在遮罩输出中包含面部轮廓区域 | -| **custom_lips** | `false` | 在遮罩中包含嘴唇（与其他启用的区域合并） | -| **custom_left_eye** | `false` | 在遮罩中包含左眼区域 | -| **custom_right_eye** | `false` | 在遮罩中包含右眼区域 | -| **custom_irises** | `false` | 在遮罩中包含虹膜区域 | +| Parameter | Default | Description | +|-----------|:-------:|-------------| +| **model_name** | `mediapipe_face_fp32.safetensors` | Checkpoint in `ComfyUI/models/detection/`. If missing, download the model above | +| **detector_variant** | `short` | **short** — tuned for close/large faces (~2 m range). **full** — covers smaller/farther faces (~5 m), slower. **both** — runs both detectors and keeps whichever found more faces per frame (~2× cost) | +| **num_faces** | `1` | Maximum faces to return per frame. `0` = no cap (return all detected). Range: 0–16 | +| **custom_face_oval** | `false` | Include face-outline region in the mask output | +| **custom_lips** | `false` | Include lips in the mask (union with other enabled regions) | +| **custom_left_eye** | `false` | Include left eye region in the mask | +| **custom_right_eye** | `false` | Include right eye region in the mask | +| **custom_irises** | `false` | Include iris regions in the mask | -遮罩切换在内部使用自定义模式：仅填充勾选的区域；多个开启的区域会合并为每帧一个遮罩。 +Mask toggles use custom mode internally: only checked regions are filled; multiple ON regions are **unioned** into one mask per frame. -**子图输出：** +**Subgraph outputs:** -| 输出 | 类型 | 描述 | -|------|------|------| -| **face_landmarks** | `FACE_LANDMARKS` | 每帧人脸数据，包含 478 个 2D/3D 关键点、ARKit-52 blendshape、网格拓扑数据——输入到可视化和遮罩节点 | -| **bboxes** | `BOUNDING_BOX` | 人脸边界框——兼容 `DrawBBoxes` 节点 | -| **mask** | `MASK` | 基于启用的区域切换生成的二值遮罩 | +| Output | Type | Description | +|--------|------|-------------| +| **face_landmarks** | `FACE_LANDMARKS` | Per-frame faces with 478 2D/3D landmarks, ARKit-52 blendshapes, mesh topology data — feeds into visualization and mask nodes | +| **bboxes** | `BOUNDING_BOX` | Face bounding boxes — compatible with `DrawBBoxes` node | +| **mask** | `MASK` | Binary mask from the enabled region toggles | -### 4. 运行工作流 +### 4. Run the Workflow -1. 确保模型文件已放置在 `ComfyUI/models/detection/` 目录 -2. 在 `Load Image` 节点加载一张图片 -3. 根据需要调整检测参数 -4. 点击 `Queue` 或使用 `Ctrl(Cmd) + Enter` 运行 -5. 工作流输出网格覆盖图、边界框和遮罩预览 +1. Ensure the model file is placed in `ComfyUI/models/detection/` +2. Load an image in the `Load Image` node +3. Adjust detection parameters as needed +4. Click `Queue` or use `Ctrl(Cmd) + Enter` to run +5. The workflow outputs the mesh overlay, bounding boxes, and mask preview -## 社区资源 +## Community Resources -- [MediaPipe GitHub](https://github.com/google-ai-edge/mediapipe) — MediaPipe 上游框架 -- [Comfy-Org/mediapipe](https://huggingface.co/Comfy-Org/mediapipe) — 官方 ComfyUI 模型权重 -- [ComfyUI Subgraph 指南](https://docs.comfy.org/zh/interface/features/subgraph) — 了解子图的工作原理 +- [MediaPipe GitHub](https://github.com/google-ai-edge/mediapipe) — Upstream MediaPipe framework +- [Comfy-Org/mediapipe](https://huggingface.co/Comfy-Org/mediapipe) — Official ComfyUI model weights +- [ComfyUI Subgraph Guide](https://docs.comfy.org/interface/features/subgraph) — Learn how subgraphs work diff --git a/zh/tutorials/utility/frame-interpolation.mdx b/zh/tutorials/utility/frame-interpolation.mdx index 5c251b66d..7d102e946 100644 --- a/zh/tutorials/utility/frame-interpolation.mdx +++ b/zh/tutorials/utility/frame-interpolation.mdx @@ -38,4 +38,3 @@ translationFrom: tutorials/utility/frame-interpolation.mdx 下载工作流文件 - diff --git a/zh/tutorials/utility/image-upscale.mdx b/zh/tutorials/utility/image-upscale.mdx index 53f410749..3a2df686a 100644 --- a/zh/tutorials/utility/image-upscale.mdx +++ b/zh/tutorials/utility/image-upscale.mdx @@ -14,7 +14,6 @@ translationBlockHashes: "Tips": a966ce44 --- - 本指南介绍 ComfyUI 中的图像放大工作流，包括本地模型和合作伙伴节点选项，适用于各种使用场景。 diff --git a/zh/tutorials/utility/moge.mdx b/zh/tutorials/utility/moge.mdx index b185e0489..76f9772cf 100644 --- a/zh/tutorials/utility/moge.mdx +++ b/zh/tutorials/utility/moge.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Community Resources": d9fcd8cd --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' # ComfyUI MoGe 介绍 @@ -62,75 +60,118 @@ ComfyUI/ │ │ └── moge_1_vitl_fp16.safetensors ``` -## 工作流示例 - ---- +## Example Workflows -## 1. 深度估计 +### MoGe: Depth Estimation (`utility_moge_depth_estimation`) -**功能：** 输入一张图片，输出公尺度深度图、彩色深度预览和遮罩——即 MoGe 一次推理得到的公尺度深度结果。适合作为场景深度参考用于合成、深度特效，或作为生成网格的前置步骤。 +Upload a single RGB image and generate a colored depth preview and raw depth map. -MoGe 还会自动估计图片的相机 FOV，也可以手动输入真实 FOV 以获得更精确的结果。 +MoGe depth estimation workflow preview - - 下载 JSON 或在模板库中搜索 "MoGe Depth Estimation" + + Open in Comfy Cloud - - 获取本工作流所用的示例输入图片 + + Download JSON or search "MoGe: Depth Estimation" in Template Library -### 1.1 运行步骤 -1. 确保 `LoadMoGeModel` 节点已加载 MoGe 检查点 -2. 在 `Load Image` 节点中加载一张图片 -3. 点击 `Queue` 按钮，或使用快捷键 `Ctrl(cmd) + Enter` 运行 -4. 工作流将输出彩色深度预览、原始深度图和遮罩 +**输入素材** ---- +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 9 · `alien_world.png` + + + +### 1.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded a MoGe checkpoint +2. Load an image in the `Load Image` node +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run +4. The workflow outputs colored depth preview, raw depth preview, and a mask + +### MoGe: Perspective Geometry Estimation (`3d_moge_perspective_to_mesh`) -## 2. 透视照片转 3D 网格 +Upload an image to estimate its perspective geometry. Generate a 3D depth map and surface normals from the input, then convert to a textured GLB mesh. -**功能：** 将单张透视照片转换为带纹理的 GLB 网格，同时生成法线和深度预览。MoGe 从可见场景中估计点云、深度和法线，再转换为网格。这是**单目几何估计**——遮挡区域和物体背面会缺失或出现碎片。适合场景快速原型、参考几何体，或将深度/法线可视化展示为网格，不能替代多视角 3D 重建。 +MoGe perspective to mesh workflow preview - - 下载 JSON 或在模板库中搜索 "3D MoGe Perspective to Mesh" + + Open in Comfy Cloud - - 获取本工作流所用的示例输入图片 + + Download JSON or search "MoGe: Perspective Geometry Estimation" in Template Library -### 2.1 运行步骤 -1. 确保 `LoadMoGeModel` 节点已加载 MoGe 检查点 -2. 在 `Load Image` 节点中加载一张透视照片 -3. （可选）查看 OpenGL 和 DirectX 法线预览 -4. 点击 `Queue` 或使用 `Ctrl(cmd) + Enter` 运行 +**输入素材** ---- +Upload this file to the matching `LoadImage` node: -## 3. 全景图转 3D 网格 + + + `LoadImage` node 9 · `modern_living_room.png` + + -**功能：** 将 360° 等距柱状全景图转换为带纹理的 GLB 网格。该工作流使用 `MoGePanoramaInference` 将全景图分割为 12 个透视视角，分别独立进行单目几何估计后合并为单个网格。每个分段仍然是单视图估计，因此结果是粗略的场景重建——适合获得 360° 场景的空间概览，但遮挡区域和表面后的几何结构会缺失或碎片化。 +
+ Input image +
+ +This is **monocular geometry estimation**: occluded areas and object backsides will be missing or fragmented. Useful for quick scene prototyping, reference geometry, or visualizing depth and normals as a mesh, but not a replacement for multi-view 3D reconstruction. + +### 2.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded a MoGe checkpoint +2. Load a perspective photo in the `Load Image` node +3. (Optional) View the OpenGL and DirectX normal previews +4. Click `Queue` or use `Ctrl(cmd) + Enter` to run + +### Moge: Panorama to Mesh (`3d_moge_panorama_to_mesh`) + +Upload an equirectangular 360° panorama image and generate a textured GLB mesh with vertex colors. + +MoGe panorama to mesh workflow preview - - 下载 JSON 或在模板库中搜索 "3D MoGe Panorama to Mesh" + + Open in Comfy Cloud - - 获取本工作流所用的示例输入图片 + + Download JSON or search "Moge: Panorama to Mesh" in Template Library -### 3.1 运行步骤 -1. 确保 `LoadMoGeModel` 节点已加载 MoGe 检查点 -2. 在 `Load Image` 节点中加载一张等距柱状投影全景图 -3. 点击 `Queue` 按钮，或使用快捷键 `Ctrl(cmd) + Enter` 运行 +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 28 · `lego_street_panorama.png` + + + +
+ Input panorama +
+ +The workflow uses `MoGePanoramaInference` to split the panorama into 12 perspective views, run monocular geometry estimation on each view independently, then merge them into a single mesh. + +### 3.1 Steps to Run + +1. Ensure the `LoadMoGeModel` node has loaded one of the MoGe checkpoints +2. Load an equirectangular panorama image in the `Load Image` node +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run --- -## 社区资源 +## Community Resources -- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe) — 研究论文和代码 -- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe) — 官方 ComfyUI 模型权重 +- [MoGe GitHub (microsoft/MoGe)](https://github.com/microsoft/MoGe): Research paper and code +- [Comfy-Org/MoGe](https://huggingface.co/Comfy-Org/MoGe): Official ComfyUI model weights diff --git a/zh/tutorials/utility/pose-detection-sdpose.mdx b/zh/tutorials/utility/pose-detection-sdpose.mdx index 81f0c1d4d..1aefc3039 100644 --- a/zh/tutorials/utility/pose-detection-sdpose.mdx +++ b/zh/tutorials/utility/pose-detection-sdpose.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": 409c3058 --- - import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx" @@ -31,69 +30,161 @@ SDPose + RT-DETRv4 已在 ComfyUI 中原生支持（PR [#12748](https://github.c > **局限性：** 检测精度取决于图像分辨率和目标可见性。极度遮挡或非常小的目标可能产生较少的关键点。 -## SDPose 工作流 +## SDPose Workflows -根据你的使用场景，提供了四种工作流： +Four workflows are available depending on your use case: -| 工作流 | 输入 | 输出 | 应用场景 | +| Workflow | Input | Output | Use Case | |----------|-------|--------|----------| -| 多人（图像） | 单张图像 | 姿态图 + 边界框 | 多人照片 | -| 多人（视频） | 视频 | 逐帧姿态图 + 边界框 | 视频姿态追踪 | -| OOD 图像转姿态 | 单张图像 | 姿态图 | 风格迁移 / 图像转姿态 | -| OOD 视频转姿态图 | 视频 | 逐帧姿态图 | 视频转姿态动画 | +| Multi-Person (Image) | Single image | Pose map + BBoxes | Photos with multiple people | +| Multi-Person (Video) | Video | Per-frame pose map + BBoxes | Video pose tracking | +| OOD Image to Pose | Single image | Pose map | Style transfer / image-to-pose | +| OOD Video to Pose Map | Video | Per-frame pose map | Video-to-pose animation | + +### 1. Download Workflows + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find SDPose workflows under the Utility category. -### 1. 下载工作流 +### SDPose: Image Multi-Person Detection (`utility_sdpose_multi_person`) -将你的 ComfyUI 更新到最新版本，然后前往 `工作流` -> `浏览模板`，在“Utility”类别下找到 SDPose 工作流。 +Upload an image to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose image multi-person detection workflow preview - - 在 Comfy Cloud 中运行 + + Open in Comfy Cloud - - 下载 JSON + + Download JSON or search "SDPose: Image Multi-Person Detection" in Template Library +**输入素材** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 679 · `group_photo.png` + + + +
+ Input image +
+ +### SDPose: Video Multi-Person Detection (`utility_sdpose_multi_person_video`) + +Upload a video to detect human poses. Supports detection for both single individuals and multiple people within the same scene. + +SDPose video multi-person detection workflow preview + - - 在 Comfy Cloud 中运行 + + Open in Comfy Cloud - - 下载 JSON + + Download JSON or search "SDPose: Video Multi-Person Detection" in Template Library +**输入素材** + +Upload this file to the matching `LoadVideo` node: + - - 在 Comfy Cloud 中运行 + + `LoadVideo` node 694 · `man_playing_violin.mp4` - - 下载 JSON + + +
+ +
+ +### SDPose-OOD: Image to Pose Map (`utility_sdpose_ood_image_to_pose`) + +Upload an image to extract pose keypoints and generate a corresponding pose map using the SDPose-OOD model. + +SDPose-OOD image to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Image to Pose Map" in Template Library +**输入素材** + +Upload this file to the matching `LoadImage` node: + - - 在 Comfy Cloud 中运行 + + `LoadImage` node 667 · `dancer.png` - - 下载 JSON + + +**输出示例** + +
+ Input image + SDPose-OOD image to pose map example output +
+ +### SDPose-OOD: Video to Pose Map (`utility_sdpose_ood_video_to_pose_map`) + +Upload a video to extract pose keypoints and generate a pose map. The workflow supports multiple person detection and uses an enhanced SDPose model for accurate whole-body feature extraction. + +SDPose-OOD video to pose map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SDPose-OOD: Video to Pose Map" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 679 · `man_in_the_rain.mp4` -### 2. 下载模型 +**输出示例** + +
+ + +
+ +### 2. Download Models -SDPose 和 RT-DETRv4 模型文件托管在 [Comfy-Org SDPose 模型仓库](https://huggingface.co/Comfy-Org/SDPose) 中。 +The SDPose and RT-DETRv4 model checkpoints are hosted on the [Comfy-Org SDPose model repository](https://huggingface.co/Comfy-Org/SDPose). -**checkpoints**（SDPose 模型）： -- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors) +**checkpoints** (SDPose model): +- [sdpose_wholebody_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/checkpoints/sdpose_wholebody_fp16.safetensors) -**diffusion_models**（RT-DETRv4 检测器）： -- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors)（推荐） -- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors)（全精度，体积更大） +**diffusion_models** (RT-DETRv4 detector): +- [rt_detr_v4-x-hgnet_fp16.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors) (recommended) +- [rt_detr_v4-x-hgnet_fp32.safetensors](https://huggingface.co/Comfy-Org/SDPose/blob/main/diffusion_models/rt_detr_v4-x-hgnet_fp32.safetensors) (full precision, larger) -将模型放置在以下目录结构中： +Place them in the following directory structure: ``` 📂 ComfyUI/ @@ -105,44 +196,44 @@ SDPose 和 RT-DETRv4 模型文件托管在 [Comfy-Org SDPose 模型仓库](https └── rt_detr_v4-x-hgnet_fp32.safetensors ``` -### 3. 使用工作流 +### 3. Using the Workflows -#### 多人（图像） +#### Multi-Person (Image) -- **输入** — 通过 `加载图像` 节点加载一张图像。使用包含一人或多人的图像（示例：`group_photo.png`）。 -- **检测** — `Image to Pose Map (SDPose Multi-Person)` 子图处理图像并输出： - - **IMAGE** — 叠加在图像上的姿态骨架可视化结果 - - **keypoints** — 原始全身关键点数据 - - **bboxes** — 边界框坐标 -- **绘制选项** — 配置要绘制的身体部位： - - `draw_body`、`draw_hands`、`draw_face`、`draw_feet` — 切换可见性 - - `stick_width`、`face_point_size` — 调整视觉样式 - - `score_threshold` — 显示关键点的最低置信度 -- **检测选项**： - - `resize_type.longer_size` — 检测前对较长边进行缩放 - - `max_detections` — 最大检测人数 - - `detect_threshold` — 检测置信度阈值 - - `detect_class` — 要检测的对象类别（默认：person） +- **Input** — Load an image via the `Load Image` node. Use an image with one or more people (example: `group_photo.png`). +- **Detection** — The `Image to Pose Map (SDPose Multi-Person)` subgraph processes the image and outputs: + - **IMAGE** — pose skeleton visualization overlaid on the image + - **keypoints** — raw whole-body keypoint data + - **bboxes** — bounding box coordinates +- **Drawing Options** — Configure which body parts to draw: + - `draw_body`, `draw_hands`, `draw_face`, `draw_feet` — toggle visibility + - `stick_width`, `face_point_size` — adjust visual style + - `score_threshold` — minimum confidence for displaying keypoints +- **Detection Options**: + - `resize_type.longer_size` — scale the longer dimension before detection + - `max_detections` — maximum number of people to detect + - `detect_threshold` — detection confidence threshold + - `detect_class` — object class to detect (default: person) -#### 多人（视频） +#### Multi-Person (Video) -与图像工作流相同，但会顺序处理视频帧。使用 `加载视频` 输入视频文件，并使用 `保存视频` 导出结果。 +Same as the image workflow but processes video frames sequentially. Use `Load Video` to input a video file and `Save Video` to export the result. -#### OOD 图像转姿态 +#### OOD Image to Pose -利用 SDPose 模型从图像生成干净的人体姿态图，不包含边界框可视化。适用于风格迁移，即你想从一张图像中提取骨架姿态并应用到另一张图像上。 +Uses the SDPose model to generate a clean pose map from an image, without bounding box visualization. This is useful for style transfer where you want to extract the skeleton pose from one image and apply it to another. -#### OOD 视频转姿态图 +#### OOD Video to Pose Map -从视频生成逐帧姿态图。输出是一个视频文件，其中每一帧都包含提取的姿态骨架，适用于下游动画或 ControlNet 工作流。 +Generates per-frame pose maps from a video. The output is a video file where each frame contains the extracted pose skeleton, suitable for downstream animation or ControlNet workflows. - - 这些工作流使用子图节点进行模块化处理。请查阅子图文档，了解如何自定义和扩展工作流。 + + These workflows use Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflows. -## 附加说明 +## Additional Notes -- **模型目录** — SDPose 模型文件放在 `models/checkpoints/` 中，RT-DETRv4 检测器放在 `models/diffusion_models/` 中 -- **输入图像示例** — 工作流模板的 `input/` 目录中提供了 `group_photo.png` 文件以供测试 -- **关键点输出** — POSE_KEYPOINT 类型可以连接到接受姿态数据进行条件生成的下游节点 -- **需要更新** — 较新的 ComfyUI 版本才支持 SDPose + RT-DETRv4。请确保你的 ComfyUI 是最新版本。 +- **Model directory** — the SDPose checkpoint goes in `models/checkpoints/`, and the RT-DETRv4 detector goes in `models/diffusion_models/` +- **Input image example** — the `group_photo.png` file is available in the workflow template's `input/` directory for testing +- **Keypoint output** — the POSE_KEYPOINT type can be connected to downstream nodes that accept pose data for conditional generation +- **Update required** — SDPose + RT-DETRv4 support is available in recent ComfyUI versions. Make sure your ComfyUI is up to date. diff --git a/zh/tutorials/utility/preprocessors.mdx b/zh/tutorials/utility/preprocessors.mdx index 120210edb..0efec153f 100644 --- a/zh/tutorials/utility/preprocessors.mdx +++ b/zh/tutorials/utility/preprocessors.mdx @@ -12,7 +12,6 @@ translationBlockHashes: "Normals extraction": 65c791bb --- - ## 什么是预处理器？ @@ -27,84 +26,175 @@ translationBlockHashes: - 更容易调试和调优 - 更可预测的图像和视频结果 -## 深度估计 +## Depth estimation + +Depth estimation converts a flat image into a depth map representing relative distance within a scene. This structural signal is foundational for controlled generation, spatially aware edits, and relighting workflows. + +This workflow emphasizes: +- Clean, stable depth extraction +- Consistent normalization for downstream use +- Easy integration with ControlNet and image-edit pipelines + +Depth outputs can be reused across multiple passes, making it easier to iterate without re-running expensive upstream steps. + +### Video to Depth Map (`utility-depthAnything-v2-relative-video`) + +Convert a video to a temporally stable depth map. + +Video to Depth Map workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Depth Map" in Template Library + + + +**输入素材** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 6 · `roller_coaster.mp4` + + + +
+ +
+ +## Lineart conversion + +Lineart preprocessors distill an image down to its essential edges and contours, removing texture and color while preserving structure. + +This workflow is designed to: +- Produce clean, high-contrast lineart +- Minimize broken or noisy edges +- Provide reliable structural guidance for stylization and redraw workflows + +Lineart pairs especially well with depth and pose, offering strong structural constraints without overconstraining style. + +### Video to Lineart / Canny (`utility-lineart-video`) + +Convert a video to lineart or Canny edges for control processors. + +Video to Lineart workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "Video to Lineart / Canny" in Template Library + + + +**输入素材** + +Upload this file to the matching `VHS_LoadVideo` node: + + + + `VHS_LoadVideo` node 2 · `utility-lineart-video-input.mp4` + + + +
+ +
-深度估计将平面图像转换为表示场景中相对距离的深度图。这种结构信号是受控生成、空间感知编辑和重新打光工作流的基础。 +## Pose detection -此工作流强调： -- 干净、稳定的深度提取 -- 一致的归一化以供下游使用 -- 与 ControlNet 和图像编辑管道的轻松集成 +Pose detection extracts body keypoints and skeletal structure from images, enabling precise control over human posture and movement. -深度输出可以在多个处理过程中重复使用，使迭代更容易，无需重新运行昂贵的上游步骤。 +This workflow focuses on: +- Clear, readable pose outputs +- Stable keypoint detection suitable for reuse across frames +- Compatibility with pose-based ControlNet and animation pipelines - - 在 Comfy Cloud 上运行 - +By isolating pose extraction into a dedicated workflow, pose data becomes easier to inspect, refine, and reuse. - - 下载 JSON - +### Video to Pose Map - OpenPose (`utility-openpose-video`) -## 线稿转换 +Convert a video to a temporally stable pose control map. -线稿预处理器将图像提炼为其基本边缘和轮廓，去除纹理和颜色，同时保留结构。 +Video to Pose Map workflow preview -此工作流旨在： -- 生成干净、高对比度的线稿 -- 最小化断裂或噪声边缘 -- 为风格化和重绘工作流提供可靠的结构指导 + + + Open in Comfy Cloud + + + Download JSON or search "Video to Pose Map - OpenPose" in Template Library + + -线稿与深度和姿态配合特别好，提供强大的结构约束而不会过度约束风格。 +**输入素材** - - 在 Comfy Cloud 上运行 - +Upload this file to the matching `VHS_LoadVideo` node: - - 下载 JSON - + + + `VHS_LoadVideo` node 2 · `pose_input.mp4` + + -## 姿态检测 +
+ +
-姿态检测从图像中提取身体关键点和骨骼结构，实现对人体姿势和动作的精确控制。 +## Normals extraction -此工作流专注于： -- 清晰、可读的姿态输出 -- 适合跨帧重用的稳定关键点检测 -- 与基于姿态的 ControlNet 和动画管道的兼容性 +Normals estimation converts a flat image into a surface normal map—a per-pixel direction field that describes how each part of a surface is oriented (typically encoded as RGB). This signal is useful for relighting, material-aware stylization, and highly structured edits. -通过将姿态提取隔离到专用工作流中，姿态数据变得更容易检查、优化和重用。 +This workflow emphasizes: +- Clean, stable normal extraction with minimal speckling +- Consistent orientation and normalization for reliable downstream use +- ControlNet-ready outputs for relighting, refinement, and structure-preserving edits +- Reuse across passes so you can iterate without re-running earlier steps - - 在 Comfy Cloud 上运行 - +Normal outputs can be used to: +- Drive relight/shading changes while preserving geometry +- Add a stronger 3D-like structure to stylization and redraw pipelines +- Improve consistency across frames when paired with pose/depth for animation work - - 下载 JSON - +### Video to Normal Map (`utility-normal_crafter-video`) -## 法线提取 +Convert a video to a temporally stable normal map. -法线估计将平面图像转换为表面法线图——一个描述表面每个部分朝向的逐像素方向场（通常编码为 RGB）。这种信号对于重新打光、材质感知风格化和高度结构化的编辑非常有用。 +Video to Normal Map workflow preview -此工作流强调： -- 干净、稳定的法线提取，最小化斑点 -- 一致的方向和归一化以供可靠的下游使用 -- ControlNet 就绪的输出，用于重新打光、优化和保持结构的编辑 -- 跨处理过程重用，无需重新运行早期步骤即可迭代 + + + Open in Comfy Cloud + + + Download JSON or search "Video to Normal Map" in Template Library + + -法线输出可用于： -- 在保持几何形状的同时驱动重新打光/着色变化 -- 为风格化和重绘管道添加更强的 3D 结构 -- 与姿态/深度配合用于动画工作时提高跨帧一致性 +**输入素材** - - 在 Comfy Cloud 上运行 - +Upload this file to the matching `VHS_LoadVideo` node: - - 下载 JSON - + + + `VHS_LoadVideo` node 3 · `normals_input.mp4` + + +
+ +
diff --git a/zh/tutorials/utility/remove-background-birefnet.mdx b/zh/tutorials/utility/remove-background-birefnet.mdx index 9a248d3d2..9363af0e4 100644 --- a/zh/tutorials/utility/remove-background-birefnet.mdx +++ b/zh/tutorials/utility/remove-background-birefnet.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": d07d5c84 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -30,27 +29,44 @@ BiRefNet 在 ComfyUI 中获得原生支持（PR [#12747](https://github.com/Comf > **限制：** 背景极其杂乱或主体与背景融合的情况下，遮罩精度可能下降。模型每次处理一张图像。 -## BiRefNet 背景移除工作流 +## BiRefNet Background Removal Workflow + +### 1. Download Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "BiRefNet: Remove Background" under the Utility category. + +### BiRefNet: Remove Background (`utility_birefnet_remove_background`) -### 1. 工作流文件下载 +Upload an image with any background. Generate a version with the background removed and a precision segmentation mask. -请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` 找到 Utility 类别下的 "BiRefNet: Remove Background"。 +BiRefNet remove background workflow preview - - 下载工作流 + + + Open in Comfy Cloud + + Download JSON or search "BiRefNet: Remove Background" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadImage` node: - - Open in cloud + + + `LoadImage` node 17 · `the_lily_veil.png` + -### 2. 手动下载模型 +### 2. Download Models -BiRefNet 模型托管在 [Comfy-Org BiRefNet 模型仓库](https://huggingface.co/Comfy-Org/BiRefNet)。 +The BiRefNet model is hosted on the [Comfy-Org BiRefNet model repository](https://huggingface.co/Comfy-Org/BiRefNet). -- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/resolve/main/background_removal/birefnet.safetensors) +- [birefnet.safetensors](https://huggingface.co/Comfy-Org/BiRefNet/blob/main/background_removal/birefnet.safetensors) -放置到以下目录结构： +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -59,21 +75,21 @@ BiRefNet 模型托管在 [Comfy-Org BiRefNet 模型仓库](https://huggingface.c └── birefnet.safetensors ``` -### 3. 使用工作流 +### 3. Using the Workflow -- **图像** — 通过 `Load Image` 节点加载图像（放入 ComfyUI 的 `input/` 文件夹） -- `Remove Background (BiRefNet)` 子图会处理图像并输出： - - **IMAGE** — 背景透明的 RGBA 结果 - - **mask** — 提取的前景遮罩 +- **Image** — Load an image via the `Load Image` node (place it in the ComfyUI `input/` folder) +- The `Remove Background (BiRefNet)` subgraph processes the image and outputs: + - **IMAGE** — the result with a transparent background (RGBA) + - **mask** — the extracted foreground mask -输出可预览，也可作为其他节点的输入进行合成、进一步编辑或保存。 +Outputs can be previewed and used as inputs to other nodes for compositing, further editing, or saving. - - 本工作流使用了 Subgraph 节点实现模块化处理。查阅 Subgraph 文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 补充说明 +## Additional Notes -- **模型目录** — 模型必须放置在 `ComfyUI/models/background_removal/` 目录，而非 `checkpoints` 文件夹 -- **更新需求** — BiRefNet 支持需要较新版本的 ComfyUI，请确保已更新到最新版本 -- **RGBA 输出** — 透明背景结果可直接合成到新背景上，或用于下游工作流 +- **Model directory** — the model must be placed in `ComfyUI/models/background_removal/`, not the `checkpoints` folder +- **Update required** — BiRefNet support is available in recent ComfyUI versions. Make sure your ComfyUI is up to date. +- **RGBA output** — the transparent background result can be directly composited onto new backgrounds or used in downstream workflows diff --git a/zh/tutorials/utility/seedvr2.mdx b/zh/tutorials/utility/seedvr2.mdx index be26b6786..8ac72ffab 100644 --- a/zh/tutorials/utility/seedvr2.mdx +++ b/zh/tutorials/utility/seedvr2.mdx @@ -60,79 +60,137 @@ ComfyUI/ │ └── seedvr2_ema_vae_fp16.safetensors ``` -## 1. 图像缩放（3B INT8） +## Example Workflows -**功能说明：** 使用 SeedVR2 3B INT8 模型缩放单张图像。INT8 量化变体在质量和显存占用之间取得了良好平衡。 +### SeedVR2 3B Int8: Upscale Image (`utility_seedvr2_3b_int8_upscale_image`) - - 下载 JSON 或在模板库中搜索 „SeedVR2 3B Int8: Upscale Image“ +Upscale images using SeedVR2 3B Int8, a one-step diffusion-based video restoration model that produces high-quality results with improved temporal consistency. + +SeedVR2 3B Int8 upscale image workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Image" in Template Library + + +**输入素材** + +Upload this file to the matching `LoadImage` node: - - 获取此工作流的示例输入图像 + + + `LoadImage` node 1 · `watch_macro_shot.png` + -![SeedVR2 3B INT8 缩放预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_3b_int8_upscale_image.png) +**输出示例** -### 1.1 运行步骤 +
+ Input image + SeedVR2 3B Int8 upscale example output +
-1. 将图像放入 `ComfyUI/input/` 文件夹中，然后在 `加载图像` 节点中选择它 -2. 在 SeedVR2 模型加载器中选择 `seedvr2_3b_int8_convrot.safetensors` 检查点 -3. 点击 `队列` 或使用 `Ctrl(cmd) + 回车` 运行 +### 1.1 Steps to Run ---- +1. Place your image in the `ComfyUI/input/` folder and select it in the `Load Image` node +2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run + +### SeedVR2 7B Int8: Upscale Image (`utility_seedvr2_7b_int8_upscale_image`) -## 2. 图像缩放（7B INT8） +Upscale images using SeedVR2 7B Int8, a one-step diffusion model that enhances resolution through adversarial training and adaptive window attention. -**功能：** 使用 SeedVR2 7B INT8 模型缩放单张图像。更大的 7B 模型通过 INT8 量化提供更高质量的结果，同时高效利用 VRAM。 +SeedVR2 7B Int8 upscale image workflow preview - - 下载 JSON，或在模板库中搜索“SeedVR2 7B Int8: Upscale Image” + + + Open in Comfy Cloud + + Download JSON or search "SeedVR2 7B Int8: Upscale Image" in Template Library + + + +**输入素材** - - 获取此工作流的示例输入图像 +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 1 · `indoor_portrait.png` + -![SeedVR2 7B INT8 缩放预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/utility_seedvr2_7b_int8_upscale_image.png) +**输出示例** -### 2.1 运行步骤 +
+ Input image + SeedVR2 7B Int8 upscale example output +
-1. 将图像放入 `ComfyUI/input/` 文件夹，并在“加载图像”节点中选择该图像 -2. 在 SeedVR2 模型加载器中选择 `seedvr2_7b_int8_convrot.safetensors` 检查点 -3. 点击“队列”或使用 `Ctrl(Cmd) + Enter` 运行 +### 2.1 Steps to Run ---- +1. Place your image in the `ComfyUI/input/` folder and select it in the `Load Image` node +2. Select the `seedvr2_7b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run -## 3. 视频放大（3B INT8） +### SeedVR2 3B Int8: Upscale Video (`utility_seedvr2_3b_int8_upscale_video`) -**功能：** 使用 SeedVR2 3B INT8 模型对视频进行放大。该工作流在提升分辨率的同时，保持帧间的时间一致性。适用于修复老旧或损坏的素材，以及对低分辨率视频进行放大。 +Upscale and restore video footage using SeedVR2 3B Int8, a one-step diffusion model that enhances resolution while maintaining temporal consistency across frames. - - 下载 JSON，或在模板库中搜索 "SeedVR2 3B Int8: Upscale Video" +SeedVR2 3B Int8 upscale video workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SeedVR2 3B Int8: Upscale Video" in Template Library + - - 获取该工作流的示例输入视频 +**输入素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 73 · `grainy_perfume_shot_crf32.mp4` + + +**输出示例** + +
+ + +
-### 3.1 运行步骤 +### 3.1 Steps to Run -1. 将您的视频放入 `ComfyUI/input/` 文件夹，并在 `Load Video` 节点中选择它。 -2. 在 SeedVR2 模型加载器中选择 `seedvr2_3b_int8_convrot.safetensors` 检查点。 -3. 点击 `Queue` 或使用 `Ctrl(cmd) + Enter` 运行。 +1. Place your video in the `ComfyUI/input/` folder and select it in the `Load Video` node +2. Select the `seedvr2_3b_int8_convrot.safetensors` checkpoint in the SeedVR2 model loader +3. Click `Queue` or use `Ctrl(cmd) + Enter` to run -### 性能 +### Performance -更高的目标分辨率需要更长的处理时间。与 FP16 相比，INT8 变体提供了高效的推理，并减少了 VRAM 占用。 +Higher target resolutions require more processing time. The INT8 variant provides efficient inference with reduced VRAM usage compared to FP16. --- -## 社区资源 +## Community Resources -- [SeedVR2 项目页面](https://iceclear.github.io/projects/seedvr2/): 官方网站 -- [ByteDance SeedVR 代码库 (GitHub)](https://github.com/ByteDance-Seed/SeedVR): 原始研究代码和论文 -- [Comfy-Org/SeedVR2 (HuggingFace)](https://huggingface.co/Comfy-Org/SeedVR2): ComfyUI 模型权重 -- [ByteDance-Seed/SeedVR2-3B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-3B): 原始 3B 模型权重 -- [ByteDance-Seed/SeedVR2-7B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-7B): 原始 7B 模型权重 -- [论文 (arXiv)](https://arxiv.org/abs/2506.05301) +- [SeedVR2 Project Page](https://iceclear.github.io/projects/seedvr2/): Official project website +- [ByteDance SeedVR Codebase (GitHub)](https://github.com/ByteDance-Seed/SeedVR): Original research code and paper +- [Comfy-Org/SeedVR2 (HuggingFace)](https://huggingface.co/Comfy-Org/SeedVR2): Official ComfyUI model weights +- [ByteDance-Seed/SeedVR2-3B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-3B): Original 3B model weights +- [ByteDance-Seed/SeedVR2-7B (HuggingFace)](https://huggingface.co/ByteDance-Seed/SeedVR2-7B): Original 7B model weights +- [Paper (arXiv)](https://arxiv.org/abs/2506.05301) diff --git a/zh/tutorials/utility/video-segment-sam3.mdx b/zh/tutorials/utility/video-segment-sam3.mdx index 1883a1d74..3221a77f0 100644 --- a/zh/tutorials/utility/video-segment-sam3.mdx +++ b/zh/tutorials/utility/video-segment-sam3.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": c38c57a0 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -38,39 +37,79 @@ SAM 3.1 能根据文本提示在视频帧中分割并追踪物体。以上示例 > **限制：** 模型文本提示的 token 上限为 32 个。为获得最佳结果，请保持提示简短并聚焦于目标物体。 -## SAM 3.1 分割工作流 +## SAM 3.1 Segment Workflows + +### 1. Download Workflow -### 1. 工作流文件下载 +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find the SAM 3.1 workflows under the Utility category. -请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` 找到 Utility 类别下的 SAM 3.1 工作流。 +### SAM3: Video Segmentation (`utility_video_segment_sam3`) -**视频分割：** +Use the SAM3 model to segment the main subject or content from a video, isolating specific objects or regions. - - 下载视频工作流 +SAM3 video segmentation workflow preview + + + + Open in Comfy Cloud + + + Download JSON or search "SAM3: Video Segmentation" in Template Library + - - Open in cloud +**输入素材** + +Upload this file to the matching `LoadVideo` node: + + + + `LoadVideo` node 115 · `drinking_unicorn.mp4` + + +**输出示例** + + + +### SAM3: Image Segmentation (`utility_image_segment_sam3`) + +Use the SAM3 model to segment the main subject or content from a photo or image, isolating specific objects or regions. -**图像分割：** +SAM3 image segmentation workflow preview - - 下载图像工作流 + + + Open in Comfy Cloud + + Download JSON or search "SAM3: Image Segmentation" in Template Library + + + +**输入素材** - - Open in cloud +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 79 · `neon_guitarist.png` + + +
+ Input image +
-### 2. 手动下载模型 +### 2. Download Models -SAM 3.1 模型托管在 [Comfy-Org SAM 3.1 模型仓库](https://huggingface.co/Comfy-Org/sam3.1)。 +The SAM 3.1 model is hosted on the [Comfy-Org SAM 3.1 model repository](https://huggingface.co/Comfy-Org/sam3.1). -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) -放置到以下目录结构： +Place it in the following directory structure: ``` 📂 ComfyUI/ @@ -79,36 +118,36 @@ SAM 3.1 模型托管在 [Comfy-Org SAM 3.1 模型仓库](https://huggingface.co/ └── sam3.1_multiplex_fp16.safetensors ``` -### 3. 使用工作流 +### 3. Using the Workflows -**图像分割：** +**Image Segmentation:** -- **图像** — 通过 `Load Image` 节点加载图像（放入 ComfyUI 的 `input/` 文件夹） -- **物体提示** — 要分割物体的简短文本描述，例如 `person`、`car`、`cat` -- 输出为图像的遮罩，RGBA 预览显示分割结果 +- **Image** — Load an image via the `Load Image` node (place it in the ComfyUI `input/` folder) +- **Object Prompt** — A short text description of the object(s) to segment, e.g. `person`, `car`, `cat` +- The output is a mask applied to the image, with an RGBA preview showing the segmentation result -**视频分割：** +**Video Segmentation:** -- **视频** — 通过 `Load Video` 节点加载视频 -- **物体提示** — 同上，描述要在各帧之间追踪和分割的物体的简短文本 -- 输出包含每帧的遮罩和边界框 +- **Video** — Load a video via the `Load Video` node +- **Object Prompt** — Same as image, a short text prompt describing what to track and segment across frames +- The output provides masks and bounding boxes for each frame -**提示格式：** +**Prompt format:** -| 提示 | 作用 | -|------|------| -| SAM3 物体提示 | 描述要分割的**物体**。最多 32 个 token。 | +| Prompt | Role | +|--------|------| +| SAM3 object prompt | Short description of **what** to segment. Max 32 tokens. | -如需分别提示多个主体，用逗号分隔，并使用 `:N` 指定每个提示检测的最大物体数量： +To prompt multiple subjects separately, separate with commas and use `:N` to specify the max amount of objects detected per prompt: `eye:2, window panels:4` - - 本工作流使用了 Subgraph 节点实现模块化处理。查阅 Subgraph 文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 补充说明 +## Additional Notes -- **保持提示简短而具体** — 模型每个提示有 32 个 token 的限制 -- **多物体检测** — 使用逗号分隔不同物体类型，用 `:N` 限制每种类型的检测数量 -- **分割遮罩** — 输出遮罩可作为其他工作流的输入（例如修复、背景移除） -- **更新需求** — 确保 ComfyUI 已更新到最新版本以支持 SAM 3.1 +- **Keep prompts short and specific** — the model has a 32-token limit per prompt +- **Multi-object detection** — use commas to separate different object types, and `:N` to cap detections per type +- **Segmentation masks** — the output mask can be used as input to other workflows (e.g., inpainting, background removal) +- **Update required** — make sure ComfyUI is updated to the latest version to access SAM 3.1 support diff --git a/zh/tutorials/utility/video-upscale.mdx b/zh/tutorials/utility/video-upscale.mdx index cfe48104c..3f41f6564 100644 --- a/zh/tutorials/utility/video-upscale.mdx +++ b/zh/tutorials/utility/video-upscale.mdx @@ -15,7 +15,6 @@ translationBlockHashes: "Tips": a5dcf8f3 --- - 本指南介绍 ComfyUI 中的视频放大工作流，包括本地模型和合作伙伴节点选项，适用于各种使用场景。 diff --git a/zh/tutorials/utility/void-video-inpainting.mdx b/zh/tutorials/utility/void-video-inpainting.mdx index bab1a6c0a..fd7d1b102 100644 --- a/zh/tutorials/utility/void-video-inpainting.mdx +++ b/zh/tutorials/utility/void-video-inpainting.mdx @@ -10,7 +10,6 @@ translationBlockHashes: "Additional Notes": a3d04a46 --- - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' @@ -45,44 +44,61 @@ VOID 在 ComfyUI 中获得原生支持（PR [#13403](https://github.com/Comfy-Or > **限制：** 模糊的遮罩、杂乱的运动或占据画面大部分的物体可能仍会产生不理想的结果——提示词无法修正根本的分割错误。 -## VOID 视频修复工作流 +## VOID Video Inpainting Workflow + +### 1. Download Workflow + +Update your ComfyUI to the latest version, then go to `Workflow` -> `Browse Templates` and find "VOID: Video Inpainting" under the Utility category. + +### VOID: Video Inpainting (`utility_void_video_inpainting`) -### 1. 工作流文件下载 +Upload a video and mask the object you want to remove. Generate a clean video with the object and its physical interactions deleted. -请更新你的 ComfyUI 到最新版本，并通过菜单 `工作流` -> `浏览模板` 找到 Utility 类别下的 "VOID: Video Inpainting"。 +VOID video inpainting workflow preview - - Download workflow + + + Open in Comfy Cloud + + Download JSON or search "VOID: Video Inpainting" in Template Library + + + +**输入素材** + +Upload this file to the matching `LoadVideo` node: - - Open in cloud + + + `LoadVideo` node 4 · `snowboarder.mp4` + -### 2. 手动下载模型 +### 2. Download Models -所有模型均托管在 [Comfy-Org VOID 模型仓库](https://huggingface.co/Comfy-Org/void-model)。 +All models are hosted on the [Comfy-Org VOID model repository](https://huggingface.co/Comfy-Org/void-model). -**扩散模型** — 核心的两阶段修复模型： +**Diffusion Models** — the core two-pass inpainting model: -- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass2.safetensors) — 精炼阶段，时间稳定性更佳 -- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/diffusion_models/void_pass1.safetensors) — 主要阶段 +- [void_pass2.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass2.safetensors) — Refinement pass, better temporal stability +- [void_pass1.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/diffusion_models/void_pass1.safetensors) — Primary pass -**VAE：** +**VAE:** -- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/vae/cogvideox_vae.safetensors) +- [cogvideox_vae.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/vae/cogvideox_vae.safetensors) -**光流模型：** +**Optical Flow:** -- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/resolve/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) +- [raft_large_C_T_SKHT_V2-ff5fadd5.safetensors](https://huggingface.co/Comfy-Org/void-model/blob/main/optical_flow/raft_large_C_T_SKHT_V2-ff5fadd5.safetensors) -**SAM3 分割模型：** +**SAM3 Checkpoint** — for segmentation: -- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/resolve/main/checkpoints/sam3.1_multiplex_fp16.safetensors) +- [sam3.1_multiplex_fp16.safetensors](https://huggingface.co/Comfy-Org/sam3.1/blob/main/checkpoints/sam3.1_multiplex_fp16.safetensors) -**文本编码器：** +**Text Encoder:** -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors) +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors) ``` 📂 ComfyUI/ @@ -100,34 +116,34 @@ VOID 在 ComfyUI 中获得原生支持（PR [#13403](https://github.com/Comfy-Or │ └── void_pass1.safetensors ``` -### 3. 使用工作流 +### 3. Using the Workflow -**输入参数：** +**Inputs:** -- **源视频** — 通过 `Load Video` 节点加载视频（放入 ComfyUI `input/` 文件夹） -- **正向提示词（修复填充）** — 描述移除物体**之后**的场景。关注剩余内容和画面效果，而非被移除的物体 - - 示例：`empty kitchen counter, daylight, tiles visible` -- **负向提示词** — 可选的防伪影词表，可以留空 -- **SAM3 目标提示词** — 简短的**要遮罩移除**的对象描述。SAM3 通过语义理解为目标物体创建分割遮罩 - - 示例：`person in blue jacket`、`red cup on table` - - SAM3 提示词上限为 **32 个 token**，多个目标用逗号分隔，用 `:N` 指定每个提示检测的物体数量：`eye:2, window panels:4` +- **Source video** — Load a video via the `Load Video` node (place it in the ComfyUI `input/` folder) +- **Positive prompt (inpaint fill)** — Describe the scene **after** removal. Focus on what remains and how it looks, not on what was removed + - Example: `empty kitchen counter, daylight, tiles visible` +- **Negative prompt** — Optional anti-artifact list; can be left empty +- **SAM3 object prompt** — A short label for **what** to mask out. SAM3 uses semantic understanding to create a segmentation mask for the target object. + - Example: `person in blue jacket`, `red cup on table` + - Max tokens for SAM3 prompts is **32**. To prompt multiple subjects separately, separate with commas and use `:N` to specify the max objects detected per prompt: `eye:2, window panels:4` -**提示词分工：** +**Modes:** -| 提示词 | 作用 | +| Prompt | Role | |--------|------| -| SAM3 目标 | **移除什么**（SAM3 通过语义分割创建遮罩） | -| 正向（修复） | **如何填补空洞** | +| SAM3 object | **What** is removed (SAM3 creates the mask via semantic segmentation) | +| Positive (inpaint) | **How** the hole is filled across time | -长片段或纹理复杂的背景建议使用 **Pass 2**（精炼阶段）获得更好的时间稳定性。仅使用 **Pass 1** 速度更快，但可能出现更多抖动。 +Use **Pass 2** (refinement pass) for longer clips or textured backgrounds where temporal stability matters. **Pass 1** alone is faster but may show more jitter. - - 本工作流使用了 Subgraph 节点来实现模块化的视频处理。查阅 Subgraph 文档了解如何自定义和扩展工作流。 + + This workflow uses Subgraph nodes for modular video processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. -## 补充说明 +## Additional Notes -- **遮罩质量至关重要** — 围绕目标物体的清晰紧致的遮罩能产生最佳效果 -- **提示词写作技巧** — 描述移除后场景应自然呈现的样子，而非描述移除本身 -- **负向提示词** 仅在你看到反复出现的缺陷时使用（水印、模糊、多余肢体等） -- **双阶段工作流** — 模板会自动运行 Pass 1 然后 Pass 2；测试时也可以仅运行 Pass 1 以加快迭代 +- **Mask quality matters** — a clean, tight mask around the target object produces the best results +- **Prompt writing tip** — describe the scene as it should appear _naturally_ after removal, not the removal itself +- **Use negative prompt** only when you see repeating defects (watermarks, blur, extra limbs) +- **Two-pass workflow** — the template runs Pass 1 then Pass 2 automatically; you can also run just Pass 1 for faster iterations during testing diff --git a/zh/tutorials/video/bytedance/bernini-r.mdx b/zh/tutorials/video/bytedance/bernini-r.mdx index 68b2f3c28..bc67331d8 100644 --- a/zh/tutorials/video/bytedance/bernini-r.mdx +++ b/zh/tutorials/video/bytedance/bernini-r.mdx @@ -13,8 +13,6 @@ translationBlockHashes: "Community Resources": 9b4a0aaf --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' # ComfyUI Bernini-R 简介 @@ -82,19 +80,27 @@ ComfyUI/ **功能说明：** 生成具有匹配光照的编辑图像，并查看前/后并排对比。适用于人像和产品重光照、照片集一致的光照、电商目录摄影。 +Bernini-R 图像编辑工作流预览 + - - 下载 JSON 或在模板库中搜索 "Bernini-R" - 在 Comfy Cloud 中打开 + + 下载 JSON 或在模板库中搜索 "Bernini-R" + -
- Bernini-R 图像编辑输出 - Bernini-R 图像编辑对比 -
+**输入素材** + + + + 下载默认输入图像，或使用你自己的图像。 + + + 下载默认参考图像，或使用你自己的图像。 + + ### 运行步骤 @@ -114,16 +120,27 @@ ComfyUI/ **功能说明：** 使用 Bernini-R 生成具有一致重光照的编辑视频。连接源视频、可选的参考图像或参考视频，选择任务类型，编写提示词，然后运行。 +Bernini-R 视频编辑工作流预览 + - - 下载 JSON 或在模板库中搜索 "Bernini-R" - 在 Comfy Cloud 中打开 + + 下载 JSON 或在模板库中搜索 "Bernini-R" + -![Bernini-R 视频编辑预览](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/video_bernini_r_video_editing-1.webp) +**输入素材** + + + + 下载默认输入视频，或使用你自己的视频。 + + + 下载默认参考图像，或使用你自己的图像。 + + ### 运行步骤 diff --git a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx index 7fa04fd6d..a83149bd1 100644 --- a/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx +++ b/zh/tutorials/video/cosmos/cosmos-predict2-video2world.mdx @@ -45,7 +45,6 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- - 请下载下面的图片作为输入文件： ![输入图片](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/cosmos/predict2/input.png) @@ -66,7 +65,6 @@ huggingface: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos- [wan_2.1_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors) - 文件保存位置 ``` 📂 ComfyUI/ diff --git a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx index 2daad36be..3212700a8 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video-1-5.mdx @@ -13,7 +13,6 @@ translationBlockHashes: "Super-resolution upscaler": 89338328 --- - import UpdateReminder from "/snippets/zh/tutorials/update-reminder.mdx"; [HunyuanVideo 1.5](https://github.com/Tencent/HunyuanVideo) 是由腾讯混元团队开发的轻量级 8.3B 参数模型。它可在消费级 GPU（24GB 显存）上提供旗舰级质量的视频生成，大幅降低了使用门槛，同时不影响质量。 diff --git a/zh/tutorials/video/hunyuan/hunyuan-video.mdx b/zh/tutorials/video/hunyuan/hunyuan-video.mdx index 244df58a0..1db63a43d 100644 --- a/zh/tutorials/video/hunyuan/hunyuan-video.mdx +++ b/zh/tutorials/video/hunyuan/hunyuan-video.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Try It Yourself": 076fb43f --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' +Workflow preview + 在 Comfy Cloud 中打开 @@ -79,7 +79,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Text Encoder** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - File save location ``` @@ -97,7 +96,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. 按步骤完成工作流 ![Wan2.2 Fun Camera Control 工作流步骤](/images/tutorial/video/wan/wan_2.2_14b_fun_camera.jpg) diff --git a/zh/tutorials/video/wan/wan2-2-fun-control.mdx b/zh/tutorials/video/wan/wan2-2-fun-control.mdx index d738bc2ba..32a516a7c 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-control.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-control.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Wan2.2 Fun Control Video Generation Workflow Example": cbbb7456 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Wan2.2-Fun-Control** 是 Alibaba PAI 团队推出的新一代视频生成与控制模型，通过引入创新性的控制代码（Control Codes）机制，结合深度学习和多模态条件输入，能够生成高质量且符合预设控制条件的视频。该模型采用 **Apache 2.0 许可协议**发布，支持商业使用。 @@ -42,7 +40,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' - 这里提供的工作流包含了两个版本： 1. 使用了 lightx2v 的 [Wan2.2-Lightning](https://huggingface.co/lightx2v/Wan2.2-Lightning) 4 步 LoRA : 但可能导致生成的视频动态会有损失，但速度会更快 2. 没有使用加速 LoRA 的 fp8_scaled 版本 @@ -56,7 +53,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 由于使用了4 步 LoRA 对于初次使用工作流的用户体验较好， 但可能导致生成的视频动态会有损失， 我们默认启用了使用了加速 LoRA 版本，如果你需要启用另一组的工作流，框选后使用 **Ctrl+B** 即可启用 - ### 1. 工作流及素材下载 下载下面的视频或者 JSON 文件并拖入 ComfyUI 中以加载对应的工作流 @@ -67,6 +63,8 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_fun_control/wan2.2_14B_fun_inp.mp4" > +Workflow preview + 在 Comfy Cloud 中打开 @@ -106,7 +104,6 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Text Encoder** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - File save location ``` @@ -124,7 +121,6 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. 按步骤完成工作流 ![Wan2.2 Fun Control 工作流步骤](/images/tutorial/video/wan/wan2_2/wan_2.2_14b_fun_control.jpg) diff --git a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx index 80cbea3c4..bc7b7fd8b 100644 --- a/zh/tutorials/video/wan/wan2-2-fun-inp.mdx +++ b/zh/tutorials/video/wan/wan2-2-fun-inp.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Wan2.2 Fun Inp Start-End Frame Video Generation Workflow Example": 692b42a8 --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' **Wan2.2-Fun-Inp** 是 Alibaba pai团队推出的首尾帧控制视频生成模型，支持输入**首帧和尾帧图像**，生成中间过渡视频，为创作者带来更强的创意控制力。该模型采用 **Apache 2.0 许可协议**发布，支持商业使用。 @@ -62,18 +60,15 @@ import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' 或者，在更新 ComfyUI 至最新版本后，下载下面的工作流并拖入 ComfyUI 中加载。 - - - Wan2.2 Fun Inp workflow - Wan2.2 Fun Inp workflow + - - Download JSON or search "Wan2.2 Fun Inp" in Template Library - 在 Comfy Cloud 中打开 + + 下载 JSON，或在模板库中搜索「Wan2.2 Fun Inp」 + ### Input materials diff --git a/zh/tutorials/video/wan/wan2-2-s2v.mdx b/zh/tutorials/video/wan/wan2-2-s2v.mdx index 47076544f..545a51064 100644 --- a/zh/tutorials/video/wan/wan2-2-s2v.mdx +++ b/zh/tutorials/video/wan/wan2-2-s2v.mdx @@ -18,7 +18,6 @@ translationFrom: tutorials/video/wan/wan2-2-s2v.mdx Wan2.2 S2V 代码仓库：[Github](https://github.com/aigc-apps/VideoX-Fun) Wan2.2 S2V 模型仓库：[Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2V-14B) - ## Wan2.2 S2V ComfyUI 原生工作流 @@ -33,6 +32,8 @@ Wan2.2 S2V 模型仓库：[Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2 src="https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/video/wan/wan2.2_s2v/wan2.2-s2v.mp4" > +Workflow preview + 在 Comfy Cloud 中打开 @@ -82,7 +83,6 @@ Wan2.2 S2V 模型仓库：[Hugging Face](https://huggingface.co/Wan-AI/Wan2.2-S2 **text_encoders** - [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors) - ``` ComfyUI/ ├───📂 models/ @@ -97,14 +97,12 @@ ComfyUI/ │ └── wan_2.1_vae.safetensors ``` - ### 3. 工作流说明 ![工作流说明](/images/tutorial/video/wan/wan_2.2_14b_s2v.jpg) #### 3.1 关于 Lightning LoRA - #### 3.2 关于 fp8_scaled 和 bf16 模型 你可以在 [这里](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models) 找到两种模型： diff --git a/zh/tutorials/video/wan/wan2_2.mdx b/zh/tutorials/video/wan/wan2_2.mdx index ea705fa99..d1007e720 100644 --- a/zh/tutorials/video/wan/wan2_2.mdx +++ b/zh/tutorials/video/wan/wan2_2.mdx @@ -16,8 +16,6 @@ translationBlockHashes: "Community Resources": 7463b48b --- - - import UpdateReminder from '/snippets/zh/tutorials/update-reminder.mdx' -APP モードは、ComfyUI フロントエンドバージョン 1.41.13 から正式にサポートされています。それ以前のバージョンは早期テスト段階です。 +アプリモードは、ComfyUI フロントエンドバージョン 1.41.13 から正式にサポートされています。それ以前のバージョンは早期テスト段階です。 ## APP モードに入る @@ -82,7 +80,7 @@ APP モードに入るには 2 つの方法があります。左上のアイコ ![入出力レイアウトを示すプレビューモード](/images/interface/app_mode/app_mode_5.png) -1. これはプレビューモードです。すべての入出力設定が预期通りに表示されているか確認します。 +1. これはプレビューモードです。すべての入出力設定が期待通りに表示されているか確認します。 2. 調整が必要な場合は、**Back** をクリックして前のステップに戻ります。 ### ステップ 4：デフォルトビューの設定 @@ -108,9 +106,9 @@ APP モードでは、右パネルに設定されたすべての入力オプシ 1. 右側の **Input** パネルで、実行前に入力パラメータを調整できます。 2. **Run** をクリックしてワークフローを実行します。 -3. 左サイドバーでは、APP の再構築、アセットの表示、ワークフローパネルの開くなどの追加制御を提供します。 +3. 左サイドバーでは、APP の再構築、アセットの表示、およびワークフローパネルを開くなどの追加制御を提供します。 -APP モードはモバイルおよび狭い画面レイアウトにも最適化されています。インターフェースはタブベースのビューに適応し、入力パネル、出力ビュー、メディアアセットパネル之间を切り替えることができます。 +APP モードはモバイルおよび狭い画面レイアウトにも最適化されています。インターフェースはタブベースのビューに適応し、入力パネル、出力ビュー、メディアアセットパネルの間を切り替えることができます。 ![入力、出力、アセットタブ付きの APP モードモバイルレイアウト](/images/interface/app_mode/app_mode_15.png) @@ -133,7 +131,7 @@ APP モードはモバイルおよび狭い画面レイアウトにも最適化 ![保存された APP ワークフローを示す左サイドバー](/images/interface/app_mode/app_mode_11.png) 1. 左サイドバーのボタンをクリックしてパネルを開き、保存されたすべての APP ワークフローを表示します。 -2. このパネルから保存された APP ワークフロー之间を切り替えます。 +2. このパネルから保存された APP ワークフロー間を切り替えます。 ### ワークフローの共有 @@ -150,17 +148,17 @@ APP モードで、左サイドバーの **Share** ボタンをクリックし ![ワークフローの保存を促す共有ダイアログ](/images/interface/app_mode/app_mode_12.png) 1. 左サイドバーの Share ボタンをクリックして、共有ダイアログを開きます。 -2. ワークフローが未保存の場合、ワークフロー名を入力し、**Save workflow** をクリックして続行します。 +2. ワークフローが未保存の場合、ワークフロー名を入力し、**ワークフローを保存** をクリックして続行します。 保存されると、共有ダイアログにリンクを生成するオプションが表示されます。 -![Create a link ボタン付きの共有ダイアログ](/images/interface/app_mode/app_mode_13.png) +![リンクを作成 ボタン付きの共有ダイアログ](/images/interface/app_mode/app_mode_13.png) -**Create a link** をクリックして、共有可能なワークフローの URL を生成します。 +**リンクを作成** をクリックして、共有可能なワークフローの URL を生成します。 -![コピーボタン付きの生成されたリンクを示す共有ダイアログ](/images/interface/app_mode/app_mode_14.png) +![コピーボタン付きの生成済みリンクを示す共有ダイアログ](/images/interface/app_mode/app_mode_14.png) -**Copy** をクリックしてリンクをコピーします。受信者はリンクを開いてワークフローを直接実行できます。完全なワークフロー設定がリンクに含まれています。 +**コピー** をクリックしてリンクをコピーします。受信者はリンクを開いてワークフローを直接実行できます。完全なワークフロー設定がリンクに含まれています。 共有リンクは Comfy Cloud でのみサポートされています。 diff --git a/ja/interface/appearance.mdx b/ja/interface/appearance.mdx index a1f5b9580..796cd3584 100644 --- a/ja/interface/appearance.mdx +++ b/ja/interface/appearance.mdx @@ -15,8 +15,6 @@ translationBlockHashes: "Tree Explorer": 693b6eaf "Advanced Customization with user.css": 96af170b --- - - import UserDirectory from "/snippets/ja/install/user-directory.mdx" import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.mdx" @@ -153,13 +151,13 @@ ComfyUI の外観をカスタマイズする主な方法は、組み込みのカ ## サイドバー -### 統一サイドバー幅 +### サイドバーの幅を統一 -- **機能**: 有効にすると、異なるサイドバー之间を切り替える際、サイドバーの幅は統一された幅になります。無効にすると、異なるサイドバーは切り替え時にカスタム幅を維持できます。 +- **機能**: 有効にすると、異なるサイドバー間を切り替える際、サイドバーの幅は一定の幅に統一されます。無効にすると、異なるサイドバーは切り替え時にそれぞれのカスタム幅を維持できます。 -### サイドバーサイズ +### サイドバーのサイズ -- **機能**: サイドバーのサイズを制御します。通常または小に設定できます。 +- **機能**: サイドバーのサイズを制御します。ノーマルまたはスモールに設定できます。 ### サイドバーの位置 @@ -167,9 +165,9 @@ ComfyUI の外観をカスタマイズする主な方法は、組み込みのカ ### サイドバーのスタイル -- **機能**: サイドバーの視覚スタイルを制御します。オプションには以下が含まれます： - - **Connected**: サイドバーがインターフェースの端に接続されて表示されます。 - - **Floating**: サイドバーがフローティングパネルとして表示され、インターフェースの端から視覚的に分離されます。 +- **機能**: サイドバーの視覚的なスタイルを制御します。オプションは次のとおりです： + - **接続**: サイドバーがインターフェースの端に接続されて表示されます。 + - **フローティング**: サイドバーがフローティングパネルとして表示され、インターフェースの端から視覚的に分離されます。 ![サイドバーのスタイル](/images/interface/setting/appearance/sidbar_style.jpg) diff --git a/ja/interface/credits.mdx b/ja/interface/credits.mdx index 92936b22e..6a2c8a9db 100644 --- a/ja/interface/credits.mdx +++ b/ja/interface/credits.mdx @@ -10,13 +10,11 @@ translationBlockHashes: "How to Purchase Credits?": b6aeb110 "Frequently Asked Questions": 4ccf7056 --- - - import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.mdx" -クレジットシステムは `Partner Nodes` をサポートするために追加されました。クローズドソースの AI モデルを呼び出すにはトークンの消費が必要なため、適切なクレジット管理が必要です。デフォルトでは、クレジットインターフェースは表示されません。まず `設定` -> `ユーザー` で ComfyUI アカウントにログインしてください。その後、`設定` -> `クレジット` で関連アカウントのクレジット情報を確認できます。 +クレジットシステムは `パートナーノード` をサポートするために追加されました。クローズドソースの AI モデルを呼び出すにはトークンの消費が必要なため、適切なクレジット管理が必要です。デフォルトでは、クレジットインターフェースは表示されません。まず `設定` -> `ユーザー` で ComfyUI アカウントにログインしてください。その後、`設定` -> `クレジット` で関連アカウントのクレジット情報を確認できます。 ![ComfyUI クレジットインターフェース](/images/interface/setting/menu-credits.jpg) @@ -78,7 +76,7 @@ ComfyUI は常に完全オープンソースであり、ローカルユーザー - クレジットは負の残高やクレジットラインとして使用することを意図していません。ただし、Partner Nodes が実行前に常にコストを報告するとは限らない競合状態により、1回の実行で残高を超えるクレジットを消費し、完了後に一時的に負の残高になる可能性があります。残高が負の場合、チャージして正の残高に回復するまで Partner Nodes を実行できません。API 呼び出しを行う前に十分なクレジットがあることを確認してください。 + クレジットは負の残高やクレジットラインとして使用することを意図していません。ただし、パートナーノードが実行前に常にコストを報告するとは限らない競合状態により、1回の実行で残高を超えるクレジットを消費し、完了後に一時的に負の残高になる可能性があります。残高が負の場合、チャージして正の残高に回復するまでパートナーノードを実行できません。API 呼び出しを行う前に十分なクレジットがあることを確認してください。 diff --git a/ja/interface/features/partial-execution.mdx b/ja/interface/features/partial-execution.mdx index 58c9a684f..13f162f83 100644 --- a/ja/interface/features/partial-execution.mdx +++ b/ja/interface/features/partial-execution.mdx @@ -3,7 +3,7 @@ title: "部分実行 - ComfyUI でワークフローの一部のみを実行" description: "ComfyUI における部分実行機能の使用方法と要件について" sidebarTitle: "部分実行" icon: "play" -translationSourceHash: 99b15429 +translationSourceHash: 576da459 translationFrom: interface/features/partial-execution.mdx --- diff --git a/ja/interface/features/subgraph.mdx b/ja/interface/features/subgraph.mdx index bb4af8a0a..8ee844af2 100644 --- a/ja/interface/features/subgraph.mdx +++ b/ja/interface/features/subgraph.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Unpack Subgraphs to Nodes": 2fb4b38f "Subgraph Blueprint": 98873b2f --- - - サブグラフ機能には ComfyUI フロントエンドバージョン 1.24.3 以降が必要です。この機能が表示されない場合は、以下を参照してください：[ComfyUI の更新方法](/ja/installation/update_comfyui) - 本文書の画像は nightly バージョンのフロントエンドで作成されています。実際のインターフェースを参照してください @@ -49,16 +47,16 @@ translationBlockHashes: ComfyUI でグループ化したいノードを選択します - ![Subgraph icon](/images/interface/features/subgraph/subgraph_icon.jpg) + ![サブグラフアイコン](/images/interface/features/subgraph/subgraph_icon.jpg) ツールバーでサブグラフアイコンを見つけます - ![Workflow using subgraph](/images/interface/features/subgraph/workflow_using_subgraph.jpg) + ![サブグラフを使用したワークフロー](/images/interface/features/subgraph/workflow_using_subgraph.jpg) ComfyUI は、選択したノードの入力と出力に基づいて自動的にサブグラフを作成します [サブグラフの編集](#editing-subgraphs) を参照してください。サブグラフを編集および整理して、完全に機能するノードを作成できます - ![Workflow using subgraph](/images/interface/features/subgraph/subgraph_after_edited.jpg) + ![サブグラフを使用したワークフロー](/images/interface/features/subgraph/subgraph_after_edited.jpg) @@ -102,7 +100,7 @@ ComfyUI v0.3.66 以降では、サブグラフ内部に入ることなく、パ 任意のサブグラフを選択し、「Edit Subgraph Widgets」ボタンを使用してパラメータパネルを開くことができます。 ![Open Parameters Panel](/images/interface/features/subgraph/parameters_panel_open.jpg) -開くと、パラメータパネルで直接サブグラフウィジェット的顺序と可視性を編集できます。 +開くと、パラメータパネルで直接サブグラフウィジェットの順序と可視性を編集できます。 ![Open Parameters Panel](/images/interface/features/subgraph/parameters_panel_edit.jpg) 1. 並べ替え：ウィジェットを右クリックして押し、希望の位置にドラッグします diff --git a/ja/interface/features/template.mdx b/ja/interface/features/template.mdx index 82c7cfbeb..a013f8c45 100644 --- a/ja/interface/features/template.mdx +++ b/ja/interface/features/template.mdx @@ -14,8 +14,6 @@ translationBlockHashes: "Custom node templates": 16edae72 "How to add templates for custom nodes?": f022e0b7 --- - - ワークフローテンプレートは、ComfyUI がネイティブにサポートするモデルワークフローおよびカスタムノード提供のサンプルワークフローを閲覧するための機能です。 ComfyUI のワークフローテンプレートでは、以下内容を見つけることができます： @@ -125,9 +123,9 @@ ComfyUI のワークフローテンプレートでは、以下内容を見つけ テンプレートは個別の依存関係として管理および更新されます：[`comfyui-workflow-templates`](https://pypi.org/project/comfyui-workflow-templates/)。 -ComfyUI を更新した後、ドキュメントや新たに発表されたテンプレートが表示されない場合は、対応する依存関係を更新する必要があるかもしれません。[`ComfyUI/requirements.txt`](https://github.com/Comfy-Org/ComfyUI/blob/master/requirements.txt) でバージョンを確認できます。 +ComfyUIをアップデートした後、ドキュメントや新たに発表されたテンプレートが表示されない場合は、対応する依存関係を更新する必要があるかもしれません。[`ComfyUI/requirements.txt`](https://github.com/Comfy-Org/ComfyUI/blob/master/requirements.txt) でバージョンを確認できます。 -通常、ComfyUI の更新時に以下の 3 つの依存関係も一緒にアップグレードされることがあります： +通常、ComfyUIがアップデートされたときに以下の3つの依存関係も一緒にアップグレードされることがあります： ``` comfyui-frontend-package==1.24.4 @@ -135,7 +133,7 @@ comfyui-workflow-templates==0.1.52 comfyui-embedded-docs==0.2.4 ``` -正しい更新方法がわからない場合は、ComfyUI およびその依存関係の更新方法について [Update ComfyUI](/ja/installation/update_comfyui) を参照してください。 +正しい更新方法がまだ分からない場合は、ComfyUIとその依存関係をアップデートする方法について [Update ComfyUI](/ja/installation/update_comfyui) を参照してください。 ## ComfyUI リポジトリにテンプレートを貢献するには？ diff --git a/ja/interface/maskeditor.mdx b/ja/interface/maskeditor.mdx index c6a979a17..3f935c7b9 100644 --- a/ja/interface/maskeditor.mdx +++ b/ja/interface/maskeditor.mdx @@ -18,23 +18,19 @@ translationBlockHashes: "Settings": 190f15ac "Video Tutorial": 1b274ea6 --- - - マスクエディターは ComfyUI に組み込まれたツールで、外部画像編集ソフトを使わずに画像上で直接マスクを作成・編集できます。複数の描画ツール、レイヤー管理、元に戻す/やり直し、キャンバス変換、GPU アクセラレーションによるブラシ描画をサポートしています。 ---- - ## 開き方 -マスクエディターを開くには 3 つの方法があります： +マスクエディターを開くには3つの方法があります： -1. **選択ツールバー** — Load Image ノードを選択し、ノード上部の選択ツールバーにある Mask アイコンボタンをクリックします。 -2. **画像オーバーレイ** — 画像プレビュー（ギャラリーモード）にマウスを乗せ、左上に表示される「Edit or mask image」ボタン（mask アイコン）をクリックします。 -3. **右クリックメニュー** — Load Image ノードを右クリックし、メニューから「マスクエディタで開く」を選択します。 +1. **選択ツールバー** — Load Imageノードを選択し、ノード上部の選択ツールバーにあるマスクアイコンボタンをクリックします。 +2. **画像オーバーレイ** — 画像プレビュー（ギャラリーモード）にマウスを乗せ、左上に表示される「画像を編集またはマスク」ボタン（マスクアイコン）をクリックします。 +3. **右クリックメニュー** — Load Imageノードを右クリックし、コンテキストメニューから「マスクエディターで開く」を選択します。 > マスクエディターは**グラフモード**と**アプリモード**の両方で使用できます。 -![マスクエディターを開く 3 つの方法：Selection Toolbox、Image Overlay、右クリックメニュー](/images/interface/maskeditor/maskeditor_3ways_to_open_maskeditor.png) +![マスクエディターを開く3つの方法：Selection Toolbox、Image Overlay、右クリックメニュー](/images/interface/maskeditor/maskeditor_3ways_to_open_maskeditor.png) --- @@ -48,7 +44,7 @@ translationBlockHashes: |---|-------|------|------| | 1 | **トップバー** | 上部 | 元に戻す/やり直し、キャンバス変換、反転/クリア、保存/キャンセル | | 2 | **ツールパネル** | 左側 | ツール選択とズーム情報 | -| — | **キャンバスエリア** | 中央 | 描画・編集を行うメインキャンバスエリア | +| — | **ポインタゾーン** | 中央 | 描画・編集を行うメインキャンバスエリア | | 3 | **サイドパネル** | 右側 | ツール固有の設定とレイヤーコントロール | --- @@ -65,7 +61,7 @@ translationBlockHashes: | 2 | **Paint Pen** | ペイントレイヤー（RGB） | 画像の RGB レイヤーに直接描画します。インペイントやベース画像の修正に便利です。 | | 3 | **Eraser** | マスクレイヤー / ペイントレイヤー | 既存のマスクやペイントレイヤーの一部を消去します。 | | 4 | **Paint Bucket** | マスクレイヤー | 色の類似性に基づく塗りつぶしツール。ピクセルをクリックすると接続領域にマスクを塗りつぶします。マスク済み領域をクリックするとマスクを消去します。許容値（Tolerance）で塗りつぶしの広がりを制御します。 | -| 5 | **Color Select** | マスクレイヤー | ターゲット色に一致する全ピクセルを選択的にマスクします。高度なマッチングアルゴリズム（Simple / HSL / LAB）に対応。 | +| 5 | **Color Select** | マスクレイヤー | ターゲット色に一致する全ピクセルを選択的にマスクします。高度なマッチングアルゴリズム（Simple HSL / LAB）に対応。 | --- @@ -94,9 +90,9 @@ translationBlockHashes: ## レイヤー -右側パネル下部の **レイヤー** セクションでは、各レイヤーの表示/非表示を個別に切り替えられます： +右側パネルには **レイヤー** セクションがあり、各レイヤーの表示/非表示を個別に切り替えられます： -- **マスクレイヤー**：マスクオーバーレイ。**マスク不透明度**と**マスク合成オプション**を調整可能： +- **マスクレイヤー**：マスクオーバーレイ。**不透明度**と**ブレンドモード**を調整可能： - `黒` — マスクを黒のオーバーレイとして表示 - `白` — マスクを白のオーバーレイとして表示 - `ネガティブ` — マスクを反転表示 @@ -115,14 +111,14 @@ translationBlockHashes: | 設定 | 範囲 | 説明 | |------|------|------| -| **ブラシ形状** | Arc / Rect | 丸ブラシまたは四角ブラシ | -| **カラーピッカー** | 任意の 16 進色 | ペイント色（Paint Pen 用）またはマスク色 | +| **形状** | Arc / Rect | 丸ブラシまたは四角ブラシ | +| **色** | 任意の16進色 | ペイント色（Paint Pen用）またはマスク色 | | **太さ** | 1–250 | ブラシ直径（ピクセル） | | **不透明度** | 0–1 | ブラシストロークの透明度 | | **硬さ** | 0–1 | エッジの柔らかさ。1 = 硬いエッジ、0 = 非常に柔らかい | | **ステップサイズ** | 1–100 | ブラシの打点間隔。値が低いほど滑らかなストローク | -**デフォルトにリセット** でブラシパラメーターをデフォルトに戻せます。 +**デフォルトにリセット** を使用して、デフォルトのブラシパラメータを復元します。 --- @@ -143,10 +139,10 @@ translationBlockHashes: ComfyUI の設定パネルで、以下のマスクエディターオプションを設定できます： -| 設定 | カテゴリー | デフォルト | 説明 | -|------|-----------|-----------|------| -| **Brush adjustment speed multiplier** | Mask Editor → Brush Adjustment → Sensitivity | 1.0 (0.1–2.0) | Alt+右クリックドラッグ時のブラシサイズ・硬さ変更速度。値が大きいほど速い | -| **Lock brush adjustment to dominant axis** | Mask Editor → Brush Adjustment → Use Dominant Axis | オン | 有効にすると、移動の主軸方向に応じてサイズ**または**硬さのみが調整される | +| 設定 | カテゴリ | デフォルト | 説明 | +|------|---------|-----------|------| +| **ブラシ調整速度の倍率** | マスクエディター → ブラシ調整 → 感度 | 1.0 (0.1–2.0) | Alt+右クリックドラッグ時のブラシサイズと硬さの変更速度 | +| **ブラシ調整を主軸にロック** | マスクエディター → ブラシ調整 → 主軸を使用 | オン | 有効にすると、移動の主方向に応じてサイズまたは硬さのみが調整される | --- diff --git a/ja/interface/overview.mdx b/ja/interface/overview.mdx index cfb320a79..27c5b079f 100644 --- a/ja/interface/overview.mdx +++ b/ja/interface/overview.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "New Menu Interface": 9f329aea "Old Menu Version": 5cd18dc8 --- - - ビジュアルインターフェースは、現在大多数のユーザーが ComfyUI を使用して [ComfyUI Server](/ja/development/comfyui-server/comms_overview) を呼び出し、対応するメディアリソースを生成する方法です。これは、ユーザーがワークフローを操作および整理し、デバッグし、素晴らしい作品を作成するためのビジュアルインターフェースを提供します。 通常、ComfyUI サーバーを起動すると、次のようなインターフェースが表示されます。 @@ -53,9 +51,9 @@ translationBlockHashes: **トップヘッダーバー:** 4. **上部エリア**: 現在開いているワークフローを表示します。 -5. **新規ワークフローボタン**: クリックして新しい空白のワークフローファイルを作成します。 +5. **新規ワークフローボタン**: クリックして新しい空のワークフローファイルを作成します。 6. **右側制御エリア**: 実行およびキュー制御管理。ワークフローを実行し、キューを表示できます。 -7. **ログイン状態**: デフォルトでは表示されず、ログイン後のみ表示されます。クローズドソース API ノードが必要な場合に使用されます。 +7. **ログイン状態**: デフォルトでは表示されず、ログイン後のみ表示されます。クローズドソースのパートナーノードが必要な場合に使用されます。 8. **クイックアクセスボタン**: 右側パネルを開きます。 **右下キャンバス制御:** diff --git a/ja/interface/settings/3d.mdx b/ja/interface/settings/3d.mdx index 2c2a3fc32..21635d225 100644 --- a/ja/interface/settings/3d.mdx +++ b/ja/interface/settings/3d.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Light": 9e249283 "Scene": 601ed050 --- - - import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.mdx" @@ -29,30 +27,28 @@ import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.md ![カメラタイプ](/images/interface/setting/3d/camera_type.jpg) +## ライト -## 照明 +このセクションのライト設定は、3D コンポーネントのデフォルトのライト設定を設定するために使用されます。ComfyUI の 3D 設定にある対応する設定も変更可能です。 -このセクションの照明設定は、3D コンポーネントのデフォルト照明設定を設定するために使用されます。ComfyUI の 3D 設定にある対応する設定も変更可能です。 +![ライト](/images/interface/setting/3d/light.jpg) -![照明](/images/interface/setting/3d/light.jpg) - -### 照明調整刻み幅 +### ライト調整増分 - **デフォルト値**: 0.5 -- **機能**: 3D シーンで照明強度を調整する際の刻み幅を制御します。小さい値にするとより細かな照明調整が可能になり、大きい値にすると調整ごとの変化がより顕著になります +- **機能**: 3D シーンで光の強度を調整する際のステップサイズを制御します。小さい値にするとより細かな調整が可能になり、大きい値にすると調整ごとの変化がより顕著になります -### 照明強度下限 +### ライト強度の最小値 - **デフォルト値**: 1 -- **機能**: 3D シーンで許可される最小の照明強度値を設定します。これは、任意の 3D コントロールの照明を調整する際に設定可能な最低明度を定義します +- **機能**: 3D シーンで許可される光の強度の最小値を設定します。任意の 3D コントロールの照明を調整する際に設定可能な最低明度を定義します -### 照明強度上限 +### ライト強度の最大値 - **デフォルト値**: 10 -- **機能**: 3D シーンで許可される最大の照明強度値を設定します。これは、任意の 3D コントロールの照明を調整する際に設定可能な最高明度の上限を定義します +- **機能**: 3D シーンで許可される光の強度の最大値を設定します。任意の 3D コントロールの照明を調整する際に設定可能な最高明度の上限を定義します -### 初期照明強度 +### 初期ライト強度 - **デフォルト値**: 3 -- **機能**: 3D シーン内のライトのデフォルト明度レベルを設定します。この値は、新しい 3D コントロールを作成する際にライトがオブジェクトを照らす強度を決定しますが、作成後は各コントロールを個別に調整できます - +- **機能**: 3D シーン内のライトのデフォルトの明るさレベルを設定します。この値は、新しい 3D コントロールを作成する際にライトがオブジェクトを照らす強度を決定しますが、作成後は各コントロールを個別に調整できます ## シーン diff --git a/ja/interface/settings/about.mdx b/ja/interface/settings/about.mdx index 1056f8d7e..49d01ba37 100644 --- a/ja/interface/settings/about.mdx +++ b/ja/interface/settings/about.mdx @@ -1,12 +1,13 @@ --- -title: "バージョン情報ページ" -description: "ComfyUI 設定システムの『About』ページについての詳しい説明" +title: "About ページ" +description: "ComfyUI設定のAboutページの詳細な説明" icon: "info" -sidebarTitle: "バージョン情報" +sidebarTitle: "About" translationSourceHash: 23ce57df translationFrom: interface/settings/about.mdx --- + import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.mdx" @@ -54,4 +55,5 @@ app.registerExtension({ icon: 'pi pi-github' } ] -}) \ No newline at end of file +}) +``` diff --git a/ja/interface/settings/comfy.mdx b/ja/interface/settings/comfy.mdx index 97e57ca1d..f6cbc387f 100644 --- a/ja/interface/settings/comfy.mdx +++ b/ja/interface/settings/comfy.mdx @@ -22,28 +22,25 @@ translationBlockHashes: "Window": afe27bb3 "Workflow": 55b8bcfa --- - - import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.mdx" -## API ノード +## パートナーノード -### API ノードの価格バッジを表示 +### パートナーノードの価格バッジを表示 - **デフォルト値**: 有効 -- **機能**: API ノードに価格バッジを表示するかどうかを制御し、ユーザーが API ノードの使用コストを識別するのに役立ちます +- **機能**: パートナーノードに価格バッジを表示するかどうかを制御し、ユーザーがパートナーノードの使用コストを識別するのに役立ちます -![API ノードの価格バッジを表示](/images/interface/setting/comfy/api_node_pricing_badge.jpg) +![パートナーノードの価格バッジを表示](/images/interface/setting/comfy/api_node_pricing_badge.jpg) -> API ノードの詳細については、[API ノード](/ja/tutorials/partner-nodes/overview) を参照してください +> パートナーノードの詳細については、[パートナーノード](/ja/tutorials/partner-nodes/overview) を参照してください -## 開発者モード +## 開発モード -### 開発者モードオプションを有効化 (API 保存など) +### 開発モードオプションを有効化 (API 保存など) - **デフォルト値**: 無効 -- **機能**: 開発者モードオプション（API 保存など）を有効にします - +- **機能**: 開発モードオプション（API 保存など）を有効にします ## トークン重みの編集 @@ -55,10 +52,10 @@ import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.md -## 言語設定 +## ロケール ### 言語 -- **オプション**: 英語、中文 (中国語)、日本語、한국어 (韓国語)、Русский (ロシア語)、Español (スペイン語)、Français (フランス語) +- **オプション**: English, 中文 (Chinese), 日本語 (Japanese), 한국어 (Korean), Русский (Russian), Español (Spanish), Français (French) - **デフォルト値**: ブラウザの言語を自動検出 - **機能**: ComfyUI インターフェースの表示言語を変更します @@ -117,7 +114,6 @@ ComfyUI のイテレーションプロセス中、一部のノードを調整し ![検索で実験的ノードを表示](/images/interface/setting/comfy/beta_node.jpg) - ## ノード検索ボックス ### ノード提案数 @@ -149,7 +145,7 @@ ComfyUI のイテレーションプロセス中、一部のノードを調整し ### ノード検索ボックスの実装 - **デフォルト値**: デフォルト -- **機能**: ノード検索ボックスの実装方法を選択します（実験的機能）。`litegraph (旧版)` を選択すると、早期の ComfyUI 検索ボックスに切り替わります +- **機能**: ノード検索ボックスの実装方法を選択します（実験的機能）。`litegraph (レガシー)` を選択すると、早期の ComfyUI 検索ボックスに切り替わります ## ノードウィジェット @@ -202,7 +198,7 @@ ComfyUI のイテレーションプロセス中、一部のノードを調整し ### ワークフロー状態を永続化し、ページ（再）読み込み時に復元 - **デフォルト値**: 有効 -- **機能**: ページ（再）読み込み時にワークフロー状態を復元するかどうかを制御し、ページ刷新後にワークフロー内容を維持します +- **機能**: ページ（再）読み込み時にワークフロー状態を復元するかどうかを制御し、ページ更新後にワークフローコンテンツを維持します ### 自動保存 - **デフォルト値**: オフ @@ -221,30 +217,30 @@ ComfyUI のイテレーションプロセス中、一部のノードを調整し - **機能**: ワークフロー内のキャンバス位置とズームレベルを保存および復元するかどうかを制御し、ワークフローを再度開くときに以前のビュー状態を復元します ### 開かれたワークフローの位置 -- **オプション**: サイドバー、トップバー、トップバー（2 段目） +- **オプション**: サイドバー、トップバー、トップバー（2段目） - **デフォルト値**: トップバー -- **機能**: 開かれたワークフロータブの表示位置を制御します。現在、サイドバー、トップバー、トップバー（2 段目）のみサポートされています +- **機能**: 開かれたワークフロータブの表示位置を制御します。現在、サイドバー、トップバー、トップバー（2段目）のみサポートされています ### ワークフロー保存時にファイル名をプロンプト - **デフォルト値**: 有効 - **機能**: ワークフロー保存時にファイル名入力をプロンプトするかどうかを制御し、ユーザーがワークフローファイル名をカスタマイズできるようにします -### ワークフロー保存時にノード ID をソート +### ワークフロー保存時にノードIDを並び替え - **デフォルト値**: 無効 -- **機能**: ワークフロー保存時にノード ID をソートするかどうかを決定し、ワークフローファイル形式をより標準化し、バージョン管理に便利にします +- **機能**: ワークフロー保存時にノードIDを並び替えるかどうかを決定し、ワークフローファイル形式をより標準化し、バージョン管理に便利にします -### 欠落ノード警告を表示 +### 不足しているノードの警告を表示 - **デフォルト値**: 有効 -- **機能**: ワークフロー内の欠落ノードの警告を表示するかどうかを制御し、ユーザーがワークフロー内の利用不可ノードを識別するのに役立ちます +- **機能**: ワークフロー内の不足しているノードの警告を表示するかどうかを制御し、ユーザーがワークフロー内で利用できないノードを識別するのに役立ちます -### 欠落モデル警告を表示 +### 不足しているモデルの警告を表示 - **デフォルト値**: 有効 -- **機能**: ワークフローファイル内のウィジェット値にモデルリンク情報を追加して、モデルファイル読み込み時のプロンプトに使用することをサポートしています。有効にすると、ローカルに対応するモデルファイルがない場合、ワークフロー内の欠落モデルの警告が表示されます +- **機能**: ワークフローファイル内のウィジェット値にモデルリンク情報を追加して、モデルファイル読み込み時のプロンプト用にサポートしています。有効にすると、ローカルに対応するモデルファイルがない場合、ワークフロー内の不足しているモデルの警告が表示されます ### ワークフロークリア時に確認を要求 - **デフォルト値**: 有効 -- **機能**: ワークフローをクリアする際に確認ダイアログを表示するかどうかを制御し、ワークフロー内容の誤クリアを防ぎます +- **機能**: ワークフローをクリアする際に確認ダイアログを表示するかどうかを制御し、ワークフローコンテンツの誤クリアを防ぎます -### ノード ID をワークフローに保存 +### ノードIDをワークフローに保存 - **デフォルト値**: 有効 -- **機能**: ワークフロー保存時にノード ID を保存するかどうかを制御し、ワークフローファイル形式をより標準化し、バージョン管理に便利にします +- **機能**: ワークフロー保存時にノードIDを保存するかどうかを制御し、ワークフローファイル形式をより標準化し、バージョン管理に便利にします diff --git a/ja/interface/settings/lite-graph.mdx b/ja/interface/settings/lite-graph.mdx index e94cf5b16..c9b8eee59 100644 --- a/ja/interface/settings/lite-graph.mdx +++ b/ja/interface/settings/lite-graph.mdx @@ -1,8 +1,8 @@ --- title: "ComfyUI LiteGraph (キャンバス) 設定" -description: "ComfyUI 图形渲染エンジン LiteGraph の設定オプションの詳細説明" +description: "ComfyUI グラフィックレンダリングエンジン LiteGraph の設定オプションの詳細説明" icon: "diagram-project" -sidebarTitle: "Lite Graph 設定" +sidebarTitle: "Lite Graph" translationSourceHash: 870926b7 translationFrom: interface/settings/lite-graph.mdx translationBlockHashes: @@ -18,8 +18,6 @@ translationBlockHashes: "Pointer": 6b9ade60 "Reroute": bb9a582a --- - - import SettingsMenuContext from "/snippets/ja/interface/settings-menu-context.mdx" @@ -30,7 +28,7 @@ LiteGraph は ComfyUI の基盤となる图形レンダリングエンジンで ### 選択ツールボックスを表示 - **デフォルト値**: 有効 -- **機能**: 選択ツールボックスは、ノードを選択した後にノード上に浮动表示されるクイックアクションツールバーで、部分的な実行、ピン留め、削除、色の変更などの一般的なクイック操作を提供します。 +- **機能**: 選択ツールボックスは、ノードを選択した後にノード上にフローティング表示されるクイックアクションツールバーで、部分的な実行、ピン留め、削除、色の変更などの一般的なクイック操作を提供します。 ![選択ツールボックスを表示](/images/interface/setting/lite-graph/selection-toolbox.jpg) @@ -41,10 +39,10 @@ LiteGraph は ComfyUI の基盤となる图形レンダリングエンジンで ![低品質レンダリング](/images/interface/setting/lite-graph/render-mode.jpg) -### 最大 FPS +### 最大FPS - **デフォルト値**: 0（画面リフレッシュレートを使用） - **範囲**: 0 - 120 -- **機能**: キャンバスのレンダリングフレームレートを制限します。0 は画面のリフレッシュレートを使用することを意味します。FPS が高いほどキャンバスのレンダリングは滑らかになりますが、より多くのパフォーマンスを消費します。値が低すぎると、より顕著なスタッター（卡顿）が発生します。 +- **機能**: キャンバスのレンダリングフレームレートを制限します。0 は画面のリフレッシュレートを使用することを意味します。FPS が高いほどキャンバスのレンダリングは滑らかになりますが、より多くのパフォーマンスを消費します。値が低すぎると、より顕著なスタッターが発生します。 ### 常にグリッドにスナップ - **デフォルト値**: 無効 @@ -58,9 +56,9 @@ LiteGraph は ComfyUI の基盤となる图形レンダリングエンジンで - **範囲**: 1 - 500 - **機能**: 自動スナップが有効な場合、または `Shift` キーを押しながらノードを移動する場合、このパラメータはスナップ用のグリッドサイズを決定します。デフォルト値は 10 で、必要に応じて調整できます。 -### 快速ズームショートカットを有効化 (Ctrl + Shift + Drag) +### 高速ズームショートカットを有効化 (Ctrl + Shift + Drag) - **デフォルト値**: 有効 -- **機能**: `Ctrl + Shift + 左マウスボタンドラッグ` による快速ズーム機能を有効にし、より高速なズーム操作方法を提供します。 +- **機能**: `Ctrl + Shift + 左マウスボタンドラッグ` による高速ズーム機能を有効にし、より高速なズーム操作方法を提供します。 - 現在、LAN 環境では API Key ログインのみサポートされています。LAN 経由で ComfyUI サービスにアクセスしている場合は、API Key を使用してログインしてください。 + 現在、LAN 環境では APIキー ログインのみサポートされています。LAN 経由で ComfyUI サービスにアクセスしている場合は、APIキー を使用してログインしてください。 - ログインサービスにはホワイトリストがあるため、一部のサーバーにデプロイされた ComfyUI では正常にログインできない可能性があります。この場合、API Key ログインを使用して解決できます。 + ログインサービスにはホワイトリストがあるため、一部のサーバーにデプロイされた ComfyUI では正常にログインできない可能性があります。この場合、APIキー ログインを使用して解決できます。 diff --git a/ja/manager/configuration.mdx b/ja/manager/configuration.mdx index bbd946e68..834e05df0 100644 --- a/ja/manager/configuration.mdx +++ b/ja/manager/configuration.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "extra_model_paths.yaml configuration": 8df46f19 "CLI tools": 154ccf25 --- - - ## 設定パス V3.38 以降、Manager はセキュリティ強化のため保護されたシステムパスを使用します。 diff --git a/ja/manager/install.mdx b/ja/manager/install.mdx index 6662e2e1d..794c2f4eb 100644 --- a/ja/manager/install.mdx +++ b/ja/manager/install.mdx @@ -10,8 +10,6 @@ translationBlockHashes: "Manual install users": e7ec6c36 "Legacy installation methods": 0239ebc3 --- - - ## デスクトップ版ユーザー [Comfy デスクトップ版](/ja/installation/desktop/windows) を使用している場合、ComfyUI-Manager はすでに同梱されており、デフォルトで有効になっています。追加のインストールは不要です。 @@ -20,12 +18,12 @@ translationBlockHashes: [Windows ポータブル版](/ja/installation/comfyui_portable_windows) を使用しているユーザーの場合、新しい ComfyUI-Manager は ComfyUI コアに組み込まれていますが、有効化する必要があります。 -1. マネージャー機能の依存関係をインストールします： +1. マネージャーの依存関係をインストールします： ```bash .\python_embeded\python.exe -m pip install -r ComfyUI\manager_requirements.txt ``` -2. マネージャー機能を有効にして ComfyUI を起動します： +2. マネージャーを有効にして ComfyUI を起動します： ```bash .\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --enable-manager pause @@ -95,7 +93,7 @@ python main.py --enable-manager --enable-manager-legacy-ui ```bash git clone https://github.com/Comfy-Org/ComfyUI-Manager comfyui-manager ``` -3. マネージャー機能の依存関係をインストールします： +3. マネージャーの依存関係をインストールします： ```bash cd comfyui-manager pip install -r requirements.txt diff --git a/ja/manager/legacy-ui.mdx b/ja/manager/legacy-ui.mdx index 377cafc6c..5c3ddc2fe 100644 --- a/ja/manager/legacy-ui.mdx +++ b/ja/manager/legacy-ui.mdx @@ -1,15 +1,13 @@ --- title: ComfyUI-Manager を使用したカスタムノードの管理（旧版 UI） sidebarTitle: 旧版 UI -description: 旧版インターフェースを使用して ComfyUI-Manager でカスタムノードをインストール、更新、管理する方法 "中国語参照訳では画像の alt テキストにコピーペーストの誤りがあります（ステップ 3 がステップ 2 と同じ、ステップ 6 がステップ 5 と同じ）。また、URL に `/zh/` プレフィックスが含まれていますが、英語ソースに従い元の URL を維持します。" +description: 旧版インターフェースを使用して ComfyUI-Manager でカスタムノードをインストール、更新、管理する方法 translationSourceHash: 665e3090 translationFrom: manager/legacy-ui.mdx translationBlockHashes: "ComfyUI custom node management": 036f30fa "Snapshot manager": 5e681705 --- - - ## ComfyUI カスタムノード管理 Manager を経由してカスタムノードをインストールする方法を学びます diff --git a/ja/manager/overview.mdx b/ja/manager/overview.mdx index 8e84c939e..cf664a1a7 100644 --- a/ja/manager/overview.mdx +++ b/ja/manager/overview.mdx @@ -1,7 +1,7 @@ --- title: 概要 sidebarTitle: 概要 -description: ComfyUI 内のカスタムノード、モデルなどを管理するための拡張機能 +description: ComfyUI内のカスタムノード、モデルなどを管理するための拡張機能 translationSourceHash: d9b5c989 translationFrom: manager/overview.mdx --- @@ -15,7 +15,7 @@ translationFrom: manager/overview.mdx - **カスタムノード管理**: カスタムノードのインストール、更新、削除、無効化 - **モデル管理**: さまざまなソースからモデルをダウンロードおよび管理 - **スナップショット管理**: インストール状態の保存と復元 -- **欠落ノード検出**: ワークフローから欠落したノードを自動的に検出およびインストール +- **不足ノード検出**: ワークフローから不足しているノードを自動的に検出およびインストール ## 次のステップ @@ -26,8 +26,8 @@ translationFrom: manager/overview.mdx 新しい UI でカスタムノードとモデルを管理します - - 従来の UI でカスタムノードとモデルを管理します + + レガシー UI でカスタムノードとモデルを管理します 設定とオプションを構成します diff --git a/ja/manager/troubleshooting.mdx b/ja/manager/troubleshooting.mdx index a15be9951..4381bfbc6 100644 --- a/ja/manager/troubleshooting.mdx +++ b/ja/manager/troubleshooting.mdx @@ -12,8 +12,6 @@ translationBlockHashes: "Security policy": b07e2c4f "Getting help": f37e18a2 --- - - ## 一般的な問題 ### カスタム git 実行ファイルパス diff --git a/ja/registry/cd.mdx b/ja/registry/cd.mdx index d857ef3d3..d165ad2e1 100644 --- a/ja/registry/cd.mdx +++ b/ja/registry/cd.mdx @@ -4,6 +4,7 @@ translationSourceHash: bde264d5 translationFrom: registry/cd.mdx --- + ### バージョン公開を CI/CD パイプラインの一部として組み込む カスタムノードのバージョンを公開するための [Github Action](https://github.com/Comfy-Org/node-publish-action) を作成しました。このアクションは、カスタムノードの新しいバージョンを Comfy Registry に公開し、開発者のワークフローに応じて設定できます。 @@ -29,3 +30,5 @@ translationFrom: registry/cd.mdx #### 手動での公開 ```yaml + +``` diff --git a/ja/registry/claim-my-node.mdx b/ja/registry/claim-my-node.mdx index c2c0fce61..4ccaad895 100644 --- a/ja/registry/claim-my-node.mdx +++ b/ja/registry/claim-my-node.mdx @@ -8,15 +8,13 @@ translationBlockHashes: "Getting Started": 4857658b "Frequently Asked Questions": 19812fbe --- - - ## 概要 -**ノードの所有権を主張する**機能により、開発者は ComfyUI Registry においてカスタムノードの所有権を主張できます。このシステムは、正当な作者のみが公開したノードを管理および更新できるようにし、コミュニティ内のセキュリティと説明責任を確保します。 +**Claim My Node**機能により、開発者は ComfyUI Registry においてカスタムノードの所有権を主張できます。このシステムは、正当な作者のみが公開したノードを管理および更新できるようにし、コミュニティ内のセキュリティと説明責任を確保します。 ## 所有者未設定のノードとは -新しい標準を備えた ComfyUI Manager から Comfy Registry へ移行するに伴い、以前 ComfyUI Manager レガシーシステムに掲載されていた多くのカスタムノードが、Registry 上で「所有者未設定」として表示されるようになりました。これらのノードは以下の特徴を持ちます： +新たな基準に基づいて ComfyUI Manager から Comfy Registry へ移行するに伴い、以前 ComfyUI Manager レガシーシステムに掲載されていた多くのカスタムノードが、Registry 上で「所有者未設定」として表示されるようになりました。これらのノードは： - 元々 ComfyUI Manager レガシーシステムで公開されていたノード - 最新基準を満たすために Comfy Registry へ移行されたノード @@ -24,7 +22,7 @@ translationBlockHashes: 開発者がこれらの移行されたノードを容易に主張できるよう、レガシーシステムから新しい Registry 標準への円滑な移行を保証しつつ、適切な所有権と管理を維持するための手段を提供しています。 -## 入門ガイド +## はじめに ノードの所有権を主張するには、以下の手順に従ってください： diff --git a/ja/registry/overview.mdx b/ja/registry/overview.mdx index 59ad6bbad..9257f0bf2 100644 --- a/ja/registry/overview.mdx +++ b/ja/registry/overview.mdx @@ -8,8 +8,6 @@ translationBlockHashes: "Publishing Nodes": 278ff69e "Frequently Asked Questions": f5ffda52 --- - - ## はじめに レジストリは、[ComfyUI-Manager](https://github.com/Comfy-Org/ComfyUI-Manager) を動かすカスタムノードの公開コレクションです。開発者は、カスタムノードの公開、バージョン管理、非推奨化、および関連指標の追跡を行うことができます。ComfyUI ユーザーは、ComfyUI-Manager を介してレジストリからカスタムノードを発見、インストール、評価できます。 diff --git a/ja/registry/publishing.mdx b/ja/registry/publishing.mdx index 5676efea2..d19d1a763 100644 --- a/ja/registry/publishing.mdx +++ b/ja/registry/publishing.mdx @@ -6,9 +6,6 @@ translationBlockHashes: "Set up a Registry Account": be894a7a "Publish to the Registry": b39c0b89 --- - - - ## レジストリアカウントの設定 以下の手順に従って、レジストリアカウントを設定し、最初のノードを公開してください。 @@ -34,7 +31,7 @@ translationBlockHashes: ### レジストリ公開用 API キーの作成 - **重要:** この API キーは、**カスタムノードをレジストリおよび ComfyUI-Manager に公開するため**のものです。ワークフローで有料の API ノードを使用したい場合は、[API ノードの概要](/ja/tutorials/partner-nodes/overview) を参照してください。 + **重要:** この API キーは、**カスタムノードをレジストリおよび ComfyUI-Manager に公開するため**のものです。ワークフローで有料のパートナーノードを使用したい場合は、[パートナーノードの概要](/ja/tutorials/partner-nodes/overview) を参照してください。 [こちら](https://registry.comfy.org/nodes) にアクセスし、API キーを作成したいパブリッシャーをクリックしてください。このキーは、CLI または GitHub Actions を介してカスタムノードをレジストリ（ComfyUI-Manager を動かすもの）に公開するために使用されます。 @@ -89,8 +86,8 @@ tests/ -- `.comfyignore` は `.gitignore` の上に重ねて適用されます — git で追跡されていないファイルはすでに除外されています。git に *残しておきたい* が、レジストリアーカイブには含めたくないファイルを除外するために `.comfyignore` を使用してください。 -- [`[tool.comfy].includes`](/registry/specifications) に列挙されたディレクトリは、`.comfyignore` のパターンに一致しても強制的に含まれます（ビルドされた `dist/` フォルダーが `.gitignore` で無視されていても配布する必要がある場合などに便利です）。 +- `.comfyignore` は `.gitignore` の上に重ねて適用されます: git で追跡されていないファイルはすでに除外されています。git に *残しておきたい* が、レジストリアーカイブには含めたくないファイルを除外するために `.comfyignore` を使用してください。 +- [`[tool.comfy].includes`](/ja/registry/specifications) に列挙されたディレクトリは、`.comfyignore` のパターンに一致しても強制的に含まれます（ビルドされた `dist/` フォルダーが `.gitignore` で無視されていても配布する必要がある場合などに便利です）。 - ローカル実行と CI が同じアーカイブを生成するように、`.comfyignore` をリポジトリにコミットしてください。 diff --git a/ja/registry/specifications.mdx b/ja/registry/specifications.mdx index 98247f901..3d004f10b 100644 --- a/ja/registry/specifications.mdx +++ b/ja/registry/specifications.mdx @@ -9,7 +9,6 @@ translationBlockHashes: "Complete Example": 24cf243a --- - # 仕様 `pyproject.toml` ファイルには、ComfyUI カスタムノード用の 2 つの主要セクション `[project]` と `[tool.comfy]` が含まれています。以下に各セクションの仕様を示します。 @@ -214,7 +213,7 @@ requires-comfyui = ">0.1.3,<1.0.0" # 0.1.3 より大きく 1.0.0 より小さ includes = ['dist'] ``` -公開アーカイブからファイルを除外する（`includes` の逆）には、公開ページの [`.comfyignore`](/ja/registry/publishing) を参照してください。 +公開アーカイブからファイルを除外する（`includes` の逆）には、公開ページの [`.comfyignore`](/ja/registry/publishing#exclude-files-with-comfyignore) を参照してください。 ## 完全な例 ```toml @@ -246,3 +245,4 @@ DisplayName = "Super Resolution Node" Icon = "https://raw.githubusercontent.com/username/super-resolution-node/main/icon.png" Banner = "https://raw.githubusercontent.com/username/super-resolution-node/main/banner.png" requires-comfyui = ">=1.0.0" # ComfyUI バージョン互換性 +``` diff --git a/ja/registry/standards.mdx b/ja/registry/standards.mdx index a25c579b5..1021f7b66 100644 --- a/ja/registry/standards.mdx +++ b/ja/registry/standards.mdx @@ -27,7 +27,7 @@ translationFrom: registry/standards.mdx すべての適用される法律および規制を遵守しなければなりません ### 5. 品質要件 -ノードは完全に機能し、文書化されており、積極的にメンテナンスされている必要があります。 +ノードは完全に機能し、十分に文書化されており、積極的にメンテナンスされている必要があります。 ### 6. フォークのガイドライン フォークされたノードは以下を満たす必要があります： @@ -38,7 +38,7 @@ translationFrom: registry/standards.mdx ## セキュリティ基準 -カスタムノードは安全である必要があります。これらの基準に違反するカスタムノードについては、書き換えのために対処を開始します。コアによって公開されるべき主要な機能がある場合は、[rfcs repo](https://github.com/comfy-org/rfcs) でリクエストしてください。 +カスタムノードは安全である必要があります。これらの基準に違反するカスタムノードについては、書き換えのために取り組みを開始します。コアによって公開されるべき主要な機能がある場合は、[rfcs repo](https://github.com/comfy-org/rfcs) でリクエストしてください。 ### eval/exec 呼び出し @@ -57,7 +57,7 @@ translationFrom: registry/standards.mdx subprocess 呼び出しを通じたランタイムパッケージのインストールは許可されていません。 #### 理由 -- ComfyUI Manager は ComfyUI に同梱され、ユーザーが依存関係をインストールできるようにします +- 最初の項目: ComfyUI Manager は ComfyUI に同梱され、ユーザーが依存関係をインストールできるようにします - 集中化された依存関係管理は、セキュリティとユーザーエクスペリエンスを向上させます - 潜在的なサプライチェーン攻撃を防ぐのに役立ちます - ComfyUI の複数の再読み込みの必要性を排除します diff --git a/ja/specs/nodedef_json.mdx b/ja/specs/nodedef_json.mdx index 6ea08ccbf..cd795de53 100644 --- a/ja/specs/nodedef_json.mdx +++ b/ja/specs/nodedef_json.mdx @@ -5,6 +5,7 @@ translationSourceHash: 05aa4856 translationFrom: specs/nodedef_json.mdx --- + ノード定義 JSON は [JSON Schema](https://json-schema.org/) を使用して定義されます。このスキーマの変更は [rfcs リポジトリ](https://github.com/comfy-org/rfcs) で議論されます。 ## v2.0 (最新) @@ -535,4 +536,5 @@ translationFrom: specs/nodedef_json.mdx } }, "$schema": "http://json-schema.org/draft-07/schema#" -} \ No newline at end of file +} +``` diff --git a/ja/specs/nodedef_json_1_0.mdx b/ja/specs/nodedef_json_1_0.mdx index 96b668ea7..eb9eb976a 100644 --- a/ja/specs/nodedef_json_1_0.mdx +++ b/ja/specs/nodedef_json_1_0.mdx @@ -5,6 +5,7 @@ translationSourceHash: d08bfe2a translationFrom: specs/nodedef_json_1_0.mdx --- + ## v1.0 ```json Node Definition v1.0 @@ -1316,4 +1317,5 @@ translationFrom: specs/nodedef_json_1_0.mdx } }, "$schema": "http://json-schema.org/draft-07/schema#" -} \ No newline at end of file +} +``` diff --git a/ja/specs/workflow_json.mdx b/ja/specs/workflow_json.mdx index b4c304cf9..be98a7f0a 100644 --- a/ja/specs/workflow_json.mdx +++ b/ja/specs/workflow_json.mdx @@ -8,8 +8,6 @@ translationBlockHashes: "Version 1.0 (Latest)": f3881a59 "Older versions": fd8590ae --- - - ワークフロー JSON は [JSON Schema](https://json-schema.org/) を使用して定義されています。このスキーマの変更については [rfcs リポジトリ](https://github.com/comfy-org/rfcs) で議論されます。 ## バージョン 1.0（最新） diff --git a/ja/specs/workflow_json_0.4.mdx b/ja/specs/workflow_json_0.4.mdx index 20bd45bf0..d4fd1b3e5 100644 --- a/ja/specs/workflow_json_0.4.mdx +++ b/ja/specs/workflow_json_0.4.mdx @@ -5,6 +5,7 @@ translationSourceHash: dd7a23ce translationFrom: specs/workflow_json_0.4.mdx --- + ## v0.4 ```json @@ -651,4 +652,5 @@ translationFrom: specs/workflow_json_0.4.mdx } }, "$schema": "http://json-schema.org/draft-07/schema#" -} \ No newline at end of file +} +``` diff --git a/ja/support/contact-support.mdx b/ja/support/contact-support.mdx index 21dc9b5e0..32fa875d0 100644 --- a/ja/support/contact-support.mdx +++ b/ja/support/contact-support.mdx @@ -1,5 +1,5 @@ --- -title: サポートへのお問い合わせ +title: サポートに連絡 description: ComfyUI のヘルプを取得し、問題を報告できます。 translationSourceHash: 54844622 translationFrom: support/contact-support.mdx diff --git a/ja/support/data-retention.mdx b/ja/support/data-retention.mdx index 71fd585f5..fa3c45f3a 100644 --- a/ja/support/data-retention.mdx +++ b/ja/support/data-retention.mdx @@ -21,4 +21,4 @@ translationFrom: support/data-retention.mdx アカウントまたは関連データの削除をご希望の場合は、[support@comfy.org](mailto:support@comfy.org) までお問い合わせください。法的義務、セキュリティ、監査、または不正防止のため、一部の記録が保持される場合があります。 - + \ No newline at end of file diff --git a/ja/support/payment/accepted-payment-methods.mdx b/ja/support/payment/accepted-payment-methods.mdx index 0c2770e69..e97c12b89 100644 --- a/ja/support/payment/accepted-payment-methods.mdx +++ b/ja/support/payment/accepted-payment-methods.mdx @@ -36,7 +36,7 @@ WeChat Pay および Alipay は特定の地域でのみ利用可能です。お - [Stripe Alipay ドキュメント](https://docs.stripe.com/payments/alipay) -![Checkout screen showing USD selected and Alipay/WeChat Pay options available](/images/support/payment/active-wechat-and-alipay.jpg) +![USD が選択され、Alipay と WeChat Pay オプションが利用可能なチェックアウト画面](/images/support/payment/active-wechat-and-alipay.jpg) *チェックアウト時に支払い通貨として **USD** を選択し (1)、Alipay および WeChat Pay のオプション (2) を利用可能にしてください。* @@ -45,7 +45,7 @@ Alipay および WeChat Pay は、Comfy Cloud サブスクリプションでは これらのオプションを有効にするには、Stripe 請求ポータルの通貨選択で **USD** を選択してください。USD が選択されると、支払い方法リストに Alipay および WeChat Pay が表示されます。 -![Stripe payment form showing USD currency selection and Alipay/WeChat options](/images/support/payment/active-wechat-and-alipay.jpg) +![USD 通貨選択と Alipay/WeChat オプションが表示された Stripe 支払いフォーム](/images/support/payment/active-wechat-and-alipay.jpg) ## 要件 diff --git a/ja/support/payment/payment-history.mdx b/ja/support/payment/payment-history.mdx index d54232d60..1fd81071f 100644 --- a/ja/support/payment/payment-history.mdx +++ b/ja/support/payment/payment-history.mdx @@ -13,23 +13,23 @@ translationFrom: support/payment/payment-history.mdx 1. ComfyUI アカウントにログインし、プロフィールメニューを開きます 2. **ユーザー設定**をクリックします 3. 設定パネルで**クレジット**タブを選択します -4. **請求書履歴**をクリックして Stripe 請求ポータルを開き、支払い履歴を表示します +4. **請求履歴**をクリックして Stripe 請求ポータルを開き、支払い履歴を表示します ### 視覚的な手順 ![ユーザー設定オプションが強調表示されたプロフィールメニュー](/images/support/payment/billing-1-account.png) *請求管理のためにプロフィールメニューを開き、**ユーザー設定**を選択します。* -![請求書履歴ボタンが強調表示された設定内のクレジットタブ](/images/support/payment/billing-2-credits.png) -***クレジット**タブに移動し、**請求書履歴**を選択します。* +![請求履歴ボタンが強調表示された設定内のクレジットタブ](/images/support/payment/billing-2-credits.png) +***クレジット**タブに移動し、**請求履歴**を選択します。* -![請求書履歴テーブルを表示する Stripe 請求ポータル](/images/support/payment/billing-4-history.jpg) +![請求履歴テーブルを表示する Stripe 請求ポータル](/images/support/payment/billing-4-history.jpg) *Stripe 請求ポータルで請求書リストを確認します。* ## 請求書のダウンロード -Stripe ポータルで、請求書履歴リストの任意の項目を開いて詳細を表示し、PDF コピーをダウンロードします。 +Stripe ポータルで、請求履歴リストの任意の項目を開いて詳細を表示し、PDF コピーをダウンロードします。 -![ダウンロードオプションが強調表示された請求書履歴の詳細](/images/support/payment/billing-5-download-invoice.jpg) +![ダウンロードオプションが強調表示された請求履歴の詳細](/images/support/payment/billing-5-download-invoice.jpg) -新しい請求書が生成される前に、請求先情報が正しいことを確認してください。[支払い情報の編集](/support/payment/editing-payment-information) からこれらを更新できます。 \ No newline at end of file +新しい請求書が生成される前に、請求先情報が正しいことを確認してください。[支払い情報の編集](/ja/support/payment/editing-payment-information) からこれらを更新できます。 \ No newline at end of file diff --git a/ja/support/payment/unsuccessful-payments.mdx b/ja/support/payment/unsuccessful-payments.mdx index f054d9248..215ccf0d2 100644 --- a/ja/support/payment/unsuccessful-payments.mdx +++ b/ja/support/payment/unsuccessful-payments.mdx @@ -70,7 +70,7 @@ translationFrom: support/payment/unsuccessful-payments.mdx - 請求日の前にカードに十分な資金があることを確認してください - 有効期限が切れる前にカードの詳細を更新してください - 銀行で Comfy Organization Inc からの請求をホワイトリストに登録してください -- 請求リマインダーの電子メール通知を有効にしてください +- 請求リマインダーのメール通知を有効にしてください ## 猶予期間とアカウントの停止 diff --git a/ja/support/subscription/changing-plan.mdx b/ja/support/subscription/changing-plan.mdx index c66468edd..2b9c74341 100644 --- a/ja/support/subscription/changing-plan.mdx +++ b/ja/support/subscription/changing-plan.mdx @@ -1,6 +1,6 @@ --- title: 購読プランの変更 -sidebarTitle: 購読プランの変更 +sidebarTitle: 購読プランを変更 description: Comfy Cloud 購読プランの変更方法について学びます translationSourceHash: 1e4ea375 translationFrom: support/subscription/changing-plan.mdx diff --git a/ja/support/subscription/managing.mdx b/ja/support/subscription/managing.mdx index 36b763bc9..45dc82db7 100644 --- a/ja/support/subscription/managing.mdx +++ b/ja/support/subscription/managing.mdx @@ -12,10 +12,10 @@ Comfy Cloud サブスクリプションを管理するには: 1. [cloud.comfy.org](https://cloud.comfy.org/?utm_source=docs) にアクセスしてログインします 2. 設定メニューに移動します -3. **Plan & Credits** に移動します +3. **プランとクレジット** に移動します サブスクリプション設定メニュー -4. **Manage subscription** をクリックします +4. **サブスクリプションを管理** をクリックします これにより Stripe に移動し、そこでサブスクリプションの詳細を管理できます。 \ No newline at end of file diff --git a/ja/troubleshooting/custom-node-issues.mdx b/ja/troubleshooting/custom-node-issues.mdx index 6f7568407..c3e4cba78 100644 --- a/ja/troubleshooting/custom-node-issues.mdx +++ b/ja/troubleshooting/custom-node-issues.mdx @@ -16,6 +16,7 @@ translationBlockHashes: --- + カスタムノードの問題をトラブルシューティングするための全体のアプローチは以下の通りです： ```mermaid @@ -53,7 +54,7 @@ python main.py --disable-all-custom-nodes - ポータブル版があるフォルダを開き、`run_nvidia_gpu.bat` または `run_cpu.bat` ファイルを見つけます + ポータブル版があるフォルダーを開き、`run_nvidia_gpu.bat` または `run_cpu.bat` ファイルを見つけます ![.bat ファイルの修正](/images/troubleshooting/Portable-disable-custom-nodes.jpg) 1. `run_nvidia_gpu.bat` または `run_cpu.bat` ファイルをコピーし、`run_nvidia_gpu_disable_custom_nodes.bat` に名前を変更します 2. コピーしたファイルをメモ帳で開きます @@ -67,10 +68,10 @@ python main.py --disable-all-custom-nodes ![ComfyUI troubleshooting](/images/troubleshooting/portable-disable-custom-nodes-cml-1.jpg) - 1. ポータブル版があるフォルダに入ります + 1. ポータブル版があるフォルダーに入ります 2. 右クリックメニュー → ターミナルを開く でターミナルを開きます ![ComfyUI troubleshooting](/images/troubleshooting/portable-disable-custom-nodes-cml-2.jpg) - 3. フォルダ名がポータブル版の現在のディレクトリであることを確認します + 3. フォルダー名がポータブル版の現在のディレクトリであることを確認します 4. 以下のコマンドを入力して、ポータブル版の python を経由して ComfyUI を起動し、カスタムノードを無効にします ``` .\python_embeded\python.exe -s ComfyUI\main.py --disable-all-custom-nodes @@ -108,6 +109,7 @@ flowchart TD K --> B L --> M[終了] ``` + ## 2 つのトラブルシューティング方法 このドキュメントでは、トラブルシューティングのためにカスタムノードを 2 つのタイプに分類します： @@ -116,7 +118,7 @@ flowchart TD - A: フロントエンド拡張機能を持つカスタムノード - B: 通常のカスタムノード -まず、不同类型的なカスタムノードで発生しうる問題と原因を理解しましょう： +まず、さまざまなタイプのカスタムノードで発生しうる問題と原因を理解しましょう： @@ -154,26 +156,26 @@ flowchart TD -## トラブルシューティングのための二分法的使用 +## トラブルシューティングに二分探索を使用する -これら 2 つの不同类型的なカスタムノードの問題の中で、カスタムノードのフロントエンド拡張機能と ComfyUI の間の競合がより一般的です。これらのノードを優先的にトラブルシューティングします。以下が全体のトラブルシューティングアプローチです： +これら2つの異なるタイプのカスタムノードの問題の中で、カスタムノードのフロントエンド拡張機能とComfyUIの間の競合がより一般的です。これらのノードを優先的にトラブルシューティングします。以下が全体のトラブルシューティングアプローチです： ### 1. カスタムノードのフロントエンド拡張機能のトラブルシューティング - ![すべてのプラグインのフロントエンド拡張機能を無効](/images/troubleshooting/disable_3rd_party.jpg) - ComfyUI を起動した後、設定の `Extensions` メニューを見つけ、画像に示されている手順に従ってすべてのサードパーティ製拡張機能を無効にします + ![すべてのプラグインフロントエンド拡張機能を無効](/images/troubleshooting/disable_3rd_party.jpg) + ComfyUIを起動した後、設定の`Extensions`メニューを見つけ、画像に示されている手順に従ってすべてのサードパーティ製拡張機能を無効にします - ComfyUI フロントエンドに入れない場合は、フロントエンド拡張機能のトラブルシューティングセクションをスキップし、[一般的なカスタムノードのトラブルシューティングアプローチ](#2-general-custom-node-troubleshooting) に進んでください + ComfyUIフロントエンドに入れない場合は、フロントエンド拡張機能のトラブルシューティングセクションをスキップし、[一般的なカスタムノードのトラブルシューティングアプローチ](#2-general-custom-node-troubleshooting-approach)に進んでください - - 最初にフロントエンド拡張機能を無効にした後、すべてのフロントエンド拡張機能が適切に無効化されていることを確認するために、ComfyUI を再起動することをお勧めします - - 問題が消えた場合、それはカスタムノードのフロントエンド拡張機能が原因でした。二分法トラブルシューティングを進めることができます + + 最初にフロントエンド拡張機能を無効にした後、すべてのフロントエンド拡張機能が適切に無効化されていることを確認するために、ComfyUIを再起動することを推奨します + - 問題が消えた場合、それはカスタムノードのフロントエンド拡張機能が原因でした。二分探索によるトラブルシューティングを進めることができます - 問題が継続する場合、それはフロントエンド拡張機能が原因ではありません。このドキュメントの他のトラブルシューティングアプローチを参照してください - + このドキュメントの冒頭で述べた方法を使用してトラブルシューティングを行い、問題のあるノードを見つけるまで、一度にカスタムノードの半分を有効にします ![フロントエンド拡張機能を有効](/images/troubleshooting/enable_extensions.jpg) 画像を参照して、フロントエンド拡張機能の半分を有効にします。拡張機能の名前が似ている場合は、それらが同じカスタムノードのフロントエンド拡張機能から来ている可能性が高いことに注意してください @@ -183,39 +185,37 @@ flowchart TD -この方法を使用すると、ComfyUI を複数回再起動する必要はありません。カスタムノードのフロントエンド拡張機能を有効/無効にした後に ComfyUI を再読み込みするだけです。さらに、トラブルシューティングの範囲はフロントエンド拡張機能を持つノードに限定されるため、検索範囲が大幅に狭まります。 +この方法を使用すると、ComfyUIを複数回再起動する必要はありません。カスタムノードのフロントエンド拡張機能を有効/無効にした後にComfyUIを再読み込みするだけです。さらに、トラブルシューティングの範囲はフロントエンド拡張機能を持つノードに限定されるため、検索範囲が大幅に狭まります。 ### 2. 一般的なカスタムノードのトラブルシューティング - - 二分法ローカリゼーション方法については、手動検索に加えて、comfy-cli を使用した自動化された二分法もあります。詳細は以下の通りです： + + 二分探索による特定方法については、手動検索に加えて、comfy-cliを使用した自動化された二分探索もあります。詳細は以下の通りです： - - - Comfy CLI の使用には、ある程度のコマンドラインの経験が必要です。慣れていない場合は、手動の二分法を使用してください。 - - [Comfy CLI](/ja/comfy-cli/getting-started) がインストールされている場合、自動化された bisect ツールを使用して問題のあるノードを見つけることができます： + +Comfy CLI の使用には、ある程度のコマンドラインの経験が必要です。慣れていない場合は、手動の二分法を使用してください。 - ```bash - # bisect セッションを開始 - comfy-cli node bisect start +[Comfy CLI](/ja/comfy-cli/getting-started) がインストールされている場合、自動化された bisect ツールを使用して問題のあるノードを見つけることができます： - # プロンプトに従います： - # - 現在有効になっているノードセットで ComfyUI をテスト - # - 問題が消えた場合は 'good' とマーク：comfy-cli node bisect good - # - 問題が継続する場合は 'bad' とマーク：comfy-cli node bisect bad - # - 問題のあるノードが特定されるまで繰り返す +```bash +# bisect セッションを開始 +comfy-cli node bisect start - # 完了したらリセット - comfy-cli node bisect reset - ``` +# プロンプトに従います： +# - 現在有効になっているノードセットで ComfyUI をテスト +# - 問題が消えた場合は 'good' とマーク：comfy-cli node bisect good +# - 問題が継続する場合は 'bad' とマーク：comfy-cli node bisect bad +# - 問題のあるノードが特定されるまで繰り返す - bisect ツールは自動的にノードを有効/無効にし、プロセスをガイドします。 +# 完了したらリセット +comfy-cli node bisect reset +``` - - +bisect ツールは自動的にノードを有効/無効にし、プロセスをガイドします。 + + 開始する前に、問題が発生した場合に備えて、custom_nodes フォルダの**バックアップを作成**してください。 @@ -240,20 +240,18 @@ flowchart TD # すべてのコンテンツをバックアップ xcopy "custom_nodes\*" "%USERPROFILE%\custom_nodes_backup\" /E /H /Y ``` - - - custom_nodes フォルダを手動でバックアップします - または、以下のコマンドラインを使用してバックアップします： - ```bash - # バックアップと一時フォルダを作成 - mkdir ~/custom_nodes_backup - mkdir ~/custom_nodes_temp + +custom_nodes フォルダを手動でバックアップします +または、以下のコマンドラインを使用してバックアップします： +```bash +# バックアップと一時フォルダを作成 +mkdir ~/custom_nodes_backup +mkdir ~/custom_nodes_temp - # すべてのコンテンツをバックアップ - cp -r custom_nodes/* ~/custom_nodes_backup/ - ``` - - +# すべてのコンテンツをバックアップ +cp -r custom_nodes/* ~/custom_nodes_backup/ +``` + ```bash # バックアップと一時フォルダを作成 mkdir /content/custom_nodes_backup @@ -389,3 +387,4 @@ flowchart TD 一般的なインストール、モデル、またはパフォーマンスの問題については、[トラブルシューティングの概要](/ja/troubleshooting/overview) および [モデルの問題](/ja/troubleshooting/model-issues) ページを参照してください。 +``` diff --git a/ja/troubleshooting/model-issues.mdx b/ja/troubleshooting/model-issues.mdx index 51fcc4779..f11d5af8d 100644 --- a/ja/troubleshooting/model-issues.mdx +++ b/ja/troubleshooting/model-issues.mdx @@ -11,8 +11,6 @@ translationBlockHashes: "Model Loading Errors": dd723562 "Model Performance Issues": 0d78150a --- - - ## モデルアーキテクチャの不一致 **症状：** 生成過程中にテンソル次元エラーが発生する、特に VAE デコード段階 diff --git a/ja/troubleshooting/overview.mdx b/ja/troubleshooting/overview.mdx index 743d95782..b9b683acb 100644 --- a/ja/troubleshooting/overview.mdx +++ b/ja/troubleshooting/overview.mdx @@ -15,8 +15,6 @@ translationBlockHashes: "Getting Help & Reporting Bugs": 81e1016e "Community Resources": 1f81eee3 --- - - 多くのフィードバックを受けており、提出される問題のほとんどがカスタムノードに関連していることがわかります。そのため、エラーレポートを提出する前に、問題が ComfyUI コアの問題によって引き起こされていないことを確認するために、[カスタムノードのトラブルシューティングガイド](/ja/troubleshooting/custom-node-issues) を必ずお読みください。 @@ -227,7 +225,7 @@ python -c "import torch; print('CUDA available:', torch.cuda.is_available()); pr ## ネットワークと API の問題 -### API ノードが機能しない +### パートナーノードが機能しない **症状:** API 呼び出しが失敗する、タイムアウトエラー、クォータ超過 @@ -373,7 +371,7 @@ pip install torch==2.3.1.post0+cxx11.abi torchvision==0.18.1.post0+cxx11.abi tor 1. **既知の問題かどうかを確認:** - [GitHub Issues](https://github.com/Comfy-Org/ComfyUI/issues) を検索 - [ComfyUI Forum](https://forum.comfy.org/) を確認 - - [Discord 議論](https://discord.com/invite/comfyorg) をレビュー + - [Discord ディスカッション](https://discord.com/invite/comfyorg) を確認 2. **基本的なトラブルシューティングを試す:** - [デフォルトワークフロー](/ja/get_started/first_generation) でテスト @@ -404,27 +402,27 @@ pip install torch==2.3.1.post0+cxx11.abi torchvision==0.18.1.post0+cxx11.abi tor -**システム情報（設定の关于ページで見つかります）:** +**システム情報（設定のアバウトページで確認できます）:** - オペレーティングシステム（Windows 11、macOS 14.1、Ubuntu 22.04 など） -- ComfyUI バージョン（設定の关于ページを確認） +- ComfyUI バージョン（設定のアバウトページを確認） - Python バージョン：`python --version` - PyTorch バージョン：`python -c "import torch; print(torch.__version__)"` - GPU モデルとドライバーバージョン - インストール方法（デスクトップ、ポータブル、手動、comfy-cli） -![設定の关于ページ](/images/troubleshooting/menu-about.jpg) +![設定のアバウトページ](/images/troubleshooting/menu-about.jpg) ```bash - # システム情報 + # System info systeminfo | findstr /C:"OS Name" /C:"OS Version" - # GPU 情報 + # GPU info wmic path win32_VideoController get name - # Python と PyTorch 情報 + # Python & PyTorch info python --version python -c "import torch; print(f'PyTorch: {torch.__version__}')" python -c "import torch; print(f'CUDA Available: {torch.cuda.is_available()}')" @@ -432,13 +430,13 @@ pip install torch==2.3.1.post0+cxx11.abi torchvision==0.18.1.post0+cxx11.abi tor ```bash - # システム情報 + # System info uname -a - # GPU 情報（Linux） + # GPU info (Linux) lspci | grep VGA - # Python と PyTorch 情報 + # Python & PyTorch info python --version python -c "import torch; print(f'PyTorch: {torch.__version__}')" python -c "import torch; print(f'CUDA Available: {torch.cuda.is_available()}')" diff --git a/ja/tutorials/3d/hunyuan3D-2.mdx b/ja/tutorials/3d/hunyuan3D-2.mdx index 35f163849..eb2d87b7e 100644 --- a/ja/tutorials/3d/hunyuan3D-2.mdx +++ b/ja/tutorials/3d/hunyuan3D-2.mdx @@ -2,18 +2,16 @@ title: "ComfyUI Hunyuan3D-2 例" description: "このガイドでは、ComfyUI で Hunyuan3D-2 を使用して 3D アセットを生成する方法を説明します。" sidebarTitle: "Hunyuan3D-2" -translationSourceHash: 61db1e41 +translationSourceHash: 10c6c4b2 translationFrom: tutorials/3d/hunyuan3D-2.mdx translationBlockHashes: - "_intro": 0078e687 - "ComfyUI Hunyuan3D-2mv Workflow Example": 1a1a8896 - "Hunyuan3D-2mv-turbo Workflow": afdb6432 - "Hunyuan3D-2 Single View Workflow": 1f454a64 + "_intro": 2fbea77b + "ComfyUI Hunyuan3D-2mv Workflow Example": 81e3ea04 + "Hunyuan3D-2mv-turbo Workflow": 15679e8b + "Hunyuan3D-2 Single View Workflow": 9384e8d2 "Community Resources": d12dc58a "Hunyuan3D 2.0 Open-Source Model Series": 363ff037 --- - - import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' # Hunyuan3D 2.0 はじめに @@ -57,9 +55,39 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して - -

Run on Comfy Cloud

- +### HY 3D 2.0 MV (`3d_hunyuan3d_multiview_to_model`) + +Hunyuan3D 2.0 MV を使用して、複数のビューから 3D モデルを生成します。 + +HY 3D 2.0 MV ワークフロープレビュー + + + + このワークフローを Comfy Cloud ですぐに実行 + + + テンプレートライブラリで JSON をダウンロード、または "HY 3D 2.0 MV" を検索 + + + +**入力画像** + +以下のファイルを対応する `LoadImage` ノードにアップロードします： + + + + `LoadImage` ノード 56 · 正面ビュー + + + `LoadImage` ノード 78 · 左ビュー + + + `LoadImage` ノード 80 · 背面ビュー + + + `LoadImage` ノード 87 · 右ビュー + + ### 1. ワークフロー @@ -72,29 +100,28 @@ Hunyuan3D-2mv ワークフローでは、マルチビュー画像を使用して ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/left.png) ![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_elf/back.png) - この例では、入力画像はすでに余分な背景を除去するように前処理されています。実際の使用では、[ComfyUI_essentials](https://github.com/cubiq/ComfyUI_essentials) のようなカスタムノードを使用して、余分な背景を自動的に除去できます。 ### 2. 手動モデルインストール -以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください +以下のモデルをダウンロードし、対応する ComfyUI フォルダーに保存してください -- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv.safetensors` に名前を変更できます ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv.safetensors // リネーム後のファイル +│ │ └── hunyuan3d-dit-v2-mv.safetensors // 名前を変更したファイル ``` ### 3. ワークフローの実行手順 ![ComfyUI hunyuan3d_2mv](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv.jpg) -1. Image Only Checkpoint Loader(img2vid model) が、ダウンロードしてリネームした `hunyuan3d-dit-v2-mv.safetensors` モデルを読み込んでいることを確認します +1. Image Only Checkpoint Loader(img2vid model) が、ダウンロードして名前を変更した `hunyuan3d-dit-v2-mv.safetensors` モデルを読み込んでいることを確認します 2. 各 `Load Image` ノードに対応するビュー画像を読み込みます 3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します @@ -104,81 +131,130 @@ ComfyUI/ Hunyuan3D-2mv-turbo ワークフローでは、Hunyuan3D-2mv-turbo モデルを使用して 3D モデルを生成します。このモデルは Hunyuan3D-2mv のステップ蒸留バージョンで、より高速な 3D モデル生成を可能にします。このバージョンのワークフローでは、`cfg` を 1.0 に設定し、`flux guidance` ノードを追加して `distilled cfg` 生成を制御します。 - -

Run on Comfy Cloud

- +### HY 3D 2.0 MV Turbo（`3d_hunyuan3d_multiview_to_model_turbo`） + +複数のビューから Hunyuan3D 2.0 MV Turbo を使用して 3D モデルを生成します。 + +HY 3D 2.0 MV Turbo ワークフローのプレビュー + + + + Comfy Cloud でこのワークフローを即座に実行 + + + JSON をダウンロードするか、テンプレートライブラリで "HY 3D 2.0 MV Turbo" を検索 + + + +**入力マテリアル** + +以下のファイルを該当する `LoadImage` ノードにアップロードします。 + + + + `LoadImage` ノード 56 · 正面ビュー + + + `LoadImage` ノード 82 · 背面ビュー + + + `LoadImage` ノード 85 · 左ビュー + + + `LoadImage` ノード 87 · 右ビュー + + ### 1. ワークフロー 以下の画像をダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください。 -![Hunyuan3D-2mv-turbo workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) +![Hunyuan3D-2mv-turbo ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/hunyuan-3d-turbo.webp) 以下の画像をダウンロードしてください。これらを入力画像として使用します。 - -![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) -![input image](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) - +![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/front.png) +![入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d_2mv_turbo/right.png) ### 2. 手動モデルインストール -以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください +以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください。 -- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/resolve/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` にリネームできます +- hunyuan3d-dit-v2-mv-turbo: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2mv/blob/main/hunyuan3d-dit-v2-mv-turbo/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2-mv-turbo.safetensors` に名前を変更できます。 ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // リネーム後のファイル +│ │ └── hunyuan3d-dit-v2-mv-turbo.safetensors // 名前を変更後のファイル ``` ### 3. ワークフローの実行手順 ![ComfyUI hunyuan3d_2mv_turbo](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2mv_turbo.jpg) -1. `Image Only Checkpoint Loader(img2vid model)` ノードが、リネームした `hunyuan3d-dit-v2-mv-turbo.safetensors` モデルを読み込んでいることを確認します -2. 各 `Load Image` ノードに対応するビュー画像を読み込みます -3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します +1. `Image Only Checkpoint Loader(img2vid model)` ノードが、名前を変更した `hunyuan3d-dit-v2-mv-turbo.safetensors` モデルを読み込んでいることを確認します。 +2. 各 `Load Image` ノードに対応するビュー画像を読み込みます。 +3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します。 ## Hunyuan3D-2 単一ビューワークフロー Hunyuan3D-2 ワークフローでは、Hunyuan3D-2 モデルを使用して 3D モデルを生成します。このモデルはマルチビューモデルではありません。このワークフローでは、`Hunyuan3Dv2ConditioningMultiView` ノードの代わりに `Hunyuan3Dv2Conditioning` ノードを使用します。 - -

Run on Comfy Cloud

- +### HY 3D 2.0 (`3d_hunyuan3d_image_to_model`) + +Hunyuan3D 2.0 を使用して、単一の画像から 3D モデルを生成します。 + +HY 3D 2.0 ワークフロープレビュー + + + + このワークフローを Comfy Cloud ですぐに実行 + + + JSON をダウンロードするか、テンプレートライブラリで "HY 3D 2.0" を検索 + + + +**入力素材** + +以下のファイルを該当する `LoadImage` ノードにアップロードします。 + + + + `LoadImage` ノード 56 · `3d_hunyuan3d_image_to_model_input_image.png` + + ### 1. ワークフロー 以下の画像をダウンロードし、ComfyUI にドラッグしてワークフローを読み込んでください。 -![Hunyuan3D-2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) +![Hunyuan3D-2 ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan3d-non-multiview-train.webp) 以下の画像をダウンロードしてください。これを入力画像として使用します。 -![ComfyUI Hunyuan 3D 2 workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan_3d_v2_non_multiview_train.png) +![ComfyUI Hunyuan 3D 2 ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/3d/hunyuan3d-2/hunyuan_3d_v2_non_multiview_train.png) ### 2. 手動モデルインストール -以下のモデルをダウンロードし、対応する ComfyUI フォルダに保存してください +以下のモデルをダウンロードし、対応する ComfyUI フォルダーに保存してください。 -- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/resolve/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` にリネームできます +- hunyuan3d-dit-v2-0: [model.fp16.safetensors](https://huggingface.co/tencent/Hunyuan3D-2/blob/main/hunyuan3d-dit-v2-0/model.fp16.safetensors?download=true) - ダウンロード後、`hunyuan3d-dit-v2.safetensors` に名前を変更できます。 ``` ComfyUI/ ├── models/ │ ├── checkpoints/ -│ │ └── hunyuan3d-dit-v2.safetensors // リネーム後のファイル +│ │ └── hunyuan3d-dit-v2.safetensors // 名前を変更したファイル ``` ### 3. ワークフローの実行手順 ![ComfyUI hunyuan3d_2](/images/tutorial/3d/hunyuan3d-2mv/hunyuan3d_2_non_multiview.jpg) -1. `Image Only Checkpoint Loader(img2vid model)` ノードが、リネームした `hunyuan3d-dit-v2.safetensors` モデルを読み込んでいることを確認します -2. `Load Image` ノードに画像を読み込みます -3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します +1. `Image Only Checkpoint Loader(img2vid model)` ノードが、名前を変更した `hunyuan3d-dit-v2.safetensors` モデルを読み込んでいることを確認します。 +2. `Load Image` ノードに画像を読み込みます。 +3. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します。 ## コミュニティリソース diff --git a/ja/tutorials/3d/triposplat.mdx b/ja/tutorials/3d/triposplat.mdx index 3c1dbffce..85419cc20 100644 --- a/ja/tutorials/3d/triposplat.mdx +++ b/ja/tutorials/3d/triposplat.mdx @@ -2,13 +2,13 @@ title: "TripoSplat 画像からガウシアンスプラット ComfyUI ワークフロー例" description: "TripoSplat を使用して、単一の 2D 画像から高品質な 3D ガウシアンスプラット表現を生成します。密度とレンダリング予算を制御可能。" sidebarTitle: "TripoSplat" -translationSourceHash: 5ca7b463 +translationSourceHash: a6fa8405 translationFrom: tutorials/3d/triposplat.mdx translationBlockHashes: - "_intro": 07dba1db - "How it works": 9da65e03 + "_intro": 8b993083 + "How it works": a8a8cf2e "Workflow node guide": 83cd15a0 - "Steps to run": e4ad7fa4 + "Steps to run": 3bf53f69 "Output options": 9a1cc408 "Model downloads": c6ba2321 --- @@ -20,22 +20,45 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' 複数視点の入力が必要だったり、主にメッシュを生成する従来の 3D 再構築手法とは異なり、TripoSplat は **ガウシアンスプラット** 表現を作成します。これは、数千の色付き 3D ガウシアンを空間に配置してシーンを表現するレンダリング技術で、高速で高品質なレンダリングと制御可能な密度・予算を実現します。 -TripoSplat ワークフロー +### TripoSplat: Image to Gaussian Splat (`3d_triposplat_image_to_gaussian_splat`) + +1枚の2D画像をアップロードします。制御可能な密度と予算で、高品質な3Dガウシアンスプラット表現を生成してレンダリングできます。 + +TripoSplat ワークフロープレビュー - - JSON をダウンロード、またはテンプレートライブラリで "TripoSplat" を検索 + + + このワークフローを Comfy Cloud で即座に実行 + + + JSON をダウンロード、またはテンプレートライブラリで "TripoSplat: Image to Gaussian Splat" を検索 + + + +**入力素材** + +このファイルを該当する `LoadImage` ノードにアップロードします。 + + + + `LoadImage` ノード 99 · `white-hotel-on-rocky-island.png` + + +
+ 入力画像 +
## 仕組み TripoSplat は **フィードフォワードアーキテクチャ** を使用し、単一の RGB 画像を受け取って 3D ガウシアンプリミティブのセットを直接予測します。パイプラインの流れ： -1. **画像エンコード** — 入力画像がビジョンエンコーダー（DINOv2）で処理される -2. **トライプレーン生成** — 特徴がトライプレーン表現にデコードされる -3. **ガウシアン予測** — トライプレーンをサンプリングしてガウシアンパラメータ（位置、スケール、回転、不透明度、色）を生成 -4. **レンダリング** — 微分可能なスプラッティングを使用して任意の視点からガウシアンをレンダリング +1. **画像エンコード**: 入力画像がビジョンエンコーダー（DINOv2）で処理される +2. **トライプレーン生成**: 特徴がトライプレーン表現にデコードされる +3. **ガウシアン予測**: トライプレーンをサンプリングしてガウシアンパラメータ（位置、スケール、回転、不透明度、色）を生成 +4. **レンダリング**: 微分可能なスプラッティングを使用して任意の視点からガウシアンをレンダリング このワークフローはサブグラフノードを使用してモジュール化された処理を行います。サブグラフのドキュメントを参照して、ワークフローをカスタマイズおよび拡張する方法を学んでください。 @@ -88,10 +111,10 @@ TripoSplat は **フィードフォワードアーキテクチャ** を使用し ## 実行手順 -1. **画像を読み込む** — **LoadImage** ノードで単一の 2D 画像を読み込みます -2. **TripoSplat サブグラフを実行** — モデルがガウシアンスプラット表現を生成します -3. **出力形式を選択** — GLB、SPZ、動画、またはメッシュに変換 -4. **結果を表示** — 生成された 3D ファイルまたはレンダリングプレビューを確認 +1. **画像を読み込む**: **LoadImage** ノードで単一の 2D 画像を読み込みます +2. **TripoSplat サブグラフを実行**: モデルがガウシアンスプラット表現を生成します +3. **出力形式を選択**: GLB、SPZ、動画としてエクスポート、またはメッシュにレンダリング +4. **結果を表示**: 生成された 3D ファイルまたはレンダリングプレビューを確認 ## 出力オプション diff --git a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx index 97dd4f880..e08b88bf0 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1-5.mdx @@ -15,6 +15,9 @@ translationBlockHashes: --- + + + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ## ComfyUIにおけるACE-Step 1.5について diff --git a/ja/tutorials/audio/ace-step/ace-step-v1.mdx b/ja/tutorials/audio/ace-step/ace-step-v1.mdx index 2a2e468ce..9cf1051d1 100644 --- a/ja/tutorials/audio/ace-step/ace-step-v1.mdx +++ b/ja/tutorials/audio/ace-step/ace-step-v1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI ACE-Step ネイティブサンプル" description: "本ガイドでは、ComfyUI で ACE-Step モデルを用いてダイナミックな音楽を作成する方法を説明します" sidebarTitle: "ACE-Step 1.0" -translationSourceHash: 0090114c +translationSourceHash: b3044926 translationFrom: tutorials/audio/ace-step/ace-step-v1.mdx translationBlockHashes: "_intro": 56ef7364 "ACE-Step ComfyUI Text-to-Audio Generation Workflow Example": 3f211bda - "ACE-Step ComfyUI Audio-to-Audio Workflow": fe7381f0 + "ACE-Step ComfyUI Audio-to-Audio Workflow": 06b425cb "ACE-Step Prompt Guide": 90bf0bf7 "ACE-Step Related Resources": f74e2db9 --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx"; ACE-Step は、中国のチーム StepFun と ACE Studio が共同開発したオープンソースの音楽生成基盤モデルであり、音楽クリエイターに効率的で柔軟性が高く、高品質な音楽生成および編集ツールを提供することを目的としています。 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx index a0c3bcfa4..83f92a14d 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-1.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-1.mdx @@ -2,50 +2,58 @@ title: "Stable Audio 1.0 ComfyUI ワークフロー例" description: "Stability AI のオープンソースモデル Stable Audio 1.0 を ComfyUI で使用してテキストから音声を生成する方法を解説します。" sidebarTitle: "Stable Audio 1.0" -translationSourceHash: 627114c2 +translationSourceHash: bfce4810 translationFrom: tutorials/audio/stable-audio/stable-audio-1.mdx +translationBlockHashes: + "_intro": ca52fb1b + "Workflow": d44f0c23 + "Model download": 4c9f3106 --- - import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" -**Stable Audio 1.0** は、Stability AI 初のオープンソース音声生成モデルです。テキストプロンプトを受け取り、音声クリップを生成します。ComfyUI では標準のテキストから音声パイプラインとして動作します：CLIP がプロンプトをエンコードし、KSampler が潜在空間をノイズ除去し、VAE が音声にデコードします。 +**Stable Audio 1.0** は、Stability AI初のオープンソースのオーディオ生成モデルです。テキストプロンプトを受け取り、オーディオクリップを生成します。ComfyUIでは、標準的なテキストから音声生成のパイプラインとして動作します。CLIPがプロンプトをエンコードし、Kサンプラーが潜在空間をノイズ除去し、VAEがそれをオーディオにデコードします。 **関連リンク**： - [GitHub: Stability-AI/stable-audio-open-1.0](https://github.com/Stability-AI/stable-audio-open-1.0) ## ワークフロー - - JSON をダウンロードするか、テンプレートライブラリで"Stable Audio 1.0"を検索 - +### Stable Audio 1.0: テキストからオーディオへ (`audio_stable_audio_example`) + +テキストプロンプトからStable Audioを使用してオーディオを生成します。 - +Stable Audio 1.0 テキストからオーディオへのワークフロープレビュー + + + Comfy Cloud で開く + + JSONをダウンロードするか、テンプレートライブラリで "Stable Audio 1.0: Text to Audio" を検索 + + -![Stable Audio 1.0 ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_example-1.webp) - -**標準の ComfyUI ノード**のみを使用し、カスタムノードは不要です。Stable Audio 1.0 チェックポイントを読み込み、CLIP テキストエンコーダーでプロンプトをエンコードし、KSampler で潜在空間をノイズ除去し、VAE で音声にデコードします。 +このワークフローは**スタンダードなComfyUIノード**を使用しており、カスタムノードは不要です。Stable Audio 1.0チェックポイントをロードし、CLIPテキストエンコーダー（t5-base）でプロンプトをエンコード、Kサンプラーで潜在オーディオをデノイズし、モデルのVAEでオーディオにデコードします。 -**使用方法**： -1. **モデルを読み込む** — `CheckpointLoaderSimple` ノードで `stable-audio-open-1.0.safetensors` を使用 -2. **プロンプトを書く** — `CLIPTextEncode` ノードに説明を入力（例："heaven church electronic dance music"） -3. **再生時間を設定** — `EmptyLatentAudio` ノードの長さを調整（デフォルト 47.6 秒） -4. **実行**（`Ctrl/Cmd + Enter`）をクリックして生成。音声は `ComfyUI/output/audio/` に保存されます +**使用方法**: +1. **チェックポイントをロード** — `CheckpointLoaderSimple`ノードで`stable-audio-open-1.0.safetensors`を使用 +2. **プロンプトを入力** — `CLIPTextEncode`ノードに説明を入力（例："heaven church electronic dance music"） +3. **再生時間を設定** — `EmptyLatentAudio`ノードの長さの値を調整（デフォルト47.6秒） +4. **実行**（`Ctrl/Cmd + Enter`）をクリックして生成。オーディオは`ComfyUI/output/audio/`に保存されます。 -## モデルダウンロード +## モデルのダウンロード -ワークフローを読み込むと、モデルがない場合に ComfyUI がダウンロードリンクを提示します。手動で設定する場合、以下のファイルをダウンロードして適切なフォルダに配置してください。 +ワークフローの読み込み時に、不足しているモデルがある場合、ComfyUI はダウンロードリンクを表示します。手動で設定するには、以下のファイルをダウンロードし、適切なフォルダに配置してください。 ### チェックポイント - - 2.3GB。models/checkpoints/ に配置 + + 2.3GB。models/checkpoints/ に配置してください。 -以下のように配置します： +チェックポイントは以下の場所に配置します： ``` 📂 ComfyUI/ @@ -56,11 +64,11 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" ### テキストエンコーダー - - プロンプト処理用テキストエンコーダー。models/text_encoders/ に配置 + + プロンプトの条件付け用テキストエンコーダー。models/text_encoders/ に配置してください。 -以下のように配置します： +テキストエンコーダーは以下の場所に配置します： ``` 📂 ComfyUI/ @@ -69,4 +77,4 @@ import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" │ └── t5-base.safetensors ``` -配置後、ComfyUI で **R** キーを押してノード定義をリフレッシュすると、最新のモデルが利用可能になります。 +ファイルを配置した後、ComfyUI で **R** キーを押してノードを更新し、最新のモデルを読み込んでください。 diff --git a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx index e638d2d56..1f6191a8a 100644 --- a/ja/tutorials/audio/stable-audio/stable-audio-3.mdx +++ b/ja/tutorials/audio/stable-audio/stable-audio-3.mdx @@ -2,15 +2,16 @@ title: "Stable Audio 3 ComfyUI ワークフロー例" description: "Stability AI のオープンソースモデル Stable Audio 3 を ComfyUI で使用し、Qwen によるプロンプト拡張とカテゴリ認識リプロンプト機能を活用したテキストから音声生成を行う方法を解説します。" sidebarTitle: "Stable Audio 3" -translationSourceHash: 0f15443c +translationSourceHash: 1be77476 translationFrom: tutorials/audio/stable-audio/stable-audio-3.mdx translationBlockHashes: "_intro": 7aae7ae5 - "Available workflows": 85ee7552 - "Model download": d729490a + "Available workflows": b4e9e047 + "Model download": 384798ea --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" @@ -29,41 +30,47 @@ Stable Audio 3 には3つのバリエーションがあります： ## 利用可能なワークフロー -### Stable Audio 3 Medium +### Stable Audio 3.0 Medium (`audio_stable_audio_3_medium`) - - JSON をダウンロードするか、テンプレートライブラリで"Stable Audio 3 Medium"を検索 - +短いテキストアイデア、オプションの再生時間、シード、カテゴリを入力します。AIによるテキスト拡張をオプションで使用して、Stable Audio 3でステレオオーディオ（音楽、効果音、楽器）を生成します。 + +Stable Audio 3 Medium ワークフローのプレビュー - + + Comfy Cloud で開く - -![Stable Audio 3 Medium ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium-1.webp) + + JSON をダウンロードするか、テンプレートライブラリで "Stable Audio 3.0 Medium" を検索 + + **Stable Audio 3 Medium** ワークフローは、完全なテキストから音声生成パイプラインです。短いテキストアイデア、任意の再生時間、シード、カテゴリを入力すると、Qwen を使用したカテゴリ認識リプロンプトテンプレートでプロンプトを拡張し、Stable Audio 3 チェックポイントでステレオ音声を生成します。 **使用方法**： 1. **テキストアイデア** — 生成したい音声の簡単な説明を入力（例：「重いベースのアップテンポなエレクトロニックダンスミュージック」） -2. **再生時間** — クリップの長さ（秒）を設定 -3. **シード** — 再現性を制御 +2. **再生時間** — クリップの長さ（秒）を設定（デフォルトは異なる） +3. **シード** — 再現性を制御するためにシード値を調整 4. **カテゴリ** — リプロンプトプリセットを選択：**Music**（音楽）、**Instrument**（楽器）、**SFX**（効果音）、**One-shot**（単発音） -5. **リプロンプトを有効化** — `use_reprompt` をオンにして Qwen が短いアイデアを詳細なプロンプトに拡張 +5. **リプロンプトを有効化** — `use_reprompt` をオンにして Qwen が短いアイデアを詳細なプロンプトに拡張してから生成 6. **実行**（`Ctrl/Cmd + Enter`）をクリックして生成。音声は `ComfyUI/output/audio/` に保存されます -### Stable Audio 3 Medium Base +### Stable Audio 3.0 Medium Base (`audio_stable_audio_3_medium_base`) - - JSON をダウンロードするか、テンプレートライブラリで"Stable Audio 3 Medium Base"を検索 - +オーディオ、音楽、効果音の短いテキスト説明を入力します。ワークフローは Qwen でプロンプトを拡張し、Stable Audio 3 からステレオオーディオクリップを生成します。 + +Stable Audio 3 Medium Base ワークフローのプレビュー - + + Comfy Cloud で開く + + JSON をダウンロードするか、テンプレートライブラリで "Stable Audio 3.0 Medium Base" を検索 + + -![Stable Audio 3 Medium Base ワークフロー](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/audio_stable_audio_3_medium_base-1.webp) - -Qwen リプロンプト機能を省いたシンプルなバージョン。完全なテキストプロンプトを直接 Stable Audio 3 モデルに渡します。すでに詳細なプロンプトがある場合や、高速に生成したい場合に使用します。 +Qwen リプロンプト機能を省いたシンプルなバージョン。完全なテキストプロンプトを想定し、そのままモデルに渡します。すでに詳細なプロンプトがある場合や、高速に生成したい場合に使用します。 **使用方法**： 1. **テキストプロンプト** — 生成したい音声の詳細な説明を入力 @@ -73,17 +80,19 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 ## モデルダウンロード -ワークフローを読み込むと、モデルがない場合に ComfyUI がダウンロードリンクを提示します。手動で設定する場合、以下のファイルをダウンロードして適切なフォルダに配置してください。 +ワークフローを読み込むと、不足しているモデルがあれば ComfyUI がダウンロードリンクを提示します。手動で設定する場合、以下のファイルをダウンロードして適切なフォルダに配置してください。 ### チェックポイント - + + Medium ワークフロー用。models/checkpoints/ に配置 - + Medium Base ワークフロー用。models/checkpoints/ に配置 + 以下のように配置します： @@ -97,13 +106,15 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 ### テキストエンコーダー - + + すべての Stable Audio 3 ワークフローで必要。models/text_encoders/ に配置 - + Medium ワークフローで必要（Qwen リプロンプト）。models/text_encoders/ に配置 + 以下のように配置します： @@ -115,4 +126,4 @@ Qwen リプロンプト機能を省いたシンプルなバージョン。完全 │ └── qwen3.5_2b_bf16.safetensors ``` -配置後、ComfyUI で **R** キーを押してノード定義をリフレッシュすると、最新のモデルが利用可能になります。 +配置後、ComfyUI で **R** キーを押してノードを更新し、最新のモデルを読み込みます。 diff --git a/ja/tutorials/basic/multiple-loras.mdx b/ja/tutorials/basic/multiple-loras.mdx index 88ddeae12..4029df45e 100644 --- a/ja/tutorials/basic/multiple-loras.mdx +++ b/ja/tutorials/basic/multiple-loras.mdx @@ -2,7 +2,7 @@ title: "ComfyUI での複数 LoRA の適用例" sidebarTitle: "複数の LoRA" description: "本ガイドでは、ComfyUI で複数の LoRA モデルを同時に適用する方法を紹介します。" -translationSourceHash: 4726e42f +translationSourceHash: 47f6f952 translationFrom: tutorials/basic/multiple-loras.mdx --- diff --git a/ja/tutorials/basic/upscale.mdx b/ja/tutorials/basic/upscale.mdx index 8468e285c..5f157fdf3 100644 --- a/ja/tutorials/basic/upscale.mdx +++ b/ja/tutorials/basic/upscale.mdx @@ -2,11 +2,11 @@ title: "ComfyUI 画像アップスケールワークフロー" description: "本ガイドでは、AI描画における画像アップスケールの概念を解説し、ComfyUIで画像アップスケールワークフローを実装する方法を紹介します" sidebarTitle: "アップスケール" -translationSourceHash: 445e2878 +translationSourceHash: ef55806b translationFrom: tutorials/basic/upscale.mdx translationBlockHashes: "What is Image Upscaling?": 075b71b9 - "Upscaling Workflow": 7d8b6c3c + "Upscaling Workflow": 9f559d39 "Text-to-Image Combined Workflow": b0d0c767 "Additional Tips": 5352d8c0 --- @@ -28,14 +28,14 @@ translationBlockHashes: より高度なアップスケールオプションをお探しですか？包括的な[画像アップスケールガイド](/ja/tutorials/utility/image-upscale)をご覧ください。こちらでは、ポートレート強化、製品撮影など、さまざまなモデルとユースケースについて解説しています。 -## アップスケールワークフロー +## アップスケーリングワークフロー ### モデルのインストール 必要なESRGANモデルのダウンロード手順： -[OpenModelDB](https://openmodeldb.info/) にアクセスし、アップスケールモデル（例：RealESRGAN）を検索・ダウンロードします。 +[OpenModelDB](https://openmodeldb.info/) にアクセスし、アップスケーリングモデル（例：RealESRGAN）を検索・ダウンロードします。 ![openmodeldb](/images/tutorial/basic/upscale/upscale_OpenModelDB.jpg) @@ -55,7 +55,7 @@ translationBlockHashes: ### ワークフローとアセット -以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして、基本的なアップスケールワークフローを読み込みます： +以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして、基本的なアップスケーリングワークフローを読み込みます： ![Upscale workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/basic/upscale_workflow.png) @@ -75,7 +75,7 @@ translationBlockHashes: 2. 入力画像を `Load Image` ノードにアップロード 3. `Queue` ボタンをクリックするか、ショートカット `Ctrl（Macの場合はCmd） + Enter` を押して画像を生成 -このワークフローの核となるのは、`Load Upscale Model` および `Upscale Image (Using Model)` の2つのノードです。これらは入力画像を受け取り、選択したモデルを用いて画像をアップスケールします。 +このワークフローの核となるのは、`Load Upscale Model` および `Upscale Image (Using Model)` の2つのノードです。これらは入力画像を受け取り、選択したモデルを用いて画像をアップスケーリングします。 ## テキストから画像を生成するワークフローとの統合 diff --git a/ja/tutorials/flux/flux-1-controlnet.mdx b/ja/tutorials/flux/flux-1-controlnet.mdx index deaffe414..3ec7ae34b 100644 --- a/ja/tutorials/flux/flux-1-controlnet.mdx +++ b/ja/tutorials/flux/flux-1-controlnet.mdx @@ -2,18 +2,18 @@ title: "ComfyUI Flux.1 ControlNet の使用例" description: "本ガイドでは、Flux.1 ControlNet を用いたワークフローの使用例を紹介します。" sidebarTitle: "Flux.1 ControlNet" -translationSourceHash: a02fc86f +translationSourceHash: e3f37cab translationFrom: tutorials/flux/flux-1-controlnet.mdx translationBlockHashes: "_intro": a69fbfae "FLUX.1 ControlNet Model Introduction": c242319b - "FLUX.1-Canny-dev Complete Version Workflow": 0c4240ac - "FLUX.1-Depth-dev-lora Workflow": ebfe0fb5 + "FLUX.1-Canny-dev Complete Version Workflow": 5e5f6373 + "FLUX.1-Depth-dev-lora Workflow": 175fe60a "Community Versions of Flux Controlnets": 29728dcf --- -![Flux.1 Canny Controlnet](/images/tutorial/flux/flux-1-canny-controlnet.png) -![Flux.1 Depth Controlnet](/images/tutorial/flux/flux-1-depth-controlnet.png) +![Flux.1 Canny コントロールネット](/images/tutorial/flux/flux-1-canny-controlnet.png) +![Flux.1 Depth コントロールネット](/images/tutorial/flux/flux-1-depth-controlnet.png) ## FLUX.1 ControlNet モデルの概要 @@ -37,7 +37,7 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 すべてのワークフロー画像のメタデータには、対応するモデルのダウンロード情報が記載されています。ワークフローの読み込みには以下の方法をご利用ください： - 画像を ComfyUI へ直接ドラッグ＆ドロップ -- またはメニュー `Workflows` → `Open（Ctrl+O）` を選択 +- またはメニュー `Workflows` -> `Open（ctrl+o）` を選択 デスクトップ版を使用していない場合、あるいは一部のモデルが自動ダウンロードされない場合は、手動インストールのセクションに従い、モデルファイルを対応するフォルダーに保存してください。 @@ -46,9 +46,11 @@ Depth 版では深度マップ抽出技術を用いて元画像の空間構造 - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -### Flux.1 Canny Model (`flux_canny_model_example`) +## FLUX.1-Canny-dev 完全バージョンワークフロー -Generate images guided by edge detection using Flux.1 Canny. +### Flux.1 Canny モデル (`flux_canny_model_example`) + +エッジ検出による画像生成を、Flux.1 Canny を使用して行います。 Flux.1 Canny ワークフロープレビュー @@ -57,17 +59,17 @@ Generate images guided by edge detection using Flux.1 Canny. Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロードするか、テンプレートライブラリで "Flux.1 Canny" を検索 **入力素材** -次の `\1` ノードにこのファイルをアップロード： +このファイルを対応する `画像を読み込む` ノードにアップロードします: - `LoadImage` node 17 · `flux_canny_model_example_input_image.png` + `画像を読み込む` ノード 17 · `flux_canny_model_example_input_image.png` @@ -75,31 +77,23 @@ Generate images guided by edge detection using Flux.1 Canny. flux_canny_model_example_input_image.png
-**## 1. ワークフローおよび関連アセット - -下記のワークフロー画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 - -![ComfyUI ワークフロー - ControlNet](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-canny-dev.png) - -下記の画像をダウンロードし、入力画像として使用します。 +### 1. ワークフローとアセット -![ComfyUI Flux.1 Canny Controlnet 入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-canny-dev-input.png) - -### 2. モデルの手動インストール +### 2. 手動モデルインストール -以前に [Flux 関連ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image) を使用済みの場合、**flux1-canny-dev.safetensors** のみをダウンロードすれば十分です。 -ただし、[black-forest-labs/FLUX.1-Canny-dev](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev) のライセンス条件に同意する必要があります。そのため、[black-forest-labs/FLUX.1-Canny-dev](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev) のページにアクセスし、下図のように該当リポジトリの利用規約に同意済みであることを確認してください。 -![Flux Agreement](/images/tutorial/flux/flux1_canny_dev_agreement.jpg) +以前に [Flux 関連ワークフローの完全バージョン](/ja/tutorials/flux/flux-1-text-to-image) をご利用いただいたことがある場合は、**flux1-canny-dev.safetensors** モデルファイルのみをダウンロードする必要があります。 +まず [black-forest-labs/FLUX.1-Canny-dev](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev) の利用規約に同意する必要がありますので、[black-forest-labs/FLUX.1-Canny-dev](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev) のページにアクセスし、下の画像のように該当する利用規約に同意したことを確認してください。 +![Flux 同意](/images/tutorial/flux/flux1_canny_dev_agreement.jpg) -必要なモデル一覧： +完全なモデルリスト: - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) - [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true)（対応リポジトリの利用規約に事前に同意していることをご確認ください） +- [flux1-canny-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev/blob/main/flux1-canny-dev.safetensors?download=true) (該当リポジトリの利用規約に同意していることを確認してください) -ファイルの保存先ディレクトリ構成： +ファイル保存場所: ``` ComfyUI/ ├── models/ @@ -112,51 +106,53 @@ ComfyUI/ │ └── flux1-canny-dev.safetensors ``` -### 3. ワークフロー実行手順（ステップ・バイ・ステップ） +### 3. ステップバイステップのワークフロー実行 -![ComfyUI Flux.1 Canny Controlnet 実行手順](/images/tutorial/flux/flow_diagram_flux_1_canny_dev.jpg) +![ComfyUI Flux.1 Canny Controlnet ステッププロセス](/images/tutorial/flux/flow_diagram_flux_1_canny_dev.jpg) -1. `Load VAE` ノードで `ae.safetensors` が正しく読み込まれていることを確認してください。 -2. `Load Diffusion Model` ノードで `flux1-canny-dev.safetensors` が正しく読み込まれていることを確認してください。 -3. `DualCLIPLoader` ノードで以下のモデルが読み込まれていることを確認してください： +1. `VAEを読み込む` ノードで `ae.safetensors` が読み込まれていることを確認します +2. `拡散モデルを読み込む` ノードで `flux1-canny-dev.safetensors` が読み込まれていることを確認します +3. `DualCLIPLoader` ノードで以下のモデルが読み込まれていることを確認します: - clip_name1: t5xxl_fp16.safetensors - clip_name2: clip_l.safetensors -4. `Load Image` ノードで上記で提供した入力画像をアップロードしてください。 -5. `Queue` ボタンをクリックするか、ショートカット `Ctrl（Cmd） + Enter` を押してワークフローを実行してください。 +4. `画像を読み込む` ノードで提供された入力画像をアップロードします +5. `キュー` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します -### 4. 実験を開始しましょう +### 4. 実験を始めましょう -[FLUX.1-Depth-dev](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev) モデルを用いて、Depth 版のワークフローを実行してみてください。 +[FLUX.1-Depth-dev](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev) モデルを使用して、Depth バージョンのワークフローを完成させてみてください -以下の画像を入力としてご利用いただけます： +以下の画像を入力として使用できます ![ComfyUI 室内深度マップ](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/controlnet/depth-t2i-adapter_input.png) -また、以下のカスタムノードを用いて画像の前処理を行うことも可能です： +または、以下のカスタムノードを使用して画像の前処理を完了させてください: - [ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet) - [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) -### Flux.1 Depth Lora (`flux_depth_lora_example`) +## FLUX.1-Depth-dev-lora ワークフロー + +### Flux.1 Depth Lora（`flux_depth_lora_example`） -Generate images guided by depth information using Flux.1 LoRA. +Flux.1 LoRAを使用して、深度情報に基づいた画像を生成します。 -Flux.1 Depth LoRA ワークフロープレビュー +Flux.1 Depth LoRA ワークフローのプレビュー - - Comfy Cloud で開く + + Comfy Cloudで開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSONをダウンロードするか、テンプレートライブラリで「Flux.1 Depth LoRA」を検索してください **入力素材** -次の `\1` ノードにこのファイルをアップロード： +このファイルを対応する `LoadImage` ノードにアップロードしてください： - `LoadImage` node 17 · `flux_depth_lora_example_input_image.png` + `LoadImage`ノード17 · `flux_depth_lora_example_input_image.png` @@ -164,30 +160,22 @@ Generate images guided by depth information using Flux.1 LoRA. flux_depth_lora_example_input_image.png
-### 1. ワークフローおよび関連アセット +### 1. ワークフローとアセット -下記のワークフロー画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 - -![ComfyUI ワークフロー - ControlNet](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-depth-dev-lora.png) - -下記の画像をダウンロードし、入力画像として使用します。 - -![ComfyUI Flux.1 Depth Controlnet 入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/controlnet/flux-1-depth-dev-lora-input.png) - -### 2. モデルの手動ダウンロード +### 2. 手動モデルダウンロード -以前に [Flux 関連ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image) を使用済みの場合、**flux1-depth-dev-lora.safetensors** のみをダウンロードすれば十分です。 +すでに[Flux関連ワークフローの完全版](/ja/tutorials/flux/flux-1-text-to-image)を使用している場合は、**flux1-depth-dev-lora.safetensors**モデルファイルのみをダウンロードする必要があります。 -必要なモデル一覧： +完全なモデルリスト： - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) - [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) - [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors?download=true) - [flux1-depth-dev-lora.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev-lora/blob/main/flux1-depth-dev-lora.safetensors?download=true) -ファイルの保存先ディレクトリ構成： +ファイルの保存場所： ``` ComfyUI/ ├── models/ @@ -202,38 +190,36 @@ ComfyUI/ │ └── flux1-depth-dev-lora.safetensors ``` -### 3. ワークフロー実行手順（ステップ・バイ・ステップ） +### 3. ワークフローのステップごとの実行 -![ComfyUI Flux.1 Depth Controlnet 実行手順](/images/tutorial/flux/flow_diagram_flux_1_depth_dev_lora.jpg) +![ComfyUI Flux.1 Depth Controlnet のステッププロセス](/images/tutorial/flux/flow_diagram_flux_1_depth_dev_lora.jpg) -1. `Load Diffusion Model` ノードで `flux1-dev.safetensors` が正しく読み込まれていることを確認してください。 -2. `LoraLoaderModelOnly` ノードで `flux1-depth-dev-lora.safetensors` が正しく読み込まれていることを確認してください。 -3. `DualCLIPLoader` ノードで以下のモデルが読み込まれていることを確認してください： +1. `Load Diffusion Model`ノードに`flux1-dev.safetensors`が読み込まれていることを確認してください +2. `LoraLoaderModelOnly`ノードに`flux1-depth-dev-lora.safetensors`が読み込まれていることを確認してください +3. `DualCLIPLoader`ノードに以下のモデルが読み込まれていることを確認してください： - clip_name1: t5xxl_fp16.safetensors - clip_name2: clip_l.safetensors -4. `Load Image` ノードで上記で提供した入力画像をアップロードしてください。 -5. `Load VAE` ノードで `ae.safetensors` が正しく読み込まれていることを確認してください。 -6. `Queue` ボタンをクリックするか、ショートカット `Ctrl（Cmd） + Enter` を押してワークフローを実行してください。 +4. `Load Image`ノードに提供された入力画像をアップロードしてください +5. `Load VAE`ノードに`ae.safetensors`が読み込まれていることを確認してください +6. `Queue`ボタンをクリックするか、ショートカット`Ctrl(cmd) + Enter`を使用してワークフローを実行してください -### 4. 実験を開始しましょう +### 4. 実験を始めましょう -[FLUX.1-Canny-dev-lora](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev-lora) モデルを用いて、Canny 版のワークフローを実行してみてください。 +[FLUX.1-Canny-dev-lora](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev-lora)モデルを使用して、Cannyバージョンのワークフローを試してみてください -画像の前処理には、以下のカスタムノードをご利用ください： -- [ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet) -- [ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux) +画像の前処理を完了するには、[ComfyUI-Advanced-ControlNet](https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet)または[ComfyUI ControlNet aux](https://github.com/Fannovel16/comfyui_controlnet_aux)を使用してください -## コミュニティ提供の Flux ControlNet +## コミュニティ版Fluxコントロールネット -XLab および InstantX＋Shakker Labs が、Flux 向けの ControlNet を公開しています。 +XLabおよびInstantX + Shakker LabsがFlux用のコントロールネットをリリースしています。 **InstantX:** - [FLUX.1-dev-Controlnet-Canny](https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Canny/blob/main/diffusion_pytorch_model.safetensors) - [FLUX.1-dev-ControlNet-Depth](https://huggingface.co/Shakker-Labs/FLUX.1-dev-ControlNet-Depth/blob/main/diffusion_pytorch_model.safetensors) - [FLUX.1-dev-ControlNet-Union-Pro](https://huggingface.co/Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro/blob/main/diffusion_pytorch_model.safetensors) -**XLab**: [flux-controlnet-collections](https://huggingface.co/XLabs-AI/flux-controlnet-collections) +**XLab**: [flux-controlnet-collections](https://huggingface.co/XLabs-AI/flux-controlnet-collections) -これらのファイルは `ComfyUI/models/controlnet` ディレクトリに配置してください。 +これらのファイルを `ComfyUI/models/controlnet` ディレクトリに配置してください。 -対応するワークフロー画像は、[Flux Controlnet 使用例](https://raw.githubusercontent.com/comfyanonymous/ComfyUI_examples/refs/heads/master/flux/flux_controlnet_example.png) から取得できます。入力画像には、[こちら](https://raw.githubusercontent.com/comfyanonymous/ComfyUI_examples/refs/heads/master/flux/girl_in_field.png) の画像をご利用ください。 +対応するワークフロー画像は [Flux Controlnetの例](https://raw.githubusercontent.com/comfyanonymous/ComfyUI_examples/refs/heads/master/flux/flux_controlnet_example.png) から、入力画像は[こちら](https://raw.githubusercontent.com/comfyanonymous/ComfyUI_examples/refs/heads/master/flux/girl_in_field.png)の画像を使用してください。 diff --git a/ja/tutorials/flux/flux-1-fill-dev.mdx b/ja/tutorials/flux/flux-1-fill-dev.mdx index a24d4eda0..8ddd59df5 100644 --- a/ja/tutorials/flux/flux-1-fill-dev.mdx +++ b/ja/tutorials/flux/flux-1-fill-dev.mdx @@ -2,16 +2,17 @@ title: "ComfyUI Flux.1 fill dev の使用例" description: "本ガイドでは、Flux.1 fill dev を用いた Inpainting（画像修復）および Outpainting（画像拡張）ワークフローの構築方法を解説します。" sidebarTitle: "Flux.1 fill dev" -translationSourceHash: 5cba3176 +translationSourceHash: 3ac84f99 translationFrom: tutorials/flux/flux-1-fill-dev.mdx translationBlockHashes: "_intro": 73fb6ba4 "Introduction to Flux.1 fill dev Model": 6dfbcfd4 "Flux.1 Fill dev and related models installation": cf7747cd - "Flux.1 Fill dev inpainting workflow": 9589641c - "Flux.1 Fill dev Outpainting Workflow": dd827926 + "Flux.1 Fill dev inpainting workflow": 9de866eb + "Flux.1 Fill dev Outpainting Workflow": b523a7cd --- + ![Flux.1 fill dev](/images/tutorial/flux/flux-fill-dev-demo.jpeg) ## Flux.1 fill dev モデルの概要 @@ -60,22 +61,22 @@ ComfyUI/ ### Flux.1 Inpaint (`flux_fill_inpaint_example`) -Fill missing parts of images using Flux.1 Fill Inpainting. +Flux.1 Fill Inpainting を使用して、画像の不足している部分を補完します。 -Flux.1 inpaint ワークフロープレビュー +Flux.1 インペイント ワークフロープレビュー Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Flux.1 Inpaint」を検索 **入力素材** -次の `\1` ノードにこのファイルをアップロード： +対応する `LoadImage` ノードにこのファイルをアップロードします。 @@ -91,9 +92,16 @@ Fill missing parts of images using Flux.1 Fill Inpainting.
入力画像 - Flux.1 inpaint 出力例 + Flux.1 インペイント 出力例
+### 1. インペイントワークフローとアセット + + +対応する画像には既にアルファチャンネルが含まれているため、別途マスクを描く必要はありません。 +自分でマスクを描きたい場合は、[こちら](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/inpaint/flux_fill_inpaint_input_original.png)をクリックしてマスクなしの画像を入手し、[ComfyUI Layout Inpainting の例](/ja/tutorials/basic/inpaint#using-the-mask-editor)のマスクエディターの使用方法のセクションを参照して、`Load Image` ノードでマスクを描く方法を学んでください。 + + ### 2. ワークフローの実行手順 ![ComfyUI Flux.1 Fill dev Inpainting ワークフロー](/images/tutorial/flux/flow_diagram_inpaint.jpg) @@ -103,27 +111,50 @@ Fill missing parts of images using Flux.1 Fill Inpainting. - clip_name1: `t5xxl_fp16.safetensors` - clip_name2: `clip_l.safetensors` 3. `Load VAE` ノードに `ae.safetensors` が正しく読み込まれていることを確認します。 -4. 本文書で提供される入力画像を `Load Image` ノードへアップロードします。マスクなしの画像を使用する場合は、マスクエディタを用いてマスクの描画を完了させてください。 +4. 本文書で提供される入力画像を `Load Image` ノードへアップロードします。マスクなしの画像を使用する場合は、マスクエディターを用いてマスクの描画を完了させてください。 5. `Queue` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合 Cmd）+ Enter` を押してワークフローを実行します。 -## Flux.1 Fill dev を用いた Outpainting ワークフロー +## Flux.1 Fill dev Outpainting ワークフロー + +### Flux.1 Outpaint（`flux_fill_outpaint_example`） + +Flux.1 の outpainting を使用して画像を境界の外側に拡張します。 + +Flux.1 outpaint ワークフローのプレビュー -### 1. Outpainting ワークフローおよび関連アセット + + + Comfy Cloud で開く + + + JSON をダウンロードするか、テンプレートライブラリで「Flux.1 Outpaint」を検索してください + + -以下の画像をダウンロードし、ComfyUI へドラッグ＆ドロップすることで、対応するワークフローを読み込んでください。 -![ComfyUI Flux.1 outpaint](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint.png) +**入力素材** -以下の画像をダウンロードし、入力画像として使用します。 -![ComfyUI Flux.1 outpaint input](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/outpaint/flux_fill_dev_outpaint_input.png) +このファイルを該当する `LoadImage` ノードにアップロードします： + + + + `LoadImage` ノード 17 · `flux_fill_outpaint_example_input_image.png` + + + +
+ flux_fill_outpaint_example_input_image.png +
+ +### 1. Outpainting ワークフローとアセット ### 2. ワークフローの実行手順 ![ComfyUI Flux.1 Fill dev Outpainting ワークフロー](/images/tutorial/flux/flow_diagram_outpaint.jpg) -1. `Load Diffusion Model` ノードに `flux1-fill-dev.safetensors` が正しく読み込まれていることを確認します。 +1. `Load Diffusion Model` ノードに `flux1-fill-dev.safetensors` が読み込まれていることを確認します。 2. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認します： - clip_name1: `t5xxl_fp16.safetensors` - clip_name2: `clip_l.safetensors` -3. `Load VAE` ノードに `ae.safetensors` が正しく読み込まれていることを確認します。 -4. 本文書で提供される入力画像を `Load Image` ノードへアップロードします。 -5. `Queue` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合 Cmd）+ Enter` を押してワークフローを実行します。 +3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認します。 +4. 本文書で提供される入力画像を `Load Image` ノードにアップロードします。 +5. `Queue` ボタンをクリックするか、ショートカット `Ctrl（Cmd）+ Enter` を使用してワークフローを実行します。 diff --git a/ja/tutorials/flux/flux-1-kontext-dev.mdx b/ja/tutorials/flux/flux-1-kontext-dev.mdx index 319657edc..2a7698756 100644 --- a/ja/tutorials/flux/flux-1-kontext-dev.mdx +++ b/ja/tutorials/flux/flux-1-kontext-dev.mdx @@ -2,13 +2,13 @@ title: "ComfyUI Flux Kontext Dev ネイティブワークフローの例" description: "ComfyUI Flux Kontext Dev ネイティブワークフローの例。" sidebarTitle: "Flux.1 Kontext Dev" -translationSourceHash: d2e6fb96 +translationSourceHash: 0d9735ce translationFrom: tutorials/flux/flux-1-kontext-dev.mdx translationBlockHashes: "_intro": 51c48128 "About FLUX.1 Kontext Dev": bb2f00f7 "Model Download": e6b40970 - "Flux.1 Kontext Dev Workflow": ffae9c25 + "Flux.1 Kontext Dev Workflow": 6b98e95d --- import PromptTechniques from "/snippets/ja/tutorials/flux/prompt-techniques.mdx"; @@ -17,7 +17,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -36,11 +36,11 @@ FLUX.1 Kontext シリーズ全体と同様のコア機能を提供します： ### バージョン情報 -- **[FLUX.1 Kontext [pro]** — 商用版。高速な反復編集に特化 -- **FLUX.1 Kontext [max]** — 実験版。プロンプトへの適合性がさらに強化されています -- **FLUX.1 Kontext [dev]** — オープンソース版（本チュートリアルで使用）。12B（120億）パラメータ。主に研究用途向け +- **[FLUX.1 Kontext [pro]** : 商用版、高速な反復編集に特化 +- **FLUX.1 Kontext [max]** : 実験版、プロンプトへの適合性がさらに強化されています +- **FLUX.1 Kontext [dev]** : オープンソース版（本チュートリアルで使用）、12B（120億）パラメータ、主に研究用途向け -現在、ComfyUI では上記すべてのバージョンが利用可能です。[Pro および Max 版](/ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) は API ノード経由で呼び出せますが、Dev 版（オープンソース版）については、本ガイドの手順をご参照ください。 +現在、ComfyUI では上記すべてのバージョンが利用可能です。[Pro および Max 版](/ja/tutorials/partner-nodes/black-forest-labs/flux-1-kontext) はパートナーノード経由で呼び出せますが、Dev 版（オープンソース版）については、本ガイドの手順をご参照ください。 @@ -138,3 +138,59 @@ This workflow uses the `Load Image(from output)` node to load the image to be ed 6. `Queue` ボタンをクリックするか、ショートカットキー `Ctrl（Windows/Linux）`／`Cmd（macOS） + Enter` を押してワークフローを実行します。 + +## Flux.1 Kontext Dev ワークフロー + +### Flux Kontext Dev 画像編集 (`flux_kontext_dev_basic`) + +キャラクターの一貫性を保ち、他の部分に影響を与えずに特定の部分を編集し、元のスタイルを維持するスマートな画像編集です。 + +Flux Kontext Dev ワークフロープレビュー + + + + Comfy Cloud で開く + + + JSONをダウンロードするか、テンプレートライブラリで「Flux Kontext Dev」を検索してください + + + +**入力素材** + +このファイルを対応する `LoadImage` ノードにアップロードしてください。 + + + + `LoadImage` ノード 190 · `flux_kontext_dev_basic_input_image.jpg` + + + +
+ flux_kontext_dev_basic_input_image.jpg +
+ +**出力例** + +
+ 入力画像 + Flux Kontext Dev 出力例 +
+ +このワークフローでは、`Load Image(from output)` ノードを使用して編集する画像を読み込み、編集後の画像にアクセスして複数回の編集を行いやすくしています。 + +### 1. ワークフローと入力画像のダウンロード + +### 2. ワークフローをステップバイステップで実行 + +![ワークフローステップガイド](/images/tutorial/flux/flux_1_kontext_dev_basic_step_guide.jpg) +画像内の番号を参考に、ワークフローの実行を完了してください。 + +1. `Load Diffusion Model` ノードで、`flux1-dev-kontext_fp8_scaled.safetensors` モデルを読み込みます。 +2. `DualCLIP Load` ノードで、`clip_l.safetensors` と `t5xxl_fp16.safetensors` または `t5xxl_fp8_e4m3fn_scaled.safetensors` が読み込まれていることを確認してください。 +3. `Load VAE` ノードで、`ae.safetensors` モデルが読み込まれていることを確認してください。 +4. `Load Image(from output)` ノードで、提供された入力画像を読み込みます。 +5. `CLIP Text Encode` ノードで、プロンプトを変更します。英語のみ対応しています。 +6. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します。 + + diff --git a/ja/tutorials/flux/flux-1-text-to-image.mdx b/ja/tutorials/flux/flux-1-text-to-image.mdx index 573035d7f..f84f77674 100644 --- a/ja/tutorials/flux/flux-1-text-to-image.mdx +++ b/ja/tutorials/flux/flux-1-text-to-image.mdx @@ -2,14 +2,15 @@ title: "ComfyUI Flux.1 テキストから画像へワークフローの例" description: "本ガイドでは、Flux.1 モデルについて簡潔に紹介し、フルバージョンおよび FP8 Checkpoint バージョンを含む、Flux.1 モデルを用いたテキストから画像への生成方法を解説します。" sidebarTitle: "Flux.1 テキストから画像へ" -translationSourceHash: 04546864 +translationSourceHash: add47dbc translationFrom: tutorials/flux/flux-1-text-to-image.mdx translationBlockHashes: "_intro": 36e44c17 - "Flux.1 Full Version Text-to-Image Example": 78064d53 - "Flux.1 FP8 Checkpoint Version Text-to-Image Example": daf5c58e + "Flux.1 Full Version Text-to-Image Example": afb29c94 + "Flux.1 FP8 Checkpoint Version Text-to-Image Example": 5656f14b --- + ![Flux](/images/tutorial/flux/flux_example.png) Flux は、現時点で最も大規模なオープンソースのテキストから画像へ生成するモデルの一つであり、120億（12B）パラメータを有し、オリジナルファイルサイズは約23GBです。このモデルは、元 Stable Diffusion チームのメンバーによって設立された [Black Forest Labs](https://blackforestlabs.ai/) が開発しました。 Flux は、優れた画像品質と高い柔軟性で知られており、高品質かつ多様な画像を生成できます。 @@ -45,9 +46,11 @@ Flux は、優れた画像品質と高い柔軟性で知られており、高品 ![Flux Agreement](/images/tutorial/flux/flux_agreement.jpg) +### Flux.1 Dev + ### Flux.1 Dev fp8: Text to Image (`flux_dev_checkpoint_example`) -Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices with limited VRAM, requires only one model file, but image quality is slightly reduced compared to the full version. +Flux.1 Dev fp8 量子化バージョンを使用して画像を生成します。VRAM が限られたデバイスに適しており、モデルファイルは1つだけ必要ですが、画像品質は完全版に比べてわずかに低下します。 Flux.1 Dev fp8 ワークフロープレビュー @@ -56,7 +59,7 @@ Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices wit Comfy Cloud でこのワークフローを実行 - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Flux.1 Dev fp8」を検索 @@ -64,22 +67,22 @@ Generate images using Flux.1 Dev fp8 quantized version. Suitable for devices wit ![Flux.1 Dev fp8 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_checkpoint_example.png) -#### 1. Workflow File +#### 1. ワークフローファイル -#### 2. Manual Model Installation +#### 2. 手動モデルインストール -- The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) agreement before downloading via browser. -- If your VRAM is low, you can try using [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) to replace the `t5xxl_fp16.safetensors` file. +- `flux1-dev.safetensors` ファイルは、ブラウザでダウンロードする前に [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) の契約に同意する必要があります。 +- VRAM が少ない場合は、[t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) を使用して `t5xxl_fp16.safetensors` ファイルを置き換えることを試せます。 -Please download the following model files: +以下のモデルファイルをダウンロードしてください： - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAMが32GB以上の場合は推奨。 - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) +- [flux1-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/flux1-dev.safetensors) -Storage location: +保存場所： ``` ComfyUI/ ├── models/ @@ -92,52 +95,89 @@ ComfyUI/ │ └── flux1-dev.safetensors ``` -#### 3. Steps to Run the Workflow +#### 3. ワークフローの実行手順 -Please refer to the image below to ensure all model files are loaded correctly +以下の画像を参照して、すべてのモデルファイルが正しく読み込まれていることを確認してください。 ![ComfyUI Flux Dev Workflow](/images/tutorial/flux/flow_diagram_flux_dev_t5fp16.jpg) -1. Ensure the `DualCLIPLoader` node has the following models loaded: +1. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認してください： - clip_name1: t5xxl_fp16.safetensors - clip_name2: clip_l.safetensors -2. Ensure the `Load Diffusion Model` node has `flux1-dev.safetensors` loaded -3. Make sure the `Load VAE` node has `ae.safetensors` loaded -4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow +2. `Load Diffusion Model` ノードに `flux1-dev.safetensors` が読み込まれていることを確認してください +3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認してください +4. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します -Thanks to Flux's excellent prompt following capability, we don't need any negative prompts +Fluxの優れたプロンプト追従機能のおかげで、ネガティブプロンプトは不要です +### Flux.1 Schnell + ### Flux.1 Schnell FP8 (`flux_schnell`) -Quickly generate images with Flux.1 Schnell fp8 quantized version. Ideal for low-end hardware, requires only 4 steps to generate images. +Flux.1 Schnell fp8 量子化バージョンを使用して画像を素早く生成します。ローエンドハードウェアに最適で、画像生成に必要なステップはわずか4ステップです。 -Flux.1 Schnell FP8 checkpoint ワークフロープレビュー +Flux.1 Schnell FP8 ワークフロープレビュー Comfy Cloud でこのワークフローを実行 - - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + + JSON をダウンロード、またはテンプレートライブラリで「Flux.1 Schnell」を検索 -Please download the image below and drag it into ComfyUI to load the workflow. +#### 1. ワークフローファイル -Please download [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. +#### 2. 手動モデルインストール -Ensure that the corresponding `Load Checkpoint` node loads `flux1-schnell-fp8.safetensors`, and you can try to run the workflow. + +このワークフローでは、Flux1 Dev バージョンのワークフローとは2つのモデルファイルのみが異なります。t5xxl については、より良い結果を得るために fp16 バージョンを引き続き使用できます。 +- **t5xxl_fp16.safetensors** -> **t5xxl_fp8.safetensors** +- **flux1-dev.safetensors** -> **flux1-schnell.safetensors** + + +完全なモデルファイルリスト： +- [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors?download=true) +- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) +- [flux1-schnell.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/flux1-schnell.safetensors) + +ファイルの保存場所： +``` +ComfyUI/ +├── models/ +│ ├── text_encoders/ +│ │ ├── clip_l.safetensors +│ │ └── t5xxl_fp8_e4m3fn.safetensors +│ ├── vae/ +│ │ └── ae.safetensors +│ └── diffusion_models/ +│ └── flux1-schnell.safetensors +``` + +#### 3. ワークフローの実行手順 + +![Flux Schnell バージョンのワークフロー](/images/tutorial/flux/flow_diagram_flux_schnell_t5fp8.jpg) + +1. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認してください： + - clip_name1: t5xxl_fp8_e4m3fn.safetensors + - clip_name2: clip_l.safetensors +2. `Load Diffusion Model` ノードに `flux1-schnell.safetensors` が読み込まれていることを確認してください +3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認してください +4. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します ## Flux.1 FP8 Checkpoint バージョンによるテキストから画像へ生成の例 -FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したものです。 -ある程度、このバージョンの品質は fp16 バージョンよりも劣りますが、その一方で必要な VRAM 量が少なくなり、試行運用のためにインストールするモデルファイルは1つだけで済みます。 +FP8 バージョンは、元の Flux.1 fp16 バージョンを量子化したものです。ある程度、このバージョンの品質は fp16 バージョンよりも劣りますが、その一方で必要な VRAM 量が少なくなり、試行のためにインストールするモデルファイルは1つだけで済みます。 + +### Flux.1 Dev -### Flux.1 Dev: Text to Image (`flux_dev_full_text_to_image`) +### Flux.1 Dev: テキストから画像へ (`flux_dev_full_text_to_image`) -Generate high-quality images with Flux Dev full version. Requires larger VRAM and multiple model files, but provides the best prompt following capability. +Flux Dev フルバージョンで高品質の画像を生成します。より多くの VRAM と複数のモデルファイルが必要ですが、プロンプトへの追従能力に優れています。 Flux.1 Dev text-to-image ワークフロープレビュー @@ -146,7 +186,7 @@ Generate high-quality images with Flux Dev full version. Requires larger VRAM an Comfy Cloud でこのワークフローを実行 - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Flux.1 Dev FP8」を検索 @@ -154,17 +194,30 @@ Generate high-quality images with Flux Dev full version. Requires larger VRAM an ![Flux.1 Dev FP8 checkpoint 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux_dev_full_text_to_image.png) -Please download the image below and drag it into ComfyUI to load the workflow. +下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 -Please download [flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) and save it to the `ComfyUI/models/checkpoints/` directory. +[flux1-dev-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 -Ensure that the corresponding `Load Checkpoint` node loads `flux1-dev-fp8.safetensors`, and you can try to run the workflow. +対応する `Load Checkpoint` ノードが `flux1-dev-fp8.safetensors` を読み込んでいることを確認し、ワークフローの実行を試みてください。 ### Flux.1 Schnell -下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 +### Flux.1 Schnell FP8 (`flux_schnell`) + +Flux.1 Schnell fp8 量子化バージョンで素早く画像を生成します。ローエンドハードウェアに最適で、わずか4ステップで画像を生成できます。 + +Flux.1 Schnell FP8 checkpoint ワークフロープレビュー + + + + Comfy Cloud でこのワークフローを実行 + + + JSON をダウンロード、またはテンプレートライブラリで「Flux.1 Schnell FP8」を検索 + + -![Flux Schnell fp8 Checkpoint バージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/flux/text-to-image/flux_schnell_fp8.png) +下記の画像をダウンロードし、ComfyUI にドラッグ＆ドロップしてワークフローを読み込んでください。 [flux1-schnell-fp8.safetensors](https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors?download=true) をダウンロードし、`ComfyUI/models/checkpoints/` ディレクトリに保存してください。 diff --git a/ja/tutorials/flux/flux-1-uso.mdx b/ja/tutorials/flux/flux-1-uso.mdx index 4ca12ce70..95042a6b2 100644 --- a/ja/tutorials/flux/flux-1-uso.mdx +++ b/ja/tutorials/flux/flux-1-uso.mdx @@ -2,7 +2,7 @@ title: "ByteDance USO ComfyUI ネイティブワークフローの例" description: "ByteDance の USO モデルを用いた統一スタイル・主体駆動型生成" sidebarTitle: "ByteDance USO" -translationSourceHash: e41f0254 +translationSourceHash: f50cdbb4 translationFrom: tutorials/flux/flux-1-uso.mdx --- @@ -25,24 +25,24 @@ USO は以下の3つの主要なアプローチをサポートします： -### Flux.1 Dev USO Reference Image Generation (`flux1_dev_uso_reference_image_gen`) +### Flux.1 Dev USO 参照画像生成ワークフロー（`flux1_dev_uso_reference_image_gen`） -Use reference images to control both style and subject. Keep your character's face while changing artistic style, or apply artistic styles to new scenes. +参照画像を使用してスタイルと主体の両方を制御します。芸術スタイルを変更しながらキャラクターの顔を保持したり、新しいシーンに芸術スタイルを適用したりできます。 -Flux.1 Dev USO reference image ワークフロープレビュー +Flux.1 Dev USO 参照画像ワークフロープレビュー Comfy Cloud でこのワークフローを実行 - Download the workflow JSON and drag it into ComfyUI + ワークフローのJSONをダウンロードし、ComfyUIにドラッグ＆ドロップしてください **入力素材** -次の `\1` ノードにこのファイルをアップロード： +対応する `LoadImage` ノードにこのファイルをアップロード： @@ -61,6 +61,8 @@ Use reference images to control both style and subject. Keep your character's fa Flux.1 Dev USO 出力例
+### 1. ワークフローと入力 + ### 2. モデルのダウンロードリンク **checkpoints** @@ -118,7 +120,7 @@ Use reference images to control both style and subject. Keep your character's fa 同一のワークフロー内に、スタイル参照のみを用いるバージョンも提供しています。 ![ワークフロー](/images/tutorial/flux/flux1_uso_reference_image_gen_style_reference_only.jpg) -唯一の違いは、`content reference` ノードを `Empty Latent Image` ノードに置き換え、必要な画像サイズを生成することです。 +唯一の違いは、`content reference` ノードを `Empty Latent Image` ノードに置き換えたことです。 2. また、`Style Reference` グループ全体を `Ctrl+B` でバイパスすることで、このワークフローをテキストから画像を生成する（text-to-image）ワークフローとしても利用可能です。つまり、本ワークフローには以下の4種類のバリエーションがあります： - コンテンツ（主体）参照のみを使用 diff --git a/ja/tutorials/flux/flux-2-dev.mdx b/ja/tutorials/flux/flux-2-dev.mdx index 2ea935dd8..bcb2ecade 100644 --- a/ja/tutorials/flux/flux-2-dev.mdx +++ b/ja/tutorials/flux/flux-2-dev.mdx @@ -2,16 +2,19 @@ title: "ComfyUI Flux.2 Dev の例" description: "本ガイドでは、Flux.2 モデルについて簡単に紹介し、ComfyUI で Flux.2 Dev モデルを用いたテキストから画像への生成手順を解説します。" sidebarTitle: "Flux.2 Dev" -translationSourceHash: 5465e1e9 +translationSourceHash: 755b6620 translationFrom: tutorials/flux/flux-2-dev.mdx translationBlockHashes: "_intro": 8984dcfe "About FLUX.2": 95417ef8 - "Single image generation workflow": 85a6ea77 - "Multi-image reference workflow": 416924a0 + "Flux.2 Dev (`image_flux2`)": 831ed94c + "Flux.2 Dev Text to Image (`image_flux2_text_to_image`)": d5bb064e + "Product Mockup (`image_flux2_fp8`)": 0e4e9099 "Model links": f4c1677b --- + + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -36,32 +39,92 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' -## 単一画像生成ワークフロー +## Flux.2 Dev (`image_flux2`) -FLUX.2 Dev を用いた基本的なテキストから画像への生成ワークフローです。 +複数の参照画像を一貫して使用し、プロフェッショナルなテキストレンダリングでフォトリアリスティックな画像を生成します。 + +Flux.2 Dev ワークフロープレビュー - このワークフローを直接 Comfy Cloud で開きます + このワークフローをComfy Cloudで直接開く - ローカル環境で使用するための JSON ワークフロー・ファイルをダウンロード + ローカルで使用するためのJSONワークフローファイルをダウンロード + + + +**入力素材** + +このファイルを `LoadImage` ノード **46** にアップロードします： + + + + `LoadImage` node 46 · `image_flux2_input_image.png` + + + +
+ Flux.2 Dev の入力画像 + Flux.2 Dev の出力例 +
+ +## Flux.2 Dev テキストから画像へ (`image_flux2_text_to_image`) + +照明、素材、リアルなディテールが強化されたテキストから画像への変換です。入力画像は不要です。 + +Flux.2 Dev テキストから画像へのワークフロープレビュー + + + + このワークフローをComfy Cloudで直接開く + + + ローカルで使用するJSONワークフローファイルをダウンロード -## 複数画像参照ワークフロー +**出力例** + +![Flux.2 Dev テキストから画像への出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_text_to_image.png) + +## Product Mockup (`image_flux2_fp8`) -2 枚の参照画像を用いるワークフローの例です。この実装を拡張することで、さらに多くの参照画像に対応させることも可能です。 +マルチリファレンスの一貫性を利用して、パッケージ、マグカップ、その他の製品にデザインパターンを適用し、プロダクトモックアップを作成します。 + +Flux.2 Dev プロダクトモックアップワークフロープレビュー - このワークフローを直接 Comfy Cloud で開きます + このワークフローをComfy Cloudで直接開く + + + ローカルで使用するためにJSONワークフローファイルをダウンロード + + + +**入力素材** + +これらのファイルを該当する `LoadImage` ノードにアップロードしてください： + + + + `LoadImage` ノード 42 · `image_flux2_input_ref_image.png` - - ローカル環境で使用するための JSON ワークフロー・ファイルをダウンロード + + `LoadImage` ノード 46 · `image_flux2_input_Illustration.png` +
+ Flux.2 Dev 用入力参照画像 + Flux.2 Dev 用入力イラスト +
+ +**出力例** + +![Flux.2 Dev プロダクトモックアップ出力](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_fp8.png) + ## モデルのダウンロードリンク **text_encoders** @@ -87,3 +150,4 @@ FLUX.2 Dev を用いた基本的なテキストから画像への生成ワーク │ │ └── flux2_dev_fp8mixed.safetensors │ └── 📂 vae/ │ └── flux2-vae.safetensors +``` diff --git a/ja/tutorials/flux/flux-2-klein.mdx b/ja/tutorials/flux/flux-2-klein.mdx index 73197ea0c..75c1f2404 100644 --- a/ja/tutorials/flux/flux-2-klein.mdx +++ b/ja/tutorials/flux/flux-2-klein.mdx @@ -2,24 +2,26 @@ title: "ComfyUI Flux.2 Klein 4B ガイド" description: "FLUX.2 [klein] 4B の概要と、ComfyUI におけるテキストから画像への生成および画像編集ワークフローの実行方法について解説します。" sidebarTitle: "Flux.2 Klein" -translationSourceHash: 2b9ccba8 +translationSourceHash: 5f6e7716 translationFrom: tutorials/flux/flux-2-klein.mdx translationBlockHashes: "_intro": 980a68e5 - "About FLUX.2 [klein]": 77f716a7 - "Flux.2 Klein 4B Workflows": 8c4a5de1 + "About FLUX.2 [klein]": bb088cd5 + "Flux.2 Klein 4B Workflows": b750ef37 "Flux.2 Klein 4B Model Downloads": b438d16c - "Flux.2 Klein 9B Workflows": fbbce562 + "Flux.2 Klein 9B Workflows": b3c7871b "Flux.2 Klein 9B Model Downloads": 0f38ad48 --- + + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## FLUX.2 [Klein] について -FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15日現在）最も高速なモデルであり、テキストから画像への生成と画像編集を、1つのコンパクトなアーキテクチャに統合しています。インタラクティブなワークフロー、即時プレビュー、および低遅延が求められるアプリケーション向けに設計されており、蒸留版モデルでは約1秒でエンドツーエンドの推論が可能でありながら、単一参照および複数参照による画像編集においても高品質な結果を維持します。 +FLUX.2 [Klein] は、Flux ファミリーの中で最も高速なモデルであり、テキストから画像への生成と画像編集を、1つのコンパクトなアーキテクチャに統合しています。インタラクティブなワークフロー、即時プレビュー、および低遅延が求められるアプリケーション向けに設計されており、蒸留版モデルでは約1秒でエンドツーエンドの推論が可能でありながら、単一参照および複数参照による画像編集においても高品質な結果を維持します。 **モデルの主な特長：** - 2種類の4Bモデル：最大限の柔軟性とファインチューニングに対応する「Base（非蒸留）」と、速度重視の展開に適した「Distilled（4ステップ）」 @@ -30,17 +32,93 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 ## Flux.2 Klein 4B のワークフロー +### Flux.2 [Klein] 4B：テキストから画像へ（`image_flux2_klein_text_to_image`） + +Flux.2 Klein 4B のテキストから画像へのワークフロープレビュー + + + + Comfy Cloud でこのワークフローを実行 + + + Flux.2 Klein 4B 用のテキストから画像へのワークフローをダウンロード。 + + + +**出力例** + +![Flux.2 Klein 4B のテキストから画像への例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_klein_text_to_image.png) + +### Flux.2 [Klein] 4B：画像編集（`image_flux2_klein_image_edit_4b_base`） + +Flux.2 Klein 4B の画像編集ベースワークフロープレビュー + + + + Comfy Cloud でこのワークフローを実行 + + + 4B ベースモデルを使用した画像編集ワークフローをダウンロード。 + + + +**入力素材** + +以下のファイルを該当する `LoadImage` ノードにアップロードしてください： + - - Flux.2 Klein 4B 用のテキストから画像への生成ワークフローをダウンロードします。 + + `LoadImage` ノード 76 · `robed_women.png` - - 4B Base モデルを使用した画像編集ワークフローをダウンロードします。 + + `LoadImage` ノード 81 · `pink_tone_chair.png` - - 4B 蒸留版による高速画像編集ワークフローをダウンロードします。 + + +
+ robed_women.png + pink_tone_chair.png +
+ +**出力例** + +![Flux.2 Klein 4B 画像編集ベース例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/image_flux2_klein_image_edit_4b_base.png) + +### Flux.2 [Klein] 4B Distilled：画像編集（`image_flux2_klein_image_edit_4b_distilled`） + +Flux.2 Klein 4B 画像編集蒸留版ワークフロープレビュー + + + + Comfy Cloud でこのワークフローを実行 + + + 高速な蒸留版 4B 画像編集ワークフローをダウンロード。 + + + +**入力素材** + +以下のファイルを該当する `LoadImage` ノードにアップロードしてください： + + + + `LoadImage` ノード 76 · `handbag_white.png` + + + `LoadImage` ノード 81 · `comfy_logo_blue.png` + +
+ handbag_white.png + comfy_logo_blue.png +
+ +**出力例** + +![Flux.2 Klein 4B 画像編集蒸留版例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_klein_image_edit_4b_distilled.png) + ## Flux.2 Klein 4B モデルのダウンロード @@ -71,19 +149,95 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 │ └── flux2-vae.safetensors ``` -## Flux.2 Klein 9B のワークフロー +## Flux.2 Klein 9B ワークフロー + +### Flux.2 [Klein] 9B: テキストから画像へ (`image_flux2_text_to_image_9b`) + +Flux.2 Klein 9B テキストから画像へのワークフロー プレビュー + + + + このワークフローを Comfy Cloud で実行します。 + + + Flux.2 Klein 9B 用のテキストから画像へのワークフローをダウンロードします。 + + + +**出力例** + +![Flux.2 Klein 9B テキストから画像への出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_text_to_image_9b.png) + +### Flux.2 [Klein] 9B: 画像編集 (`image_flux2_klein_image_edit_9b_base`) + +Flux.2 Klein 9B 画像編集ベースワークフロー プレビュー - - Flux.2 Klein 9B 用のテキストから画像への生成ワークフローをダウンロードします。 + + このワークフローを Comfy Cloud で実行します。 + + + 9B ベースモデルを使用した画像編集ワークフローをダウンロードします。 - - 9B Base モデルを使用した画像編集ワークフローをダウンロードします。 + + +**入力素材** + +以下のファイルを対応する `LoadImage` ノードにアップロードします。 + + + + `LoadImage` ノード 76 · `car_interior_white.jpeg` - - 9B 蒸留版による高速画像編集ワークフローをダウンロードします。 + + `LoadImage` ノード 81 · `comfy_logo_blue.png` + +
+ car_interior_white.jpeg + comfy_logo_blue.png +
+ +**出力例** + +![Flux.2 Klein 9B 画像編集ベースの出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/thumbnail/image_flux2_klein_image_edit_9b_base.png) + +### Flux.2 [Klein] 9B Distilled: 画像編集 (`image_flux2_klein_image_edit_9b_distilled`) + +Flux.2 Klein 9B 画像編集蒸留版ワークフロー プレビュー + + + + このワークフローを Comfy Cloud で実行します。 + + + 高速な蒸留版 9B 画像編集ワークフローをダウンロードします。 + + + +**入力素材** + +以下のファイルを対応する `LoadImage` ノードにアップロードします。 + + + + `LoadImage` ノード 76 · `bold_outfit_woman.jpeg` + + + `LoadImage` ノード 121 · `handbag_white.png` + + + +
+ bold_outfit_woman.jpeg + handbag_white.png +
+ +**出力例** + +![Flux.2 Klein 9B 画像編集蒸留版の出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_flux2_klein_image_edit_9b_distilled.png) + ## Flux.2 Klein 9B モデルのダウンロード @@ -116,3 +270,4 @@ FLUX.2 [Klein] は、Flux ファミリーの中で現時点で（2026年1月15 │ │ └── qwen_3_8b_fp8mixed.safetensors │ └── 📂 vae/ │ └── flux2-vae.safetensors +``` diff --git a/ja/tutorials/flux/flux1-krea-dev.mdx b/ja/tutorials/flux/flux1-krea-dev.mdx index 69fdcc34e..1d46f49ae 100644 --- a/ja/tutorials/flux/flux1-krea-dev.mdx +++ b/ja/tutorials/flux/flux1-krea-dev.mdx @@ -1,8 +1,8 @@ --- -title: "Flux.1 Krea Dev ComfyUI ワークフロー チュートリアル" -description: "Black Forest Labs が Krea と共同で開発した、最高品質のオープンソース FLUX モデルです。独特な美意識と自然なディテールに重点を置き、「AIらしさ」を回避し、卓越したリアリズムと画像品質を実現します。" +title: "Flux.1 Krea Dev ComfyUI ワークフローチュートリアル" +description: "Black Forest Labs が Krea と共同で開発した、最高品質のオープンソース FLUX モデルです。独自の美意識と自然なディテールに重点を置き、「AIらしさ」を避け、卓越したリアリズムと画像品質を実現します。" sidebarTitle: "Flux.1 Krea Dev" -translationSourceHash: 202e24cf +translationSourceHash: bc3ec641 translationFrom: tutorials/flux/flux1-krea-dev.mdx --- @@ -10,21 +10,25 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ![Flux.1 Krea Dev ポスター](/images/tutorial/flux/flux_1_krea_dev_poster.jpg) -[Flux.1 Krea Dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev) は、Black Forest Labs（BFL）と Krea が共同で開発した先進的なテキストから画像を生成するモデルです。これは現在、テキストから画像を生成することに特化した、最高品質のオープンソース FLUX モデルです。 +[Flux.1 Krea Dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev) は、Black Forest Labs（BFL）と Krea が共同で開発した先進的なテキストから画像を生成するモデルです。現在、テキストから画像生成に特化した最高品質のオープンソース FLUX モデルです。 **モデルの特長** -- **独特な美意識**: 一般的な「AIらしい外観」を避け、独自の美意識を持つ画像の生成に焦点を当てています -- **自然なディテール**: 過剰に明るいハイライト（ブローアウト）を発生させず、自然なディテール表現を維持します +- **独自の美意識**: よくある「AIらしい外観」を避け、独自の美意識を持つ画像の生成に重点を置いています +- **自然なディテール**: 白飛び（ハイライトが飛ぶこと）を発生させず、自然なディテール表現を維持します - **卓越したリアリズム**: 極めて高いリアリズムと画像品質を提供します - **完全互換アーキテクチャ**: FLUX.1 [dev] と完全に互換性のあるアーキテクチャ設計です -**モデルのライセンス** +**モデルライセンス** 本モデルは、[flux-1-dev-non-commercial-license](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/LICENSE.md) の下で公開されています。 +## Flux.1 Krea Dev ComfyUI ワークフロー + + + ### Flux.1 Krea Dev (`flux1_krea_dev`) -A fine-tuned FLUX model pushing photorealism to the max. +フォトリアリズムを極限まで追求した、ファインチューニング済みの FLUX モデルです。 Flux.1 Krea Dev ワークフロープレビュー @@ -33,7 +37,7 @@ A fine-tuned FLUX model pushing photorealism to the max. Comfy Cloud でこのワークフローを実行 - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Flux.1 Krea Dev」を検索 @@ -41,62 +45,62 @@ A fine-tuned FLUX model pushing photorealism to the max. ![Flux.1 Krea Dev 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/flux1_krea_dev.png) -#### 1. Workflow Files +#### 1. ワークフローファイル -#### 2. Manual Model Installation +#### 2. 手動でのモデルインストール -Please download the following model files: -**Diffusion model** +以下のモデルファイルをダウンロードしてください。 +**拡散モデル** - [flux1-krea-dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/FLUX.1-Krea-dev_ComfyUI/blob/main/split_files/diffusion_models/flux1-krea-dev_fp8_scaled.safetensors) -If you want to pursue higher quality and have enough VRAM, you can try the original model weights +より高品質を追求し、十分な VRAM がある場合は、オリジナルのモデル重みも試せます。 - [flux1-krea-dev.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/blob/main/flux1-krea-dev.safetensors) -The `flux1-dev.safetensors` file requires agreeing to the [black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) agreement before downloading via browser. +`flux1-dev.safetensors` ファイルをブラウザからダウンロードするには、[black-forest-labs/FLUX.1-Krea-dev](https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev/) の利用規約に同意する必要があります。 -If you have used Flux related workflows before, the following models are the same and don't need to be downloaded again +以前に Flux 関連のワークフローを使用したことがある場合、以下のモデルは共通しており、再度ダウンロードする必要はありません。 -**Text encoders** +**テキストエンコーダー** - [clip_l.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors?download=true) -- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) Recommended when your VRAM is greater than 32GB. -- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) For Low VRAM +- [t5xxl_fp16.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors?download=true) VRAM が 32GB を超える場合に推奨 +- [t5xxl_fp8_e4m3fn.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors) 低 VRAM 向け **VAE** - [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors?download=true) -File save location: +ファイルの保存先: ``` ComfyUI/ ├── models/ │ ├── diffusion_models/ -│ │ └── flux1-krea-dev_fp8_scaled.safetensors or flux1-krea-dev.safetensors +│ │ └── flux1-krea-dev_fp8_scaled.safetensors または flux1-krea-dev.safetensors │ ├── text_encoders/ │ │ ├── clip_l.safetensors -│ │ └── t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors +│ │ └── t5xxl_fp16.safetensors または t5xxl_fp8_e4m3fn.safetensors │ ├── vae/ │ │ └── ae.safetensors ``` -#### 3. Step-by-step Verification to Ensure Workflow Runs Properly +#### 3. ワークフローが正しく動作することを確認するための段階的な検証 - For low VRAM users, this model may not run smoothly on your device, you can wait for the community to provide FP8 or GGUF version. + 低 VRAM のユーザーの場合、このモデルはお使いのデバイスではスムーズに動作しない可能性があります。コミュニティが FP8 または GGUF バージョンを提供するまでお待ちください。 -Please refer to the image below to ensure all model files have been loaded correctly +以下の画像を参考に、すべてのモデルファイルが正しく読み込まれていることを確認してください。 -![ComfyUI Flux Krea Dev Workflow](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) +![ComfyUI Flux Krea Dev ワークフロー](/images/tutorial/flux/flux_1_krea_dev_guide.jpg) -1. Ensure that `flux1-krea-dev_fp8_scaled.safetensors` or `flux1-krea-dev.safetensors` is loaded in the `Load Diffusion Model` node - - `flux1-krea-dev_fp8_scaled.safetensors` is recommended for low VRAM users - - `flux1-krea-dev.safetensors` is the original weights, if you have enough VRAM like 24GB you can use it for better quality -2. Ensure the following models are loaded in the `DualCLIPLoader` node: - - clip_name1: t5xxl_fp16.safetensors or t5xxl_fp8_e4m3fn.safetensors +1. `Load Diffusion Model` ノードに `flux1-krea-dev_fp8_scaled.safetensors` または `flux1-krea-dev.safetensors` が読み込まれていることを確認します + - `flux1-krea-dev_fp8_scaled.safetensors` は低 VRAM ユーザーに推奨されます + - `flux1-krea-dev.safetensors` はオリジナルの重みです。24GB など十分な VRAM がある場合は、より高品質を得るために使用できます +2. `DualCLIPLoader` ノードに以下のモデルが読み込まれていることを確認します: + - clip_name1: t5xxl_fp16.safetensors または t5xxl_fp8_e4m3fn.safetensors - clip_name2: clip_l.safetensors -3. Ensure that `ae.safetensors` is loaded in the `Load VAE` node -4. Click the `Queue` button, or use the shortcut `Ctrl(cmd) + Enter` to run the workflow +3. `Load VAE` ノードに `ae.safetensors` が読み込まれていることを確認します +4. `Queue` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用してワークフローを実行します \ No newline at end of file diff --git a/ja/tutorials/image/anima/anima.mdx b/ja/tutorials/image/anima/anima.mdx index 6ea6e3066..5afd4f91d 100644 --- a/ja/tutorials/image/anima/anima.mdx +++ b/ja/tutorials/image/anima/anima.mdx @@ -2,15 +2,16 @@ title: "Anima Base v1 ComfyUI ワークフロー例" description: "Anima は CircleStone Labs / Comfy Org による 2B パラメータのテキストから画像生成モデルで、アニメや非フォトリアリスティックなイラスト生成に最適化されています。" sidebarTitle: "Anima Base v1" -translationSourceHash: 6668c717 +translationSourceHash: f37b6cf2 translationFrom: tutorials/image/anima/anima.mdx translationBlockHashes: "_intro": 41a50d20 - "Available workflows": 4d4fb0d6 + "Available workflows": ad35ec4b "Anima model downloads": 06a278be "Limitations": 0adbf360 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Anima** は [CircleStone Labs](https://huggingface.co/circlestone-labs/Anima) が Comfy Org と協力して開発したオープンなテキストから画像生成モデルです。**20 億** パラメータを備え、高品質な **アニメおよび非フォトリアリスティック** な画像を生成するように設計されており、キャラクター、シーン、コンセプトアートに最適です。 @@ -27,7 +28,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ## 利用可能なワークフロー -Anima は 2 つのワークフローを提供しています——標準的な使用向けのベース版と早期アクセス向けのプレビュー版です。 +Anima は 2 つのワークフローを提供しています：スタンダードな使用向けのベース版と早期アクセス向けのプレビュー版です。 両方のワークフローは **サブグラフ**（Subgraph）ノードを使用してテキストから画像への生成パイプラインを管理します。サブグラフを開いて内部ノードを確認・カスタマイズできます。 @@ -37,7 +38,7 @@ Anima は 2 つのワークフローを提供しています——標準的な ### Anima Base v1: Text to Image (`image_anima_base_v1`) -Input a text prompt describing an anime or artistic illustration. Generate a non-photorealistic image focused on anime concepts, characters, or styles. +アニメやアートイラストを説明するテキストプロンプトを入力します。アニメのコンセプト、キャラクター、スタイルに焦点を当てた非写実的な画像を生成します。 Anima Base v1 text-to-image ワークフロープレビュー @@ -46,7 +47,7 @@ Input a text prompt describing an anime or artistic illustration. Generate a non Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Anima Base v1」を検索 @@ -54,16 +55,16 @@ Input a text prompt describing an anime or artistic illustration. Generate a non ![Anima Base v1 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_base_v1.png) -#### Get started +#### はじめる -1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) -2. Go to **Template** and search for **Anima Base v1** -3. Select the **Anima Base v1: Text to Image** workflow -4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** +1. ComfyUI を最新バージョンにアップデートするか、[Comfy Cloud](https://cloud.comfy.org/?template=image_anima_base_v1&utm_source=docs&utm_medium=referral&utm_campaign=anima) を使用します +2. **テンプレート**に移動し、**Anima Base v1** を検索します +3. **Anima Base v1: Text to Image** ワークフローを選択します +4. 不足しているモデルをダウンロードし（[モデルのダウンロード](#anima-model-downloads)を参照）、プロンプトを入力して、**キュー**をクリックします ### Anima Preview: Anime Text-to-Image Generation (`image_anima_preview`) -Input a text prompt to generate an anime-style image using the Anima model. Configure settings like steps and CFG scale to control the output. +テキストプロンプトを入力して、Anima モデルを使用してアニメ風の画像を生成します。ステップ数や CFG スケールなどの設定を行い、出力を制御します。 Anima Preview text-to-image ワークフロープレビュー @@ -72,7 +73,7 @@ Input a text prompt to generate an anime-style image using the Anima model. Conf Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Anima Preview」を検索 @@ -80,12 +81,12 @@ Input a text prompt to generate an anime-style image using the Anima model. Conf ![Anima Preview 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_anima_preview.png) -#### Get started +#### はじめる -1. Update ComfyUI to the latest version or use [Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) -2. Go to **Template** and search for **Anima Preview** -3. Select the **Anima Anime Text-to-Image Generation** workflow -4. Download any missing models (see [model downloads](#anima-model-downloads)), enter your prompt, and click **Queue** +1. ComfyUI を最新バージョンにアップデートするか、[Comfy Cloud](https://cloud.comfy.org/?template=image_anima_preview&utm_source=docs&utm_medium=referral&utm_campaign=anima) を使用します +2. **テンプレート**に移動し、**Anima Preview** を検索します +3. **Anima Anime Text-to-Image Generation** ワークフローを選択します +4. 不足しているモデルをダウンロードし（[モデルのダウンロード](#anima-model-downloads)を参照）、プロンプトを入力して、**キュー**をクリックします ## Anima モデルのダウンロード diff --git a/ja/tutorials/image/boogu/boogu-image-0.1.mdx b/ja/tutorials/image/boogu/boogu-image-0.1.mdx index 5d0e1b811..7b75c26ed 100644 --- a/ja/tutorials/image/boogu/boogu-image-0.1.mdx +++ b/ja/tutorials/image/boogu/boogu-image-0.1.mdx @@ -10,6 +10,7 @@ translationBlockHashes: "Boogu-Image-0.1-Edit image editing workflow": 3c1752bb --- + import UpdateReminder from "/snippets/ja/tutorials/update-reminder.mdx" **Boogu-Image-0.1** は Apache-2.0 オープンソースの統合画像生成・編集モデルファミリーです。理解と生成を統合するシステムにより、写真、テキストレンダリング、スタイライゼーション、画像編集タスクで競争力のあるパフォーマンスを発揮します。 diff --git a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx index c9f7a17f2..2b90c55e5 100644 --- a/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx +++ b/ja/tutorials/image/cosmos/cosmos-predict2-t2i.mdx @@ -2,7 +2,7 @@ title: "Cosmos Predict2 テキストから画像へ（Text-to-Image）の ComfyUI 公式サンプル" description: "本ガイドでは、ComfyUI で Cosmos-Predict2 のテキストから画像へ（Text-to-Image）ワークフローを完了する方法を説明します" sidebarTitle: "Cosmos-Predict2" -translationSourceHash: 01df14e7 +translationSourceHash: 82b4623f translationFrom: tutorials/image/cosmos/cosmos-predict2-t2i.mdx --- @@ -11,7 +11,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' Cosmos-Predict2 は、NVIDIA が開発した次世代の物理世界向け基礎モデルであり、物理AIシナリオにおける高品質な視覚生成および予測タスクに特化して設計されています。 このモデルは、卓越した物理的正確性、環境との相互作用能力、および細部の再現性能を備えており、複雑な物理現象や動的なシーンをリアルにシミュレートすることが可能です。 -Cosmos-Predict2 は、テキストから画像へ（Text-to-Image）や動画から3Dワールドへ（Video-to-World）など、さまざまな生成手法をサポートしており、産業用シミュレーション、自動運転、都市計画、科学研究などの分野で広く活用されています。 +Cosmos-Predict2 は、テキストから画像へ（Text-to-Image）や動画から世界へ（Video-to-World）など、さまざまな生成手法をサポートしており、産業用シミュレーション、自動運転、都市計画、科学研究などの分野で広く活用されています。 GitHub: [Cosmos-predict2](https://github.com/nvidia-cosmos/cosmos-predict2) Hugging Face: [Cosmos-Predict2](https://huggingface.co/collections/nvidia/cosmos-predict2-68028efc052239369a0f2959) diff --git a/ja/tutorials/image/ernie-image/ernie-image.mdx b/ja/tutorials/image/ernie-image/ernie-image.mdx index 86667f861..209ea32c3 100644 --- a/ja/tutorials/image/ernie-image/ernie-image.mdx +++ b/ja/tutorials/image/ernie-image/ernie-image.mdx @@ -12,6 +12,7 @@ translationBlockHashes: "Examples": 0e2eb115 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **ERNIE-Image** は百度が開発したオープンなテキストから画像生成モデルで、Apache-2.0 ライセンスで公開されています。**8B** パラメータの拡散トランスフォーマー（DiT）をベースに構築されており、精密なテキストレンダリング、高い命令追従性、構造化された視覚生成を実現する高品質な画像生成が可能です。 diff --git a/ja/tutorials/image/hidream/hidream-e1.mdx b/ja/tutorials/image/hidream/hidream-e1.mdx index be8c9326e..0c0d89e0b 100644 --- a/ja/tutorials/image/hidream/hidream-e1.mdx +++ b/ja/tutorials/image/hidream/hidream-e1.mdx @@ -2,15 +2,16 @@ title: "ComfyUI ネイティブ HiDream-E1、E1.1 ワークフローの例" sidebarTitle: "HiDream-e1" description: "本ガイドでは、ComfyUI ネイティブ版 HiDream-I1 のテキストから画像を生成するワークフローの例について、理解と実行方法を説明します。" -translationSourceHash: 7554c00c +translationSourceHash: 77dc6db9 translationFrom: tutorials/image/hidream/hidream-e1.mdx translationBlockHashes: "_intro": 6829ae90 "HiDream E1 and E1.1 Workflow Related Models": 25f9249b - "HiDream E1.1 ComfyUI Native Workflow Example": 4437f406 - "HiDream E1 ComfyUI Native Workflow Example": 2dd0afb2 + "HiDream E1.1 ComfyUI Native Workflow Example": a5e35b6a + "HiDream E1 ComfyUI Native Workflow Example": 247a872f --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ![HiDream-E1 デモ](https://raw.githubusercontent.com/HiDream-ai/HiDream-E1/refs/heads/main/assets/demo.jpg) @@ -72,18 +73,18 @@ HiDream-E1 は、HiDream-ai 社が公式にオープンソース化したイン ## HiDream E1.1 の ComfyUI ネイティブ ワークフローの例 -### HiDream E1.1 Image Editing (`hidream_e1_1`) +### HiDream E1.1 画像編集 (`hidream_e1_1`) -Edit images with HiDream E1.1. Superior image quality and editing accuracy compared to HiDream-E1-Full. +HiDream E1.1 で画像を編集します。HiDream-E1-Full と比較して、優れた画質と編集精度を実現。 -HiDream E1.1 image editing ワークフロープレビュー +HiDream E1.1 画像編集ワークフロープレビュー Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「HiDream E1.1」を検索 @@ -103,10 +104,10 @@ Edit images with HiDream E1.1. Superior image quality and editing accuracy compa ### 1. HiDream E1.1 ワークフローおよび関連素材 -以下の画像をダウンロードし、対応するワークフローおよびモデルが読み込まれた状態で ComfyUI にドラッグ＆ドロップしてください： +以下の画像をダウンロードし、対応するワークフローおよびモデルが読み込まれた状態で ComfyUI にドラッグ＆ドロップしてください： ![HiDream E1.1 ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/hidream/e1.1/hidream_e1_1.png) -以下の画像を入力画像としてダウンロードしてください： +以下の画像を入力画像としてダウンロードしてください： ![HiDream E1.1 ワークフロー入力画像](https://raw.githubusercontent.com/Comfy-Org/example_workflows/refs/heads/main/image/hidream/e1.1/input.webp) ### 2. HiDream-e1 ワークフローの実行手順（ステップ・バイ・ステップ） @@ -114,21 +115,21 @@ Edit images with HiDream E1.1. Superior image quality and editing accuracy compa ![hidream_e1_1_guide](/images/tutorial/image/hidream/hidream-e1-1-guide.jpg) 以下の手順に従ってワークフローを実行してください： -1. `Load Diffusion Model` ノードが `hidream_e1_1_bf16.safetensors` を正しく読み込んでいることを確認してください。 -2. `QuadrupleCLIPLoader` 内の4つのテキストエンコーダーが正しく読み込まれていることを確認してください： -　　- clip_l_hidream.safetensors -　　- clip_g_hidream.safetensors -　　- t5xxl_fp8_e4m3fn_scaled.safetensors -　　- llama_3.1_8b_instruct_fp8_scaled.safetensors -3. `Load VAE` ノードが `ae.safetensors` を使用していることを確認してください。 -4. `Load Image` ノードで、提供された入力画像または任意の画像を読み込んでください。 -5. `Empty Text Encoder(Positive)` ノードに、**画像に対して行いたい変更内容** を入力してください。 -6. `Empty Text Encoder(Negative)` ノードに、**画像に含めたくない要素** を入力してください。 +1. `Load Diffusion Model` ノードが `hidream_e1_1_bf16.safetensors` を正しく読み込んでいることを確認してください。 +2. `QuadrupleCLIPLoader` 内の4つのテキストエンコーダーが正しく読み込まれていることを確認してください： + - clip_l_hidream.safetensors + - clip_g_hidream.safetensors + - t5xxl_fp8_e4m3fn_scaled.safetensors + - llama_3.1_8b_instruct_fp8_scaled.safetensors +3. `Load VAE` ノードが `ae.safetensors` を使用していることを確認してください。 +4. `Load Image` ノードで、提供された入力画像または任意の画像を読み込んでください。 +5. `Empty Text Encoder(Positive)` ノードに、**画像に対して行いたい変更内容** を入力してください。 +6. `Empty Text Encoder(Negative)` ノードに、**画像に含めたくない要素** を入力してください。 7. `Run` ボタンをクリックするか、ショートカットキー `Ctrl（macOS の場合は Cmd） + Enter` を押して画像生成を実行してください。 ### 3. ワークフローに関する補足事項 -- HiDream E1.1 は「合計ピクセル数が 100万」の動的入力をサポートするため、ワークフローでは `Scale Image to Total Pixels` ノードを用いてすべての入力画像を処理・変換します。このため、元の入力画像と比較してアスペクト比が変化する場合があります。 +- HiDream E1.1 は「合計ピクセル数が 100万」の動的入力をサポートするため、ワークフローでは `Scale Image to Total Pixels` ノードを用いてすべての入力画像を処理・変換します。このため、元の入力画像と比較してアスペクト比が変化する場合があります。 - fp16 版モデルを使用する場合、A100 40GB および RTX 4090D 24GB における実際のテストでは、Full バージョンでメモリ不足（Out of memory）が発生しました。そのため、ワークフローはデフォルトで `fp8_e4m3fn_fast` を推論に使用するよう設定されています。 ## HiDream E1 の ComfyUI ネイティブ ワークフローの例 @@ -144,7 +145,7 @@ Edit images with HiDream E1. Professional natural language image editing model. Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「HiDream E1 Full」を検索 @@ -169,7 +170,7 @@ Edit images with HiDream E1. Professional natural language image editing model. HiDream E1 出力例
-E1 は 2025年4月28日にリリースされたモデルで、**768×768 の固定解像度のみ** をサポートします。 +E1 は 2025年4月28日にリリースされたモデルです。 参考までに、Google Colab L4（VRAM 22.5GB）環境で 28 ステップのサンプリングを用いた場合、初回実行に約 500 秒、2回目以降の実行に約 370 秒かかります。 @@ -203,4 +204,4 @@ E1 は 2025年4月28日にリリースされたモデルで、**768×768 の固 - より良い結果を得るためには、プロンプトを複数回修正したり、複数回生成を試行する必要がある場合があります。 - 本モデルは画像スタイルの変更時に一貫性を保つのが難しく、プロンプトをできる限り詳細かつ包括的に記述することを推奨します。 -- モデルは 768×768 の解像度のみをサポートしており、他の解像度で実行した場合、画像品質が著しく低下したり、意図しない出力になることがあります。 +- モデルは 768×768 の解像度をサポートしており、他の寸法で実際にテストしたところ、画像のパフォーマンスが低下したり、他の寸法では大幅に異なる結果が生じる場合があります。 diff --git a/ja/tutorials/image/hidream/hidream-i1.mdx b/ja/tutorials/image/hidream/hidream-i1.mdx index a63e8d9bb..685d81d0c 100644 --- a/ja/tutorials/image/hidream/hidream-i1.mdx +++ b/ja/tutorials/image/hidream/hidream-i1.mdx @@ -2,16 +2,17 @@ title: "ComfyUI ネイティブ版 HiDream-I1 テキストから画像へ変換するワークフローの例" sidebarTitle: "HiDream-I1" description: "本ガイドでは、ComfyUI ネイティブ版 HiDream-I1 のテキストから画像へ変換するワークフローの実行手順を詳しく説明します" -translationSourceHash: 73e39f2a +translationSourceHash: a05c7adb translationFrom: tutorials/image/hidream/hidream-i1.mdx translationBlockHashes: "_intro": 59935bb2 "Model Features": 3ec455b0 "About This Workflow Example": 90c5d2b3 - "HiDream-I1 Workflow": 5956f05c + "HiDream-I1 Workflow": ab1c40db "Other Related Resources": e794d9ef --- + ![HiDream-I1 デモ](https://raw.githubusercontent.com/HiDream-ai/HiDream-I1/main/assets/demo.jpg) HiDream-I1 は、HiDream-ai 社が 2025 年 4 月 7 日に公式にオープンソース化したテキストから画像へ変換するモデルです。このモデルは 17B（170 億）パラメータを有し、[MIT ライセンス](https://github.com/HiDream-ai/HiDream-I1/blob/main/LICENSE) の下で公開されており、個人プロジェクト、学術研究、商用利用のすべてに対応しています。現在、複数のベンチマークテストにおいて優れた性能を発揮しています。 @@ -50,38 +51,38 @@ HiDream-ai 社は、さまざまな用途に対応するため、HiDream-I1 モ ## HiDream-I1 ワークフロー -ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモデル要件は基本的に同一であり、異なるのは [diffusion models](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/tree/main/split_files/diffusion_models) ファイルのみです。 +ComfyUI ネイティブの各 HiDream-I1 ワークフローに必要なモデル要件は基本的に同じで、[ディフュージョンモデル](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/tree/main/split_files/diffusion_models)のファイルのみが異なります。 -どのバージョンを選択すべきか迷う場合は、以下の推奨事項をご参照ください： +どのバージョンを選べばよいかわからない場合は、以下の推奨事項を参考にしてください。 -- **HiDream-I1-Full**：最高品質の画像生成が可能 -- **HiDream-I1-Dev**：高品質な画像生成と速度のバランスを重視 -- **HiDream-I1-Fast**：わずか 16 ステップで画像を生成可能。リアルタイムでの反復試行が必要なシナリオに最適 +- **HiDream-I1-Full** は最高品質の画像を生成できます +- **HiDream-I1-Dev** は高品質な画像生成と速度のバランスが取れています +- **HiDream-I1-Fast** はわずか16ステップで画像を生成でき、リアルタイムでの反復が求められるシナリオに適しています -**dev** および **fast** バージョンでは、ネガティブプロンプトは不要です。そのため、サンプリング時に `cfg` パラメータを `1.0` に設定してください。該当するワークフローには、対応するパラメータ設定が明記されています。 +**dev** および **fast** バージョンではネガティブプロンプトは不要ですので、サンプリング時に `cfg` パラメータを `1.0` に設定してください。該当するワークフロー内で対応するパラメータ設定を記載しています。 -上記 3 バージョンの完全版は、非常に大きな VRAM を必要とします。スムーズな動作には 27GB を超える VRAM が必要になる可能性があります。対応するワークフローチュートリアルでは、**fp8** 版をデモンストレーション用として使用し、大多数のユーザーが問題なく実行できるよう配慮しています。ただし、各チュートリアルでは、VRAM の状況に応じて選択可能な、異なるバージョンのモデルのダウンロードリンクも併せてご提供します。 +3つのバージョンのフルモデルはいずれも大量のVRAMを必要とします。スムーズに動作させるには27GB以上のVRAMが必要になる場合があります。対応するワークフローのチュートリアルでは、**fp8** バージョンをデモ例として使用し、ほとんどのユーザーがスムーズに実行できるようにします。ただし、各例では異なるバージョンのモデルのダウンロードリンクも提供しますので、VRAMの状況に応じて適切なファイルを選択してください。 ### モデルのインストール -以下に示すモデルファイルは、本例で共通して使用するファイルです。 -各リンクをクリックしてダウンロードし、指定された保存場所に配置してください。 -**diffusion models** については、それぞれのワークフローで個別にダウンロード方法をご案内します。 +以下のモデルファイルは共通で使用するファイルです。 +それぞれのリンクをクリックしてダウンロードし、モデルファイルの保存場所に従って保存してください。 +対応するワークフローでは、該当する **diffusion_models** のダウンロードをガイドします。 **text_encoders**： - [clip_l_hidream.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/text_encoders/clip_l_hidream.safetensors) - [clip_g_hidream.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/text_encoders/clip_g_hidream.safetensors) -- [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/text_encoders/t5xxl_fp8_e4m3fn_scaled.safetensors) このモデルは多くのワークフローで既に使用されており、すでにダウンロード済みである可能性があります。 +- [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/text_encoders/t5xxl_fp8_e4m3fn_scaled.safetensors) このモデルは多くのワークフローで使用されており、すでにダウンロード済みかもしれません。 - [llama_3.1_8b_instruct_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/text_encoders/llama_3.1_8b_instruct_fp8_scaled.safetensors) **VAE** -- [ae.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/vae/ae.safetensors) これは Flux の VAE モデルであり、以前に Flux のワークフローを使用したことがある場合、すでにダウンロード済みである可能性があります。 +- [ae.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/vae/ae.safetensors) これはFluxのVAEモデルです。以前Fluxのワークフローを使用したことがある場合、すでにダウンロード済みかもしれません。 -**diffusion models** -対応するワークフローで、該当するモデルファイルのダウンロード方法をご案内します。 +**diffusion_models** +対応するワークフローで、該当するモデルファイルのダウンロードをガイドします。 モデルファイルの保存場所 ``` @@ -95,72 +96,68 @@ ComfyUI ネイティブ版 HiDream-I1 の各種ワークフローにおけるモ │ └── 📂 vae/ │ │ └── ae.safetensors │ └── 📂 diffusion_models/ -│ └── ... # 対応するバージョンのワークフローでインストール方法をご案内します -``` - -### HiDream I1 Full (`hidream_i1_full`) +│ └── ... # 対応するバージョンのワークフローでインストールをガイドします +```### HiDream I1 フルバージョンワークフロー### HiDream I1 Full（`hidream_i1_full`） -Generate images with HiDream I1 Full. Complete version with 50 inference steps for highest quality output. +HiDream I1 Full で画像を生成します。最高品質の出力を得るための50推論ステップの完全版です。 -HiDream I1 Full ワークフロープレビュー +HiDream I1 Full ワークフローのプレビュー - Comfy Cloud でセットアップ不要で実行 + このワークフローを Comfy Cloud で即座に実行できます（セットアップ不要） - Download the workflow JSON file + ワークフローのJSONファイルをダウンロード **出力例** -![HiDream I1 Full 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_full.png) +![HiDream I1 Full の出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_full.png) -#### 1. Model File Download +#### 1. モデルファイルのダウンロード -Please select the appropriate version based on your hardware. Click the link and download the corresponding model file to save it to the `ComfyUI/models/diffusion_models/` folder. +お使いのハードウェアに応じて適切なバージョンを選択してください。リンクをクリックし、対応するモデルファイルをダウンロードして `ComfyUI/models/diffusion_models/` フォルダーに保存します。 -- FP8 version: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true) requires more than 27GB of VRAM +- FP8 バージョン: [hidream_i1_full_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp8.safetensors?download=true)（16GB以上のVRAMが必要） +- フルバージョン: [hidream_i1_full_f16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_full_fp16.safetensors?download=true)（27GB以上のVRAMが必要） -#### 2. Workflow File Download +#### 2. ワークフローファイルのダウンロード -Please download the image below and drag it into ComfyUI to load the corresponding workflow +以下の画像をダウンロードし、ComfyUI にドラッグ＆ドロップして対応するワークフローを読み込んでください。 ![HiDream-I1 Full Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_full.png) -#### 3. Complete the Workflow Step by Step +#### 3. ワークフローをステップごとに完了させる ![HiDream-I1 Full Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_full_flow_diagram.jpg) -Complete the workflow execution step by step -1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_full_fp8.safetensors` file -2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly +ワークフローの実行をステップごとに完了します。 +1. `Load Diffusion Model` ノードが `hidream_i1_full_fp8.safetensors` ファイルを使用していることを確認します。 +2. `QuadrupleCLIPLoader` で4つの対応するテキストエンコーダーが正しく読み込まれていることを確認します。 - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. Make sure the `Load VAE` node is using the `ae.safetensors` file -4. For the **full** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` -5. For the `Ksampler` node, you need to make the following settings - - Set `steps` to `50` - - Set `cfg` to `5.0` - - (Optional) Set `sampler` to `lcm` - - (Optional) Set `scheduler` to `normal` -6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation +3. `Load VAE` ノードが `ae.safetensors` ファイルを使用していることを確認します。 +4. **Full** バージョンの場合、`ModelSamplingSD3` の `shift` パラメーターを `3.0` に設定する必要があります。 +5. `Ksampler` ノードでは、次の設定を行う必要があります。 + - `steps` を `50` に設定 + - `cfg` を `5.0` に設定 + - （オプション）`sampler` を `lcm` に設定 + - （オプション）`scheduler` を `normal` に設定 +6. `Run` ボタンをクリックするか、ショートカットキー `Ctrl(cmd) + Enter` を使用して画像生成を実行します。### HiDream-I1 Devバージョンのワークフロー### HiDream I1 Dev（`hidream_i1_dev`） -### HiDream I1 Dev (`hidream_i1_dev`) - -Generate images with HiDream I1 Dev. Balanced version with 28 inference steps, suitable for medium-range hardware. +HiDream I1 Devで画像を生成します。バランスの取れたバージョンで、推論ステップ数は28、中程度のハードウェアに適しています。 HiDream I1 Dev ワークフロープレビュー - - Comfy Cloud でセットアップ不要で実行 + + ゼロセットアップでComfy Cloud上でこのワークフローを実行 - Download the workflow JSON file + ワークフローのJSONファイルをダウンロード @@ -168,48 +165,46 @@ Generate images with HiDream I1 Dev. Balanced version with 28 inference steps, s ![HiDream I1 Dev 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_dev.png) -#### 1. Model File Download -Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. +#### 1. モデルファイルのダウンロード +お使いのハードウェアに応じて適切なバージョンを選択し、リンクをクリックして対応するモデルファイルをダウンロードし、`ComfyUI/models/diffusion_models/` フォルダーに保存してください。 -- FP8 version: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) requires more than 27GB of VRAM +- FP8版: [hidream_i1_dev_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_fp8.safetensors?download=true) には16GB以上のVRAMが必要です +- フル版: [hidream_i1_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors?download=true) には27GB以上のVRAMが必要です -#### 2. Workflow File Download -Please download the image below and drag it into ComfyUI to load the corresponding workflow +#### 2. ワークフローファイルのダウンロード +以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして対応するワークフローを読み込んでください。 -![HiDream-I1 Dev Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) +![HiDream-I1 Devバージョンワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_dev.png) -#### 3. Complete the Workflow Step by Step +#### 3. ワークフローをステップごとに完了する -![HiDream-I1 Dev Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) -Complete the workflow execution step by step -1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_dev_fp8.safetensors` file -2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly +![HiDream-I1 Devバージョンフロー図](/images/tutorial/advanced/hidream/hidream_i1_dev_flow_diagram.jpg) +ワークフローの実行をステップごとに完了します。 +1. `Load Diffusion Model`ノードが`hidream_i1_dev_fp8.safetensors`ファイルを使用していることを確認してください。 +2. `QuadrupleCLIPLoader`内の対応する4つのテキストエンコーダーが正しく読み込まれていることを確認してください。 - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. Make sure the `Load VAE` node is using the `ae.safetensors` file -4. For the **dev** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `6.0` -5. For the `Ksampler` node, you need to make the following settings - - Set `steps` to `28` - - (Important) Set `cfg` to `1.0` - - (Optional) Set `sampler` to `lcm` - - (Optional) Set `scheduler` to `normal` -6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation - -### HiDream I1 Fast (`hidream_i1_fast`) +3. `Load VAE`ノードが`ae.safetensors`ファイルを使用していることを確認してください。 +4. **dev**バージョンでは、`ModelSamplingSD3`の`shift`パラメーターを`6.0`に設定する必要があります。 +5. `Kサンプラー`ノードでは、以下の設定を行う必要があります。 + - `steps`を`28`に設定 + - （重要）`cfg`を`1.0`に設定 + - （オプション）`sampler`を`lcm`に設定 + - （オプション）`scheduler`を`normal`に設定 +6. `Run`ボタンをクリックするか、ショートカット`Ctrl(cmd) + Enter`を使用して画像生成を実行します。### HiDream-I1 Fast バージョン ワークフロー### HiDream I1 Fast（`hidream_i1_fast`） -Generate images quickly with HiDream I1 Fast. Lightweight version with 16 inference steps, ideal for rapid previews on lower-end hardware. +HiDream I1 Fast を使って素早く画像を生成します。軽量版で16段階の推論ステップ数を備え、低スペックのハードウェアでも手軽にプレビューできます。 HiDream I1 Fast ワークフロープレビュー - Comfy Cloud でセットアップ不要で実行 + セットアップ不要でこのワークフローを Comfy Cloud 上で実行 - Download the workflow JSON file + ワークフローのJSONファイルをダウンロード @@ -217,36 +212,36 @@ Generate images quickly with HiDream I1 Fast. Lightweight version with 16 infere ![HiDream I1 Fast 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/hidream_i1_fast.png) -#### 1. Model File Download -Please select the appropriate version based on your hardware, click the link and download the corresponding model file to save to the `ComfyUI/models/diffusion_models/` folder. +#### 1. モデルファイルのダウンロード +お使いのハードウェアに応じて適切なバージョンを選択し、リンクをクリックして対応するモデルファイルをダウンロードし、`ComfyUI/models/diffusion_models/` フォルダーに保存してください。 -- FP8 version: [hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true) requires more than 16GB of VRAM -- Full version: [hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true) requires more than 27GB of VRAM +- FP8版：[hidream_i1_fast_fp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_fp8.safetensors?download=true)（16GB以上のVRAMが必要） +- フル版：[hidream_i1_fast_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/blob/main/split_files/diffusion_models/hidream_i1_fast_bf16.safetensors?download=true)（27GB以上のVRAMが必要） -#### 2. Workflow File Download -Please download the image below and drag it into ComfyUI to load the corresponding workflow +#### 2. ワークフローファイルのダウンロード +以下の画像をダウンロードし、ComfyUIにドラッグ＆ドロップして対応するワークフローを読み込んでください。 -![HiDream-I1 Fast Version Workflow](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) +![HiDream-I1 Fast版ワークフロー](https://raw.githubusercontent.com/Comfy-Org/example_workflows/main/hidream_i1/hidream_i1_fast.png) -#### 3. Complete the Workflow Step by Step +#### 3. ワークフローをステップごとに完了させる -![HiDream-I1 Fast Version Flow Diagram](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) +![HiDream-I1 Fast版フロー図](/images/tutorial/advanced/hidream/hidream_i1_fast_flow_diagram.jpg) -Complete the workflow execution step by step -1. Make sure the `Load Diffusion Model` node is using the `hidream_i1_fast_fp8.safetensors` file -2. Make sure the four corresponding text encoders in `QuadrupleCLIPLoader` are loaded correctly +ワークフローの実行をステップごとに完了させます。 +1. `Load Diffusion Model`（拡散モデルを読み込む）ノードが `hidream_i1_fast_fp8.safetensors` ファイルを使用していることを確認します。 +2. `QuadrupleCLIPLoader` 内の4つの対応するテキストエンコーダーが正しく読み込まれていることを確認します。 - clip_l_hidream.safetensors - clip_g_hidream.safetensors - t5xxl_fp8_e4m3fn_scaled.safetensors - llama_3.1_8b_instruct_fp8_scaled.safetensors -3. Make sure the `Load VAE` node is using the `ae.safetensors` file -4. For the **fast** version, you need to set the `shift` parameter in `ModelSamplingSD3` to `3.0` -5. For the `Ksampler` node, you need to make the following settings - - Set `steps` to `16` - - (Important) Set `cfg` to `1.0` - - (Optional) Set `sampler` to `lcm` - - (Optional) Set `scheduler` to `normal` -6. Click the `Run` button, or use the shortcut `Ctrl(cmd) + Enter` to execute the image generation +3. `Load VAE`（VAEを読み込む）ノードが `ae.safetensors` ファイルを使用していることを確認します。 +4. **高速版**では、`ModelSamplingSD3`（モデルサンプリングSD3）の `shift` パラメーターを `3.0` に設定する必要があります。 +5. `Ksampler`（Kサンプラー）ノードでは、次の設定を行う必要があります。 + - `steps` を `16` に設定 + - （重要）`cfg` を `1.0` に設定 + - （オプション）`sampler`（サンプラー）を `lcm` に設定 + - （オプション）`scheduler`（スケジューラー）を `ノーマル` に設定 +6. `Run` ボタンをクリックするか、ショートカット `Ctrl（Cmd）+ Enter` を使用して画像生成を実行します。 ## その他の関連リソース diff --git a/ja/tutorials/image/hidream/hidream-o1.mdx b/ja/tutorials/image/hidream/hidream-o1.mdx index 96a5398b7..c7c0786af 100644 --- a/ja/tutorials/image/hidream/hidream-o1.mdx +++ b/ja/tutorials/image/hidream/hidream-o1.mdx @@ -2,15 +2,16 @@ title: "ComfyUI ネイティブ版 HiDream-O1-Image ワークフローの例" sidebarTitle: "HiDream-O1-Image" description: "本ガイドでは、ComfyUI ネイティブ版 HiDream-O1-Image のテキストから画像へ、および画像編集のワークフローを説明します" -translationSourceHash: 368a086e +translationSourceHash: c6c33ec2 translationFrom: tutorials/image/hidream/hidream-o1.mdx translationBlockHashes: "_intro": e1c85b19 - "HiDream-O1-Image Full Workflow": 9c60b82c - "HiDream-O1-Image Dev Workflow": 24233bb7 + "HiDream-O1-Image Full Workflow": 1a1dbcc5 + "HiDream-O1-Image Dev Workflow": 3fed212d "Additional Notes": 762758eb --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' @@ -32,9 +33,9 @@ HiDream-O1-Image は [MIT ライセンス](https://github.com/HiDream-ai/HiDream ## HiDream-O1-Image Full ワークフロー -### HiDream O1 Full: Image generation (`image_hidream_o1`) +### HiDream O1 Full: 画像生成 (`image_hidream_o1`) -Input a text prompt and optionally upload reference images. Generate a high-resolution image up to 2048x2048 with text-to-image, editing, or subject-driven personalization. +テキストプロンプトを入力し、オプションで参照画像をアップロードします。テキストから画像、編集、または被写体駆動のパーソナライゼーションによって、最大2048x2048の高解像度画像を生成します。 HiDream O1 Full ワークフロープレビュー @@ -43,7 +44,7 @@ Input a text prompt and optionally upload reference images. Generate a high-reso Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「HiDream O1 Full」を検索 @@ -51,20 +52,53 @@ Input a text prompt and optionally upload reference images. Generate a high-reso ![HiDream O1 Full 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_hidream_o1.png) +### 1. ワークフローをダウンロード + +### 2. モデルをダウンロード + +**チェックポイント** — 再パッケージ化および量子化済み。すべてbf16を使用して最悪の外れ値を処理し、未使用のdeepstackレイヤーは削除済み： + +- [hidream_o1_image_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_fp8_scaled.safetensors) — FP8。サポートハードウェアで高速化するため、安全なMLPレイヤーでfp8/mxfp8 matmulを使用 +- [hidream_o1_image_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_mxfp8.safetensors) — MXFP8量子化バリアント +- [hidream_o1_image_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_bf16.safetensors) — フルbf16精度（最大） + +**テキストエンコーダ**（プロンプト強化）— 全バージョン共通： + +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) + +**LoRA（オプション）** — Dev蒸留はFullモデルにLoRAとして適用可能で、蒸留強度を調整できます（[Kijai](https://huggingface.co/Kijai/hidream-O1-image_comfy) 提供）： + +- [hidream_o1_dev_lora_rank_64_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16.safetensors) — フルランク +- [hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_dev_lora_rank_64_bf16_pruned_v1.safetensors) — プルーニングバリアント +- [hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors](https://huggingface.co/Kijai/hidream-O1-image_comfy/blob/main/loras/hidream_o1_image_dev_2604_lora_avg_rankg_224_bf16.safetensors) — 代替チェックポイントベースの蒸留 + +``` +📂 ComfyUI/ +├── 📂 models/ +│ ├── 📂 checkpoints/ +│ │ ├── hidream_o1_image_fp8_scaled.safetensors +│ │ ├── hidream_o1_image_mxfp8.safetensors +│ │ └── hidream_o1_image_bf16.safetensors +│ ├── 📂 loras/ +│ │ └── hidream_o1_dev_lora_rank_64_bf16.safetensors +│ └── 📂 text_encoders/ +│ └── gemma4_e4b_it_fp8_scaled.safetensors +``` + ### 3. ワークフローの使い方 - `CheckpointLoaderSimple` ノードが選択したチェックポイントを読み込んでいることを確認 - `CLIPTextEncode` ノードにプロンプトを入力 -- **テキスト→画像モード：** **"Switch to Image Edit"** トグルを **オフ**（デフォルト）にします。サンプラーはテキストプロンプトを直接使用します。 +- **テキストから画像モード：** **"Switch to Image Edit"** トグルを **オフ**（デフォルト）にします。サンプラーはテキストプロンプトを直接使用します。 - **画像編集モード：** **"Switch to Image Edit"** を **オン** にし、`Load Image` ノードで参照画像をアップロードして `HiDreamO1ReferenceImages` に接続します。 > **注意：** O1 サンプラーは潜在サンプルを出力するため、`CheckpointLoaderSimple` が読み込んだ VAE を使用して `VAEDecode` ノードでデコードする必要があります。 ## HiDream-O1-Image Dev ワークフロー -### HiDream O1 Dev (`image_hidream_o1_dev`) +### HiDream O1 Dev（`image_hidream_o1_dev`） -Input a text prompt and optional reference images. Generate a high-resolution image (up to 2048x2048) with support for text-to-image, image editing, and subject-driven personalization. +テキストプロンプトとオプションの参照画像を入力します。テキストから画像への生成、画像編集、被写体駆動のパーソナライゼーションをサポートし、最大2048×2048の高解像度画像を生成します。 HiDream O1 Dev ワークフロープレビュー @@ -73,17 +107,17 @@ Input a text prompt and optional reference images. Generate a high-resolution im Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSONをダウンロード、またはテンプレートライブラリで「HiDream O1 Dev」を検索 **入力素材** -次の `\1` ノードにこのファイルをアップロード： +該当する `LoadImage` ノードにこのファイルをアップロードしてください： - `LoadImage` node 213 · `noir_portrait.png` + `LoadImage` ノード 213 · `noir_portrait.png` @@ -98,12 +132,37 @@ Input a text prompt and optional reference images. Generate a high-resolution im HiDream O1 Dev 出力例
+### 1. ワークフローのダウンロード + +### 2. モデルのダウンロード + +**チェックポイント（Dev）** — 再パッケージ化され量子化されています。すべてbf16を使用し、最悪の外れ値に対応し、未使用のDeepStackレイヤーは削除されています： + +- [hidream_o1_image_dev_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_fp8_scaled.safetensors) — FP8、サポートされているハードウェア上で高速化するために、安全なMLPレイヤーでfp8/mxfp8行列積を使用 +- [hidream_o1_image_dev_mxfp8.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_mxfp8.safetensors) — MXFP8量子化バリアント +- [hidream_o1_image_dev_bf16.safetensors](https://huggingface.co/Comfy-Org/HiDream-O1-Image/blob/main/checkpoints/hidream_o1_image_dev_bf16.safetensors) — 完全なbf16精度（最大） + +**テキストエンコーダー**（プロンプト拡張）— すべてのバージョンで共有： + +- [gemma4_e4b_it_fp8_scaled.safetensors](https://huggingface.co/Comfy-Org/gemma-4/blob/main/text_encoders/gemma4_e4b_it_fp8_scaled.safetensors) + +``` +📂 ComfyUI/ +├── 📂 models/ +│ ├── 📂 checkpoints/ +│ │ ├── hidream_o1_image_dev_fp8_scaled.safetensors +│ │ ├── hidream_o1_image_dev_mxfp8.safetensors +│ │ └── hidream_o1_image_dev_bf16.safetensors +│ └── 📂 text_encoders/ +│ └── gemma4_e4b_it_fp8_scaled.safetensors +``` + ### 3. ワークフローの使い方 - `CheckpointLoaderSimple` が `hidream_o1_image_dev_fp8_scaled.safetensors` を読み込んでいることを確認 - Dev バージョンは 28 ステップ、CFG=1.0 で動作 — ネガティブプロンプトは不要 -- **テキスト→画像モード：** **"Switch to Image Edit"** トグルを **オフ**（デフォルト） -- **画像編集モード：** **"Switch to Image Edit"** を **オン** にし、`Load Image` で参照画像をアップロードして `HiDreamO1ReferenceImages` に接続 +- **テキストから画像へのモード：** **"Switch to Image Edit"** トグルを **オフ**（デフォルト） +- **画像編集モード：** **"Switch to Image Edit"** を **オン** にし、`LoadImage` で参照画像をアップロードして `HiDreamO1ReferenceImages` に接続 ## 補足説明 diff --git a/ja/tutorials/image/ideogram/ideogram-v4.mdx b/ja/tutorials/image/ideogram/ideogram-v4.mdx index 138bdfe33..b1e224539 100644 --- a/ja/tutorials/image/ideogram/ideogram-v4.mdx +++ b/ja/tutorials/image/ideogram/ideogram-v4.mdx @@ -2,11 +2,11 @@ title: "ComfyUI Ideogram 4.0 オープンソースモデルチュートリアル" description: "ComfyUI で Ideogram 4.0 オープンソースモデルを使用する方法" sidebarTitle: "Ideogram 4.0" -translationSourceHash: de26d2f5 +translationSourceHash: df7b7a64 translationFrom: tutorials/image/ideogram/ideogram-v4.mdx translationBlockHashes: "_intro": e87a6658 - "Ideogram 4.0 Text-to-Image Workflow": 191bc145 + "Ideogram 4.0 Text-to-Image Workflow": e9e08f22 "Live Conversation with Ideogram & ComfyOrg": 9f3809e0 --- @@ -16,18 +16,20 @@ Ideogram 4.0 は、Ideogram がオープンソースモデルとして公開し -### Ideogram v4: Text to Image (`image_ideogram4_t2i`) +## Ideogram 4.0 テキストから画像へのワークフロー -Input a text prompt or structured JSON description. Generate an image with precise layout, color, and style control using Ideogram 4.0. +### Ideogram v4：テキストから画像へ（`image_ideogram4_t2i`） -Ideogram 4.0 text-to-image ワークフロープレビュー +テキストプロンプトまたは構造化JSONの説明を入力します。Ideogram 4.0を使用して、レイアウト、色、スタイルを正確に制御しながら画像を生成します。 + +Ideogram 4.0 テキストから画像へのワークフロープレビュー Comfy Cloud で開く - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロードするか、テンプレートライブラリで「Ideogram v4：テキストから画像へ」を検索 @@ -35,37 +37,37 @@ Input a text prompt or structured JSON description. Generate an image with preci ![Ideogram 4.0 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_ideogram4_t2i.png) -### プロンプト形式 +### プロンプトの形式 -オープンソースワークフローは2つのプロンプトモードをサポート： +このオープンソースワークフローは、2つのプロンプトモードをサポートしています。 -1. **自然言語** — 素早く簡単、シンプルなアイデアに最適 -2. **構造化 JSON** — レイアウト、色、スタイルを精密に制御 +1. **自然言語** — 迅速かつ簡単で、シンプルなアイデアに最適 +2. **構造化JSON** — レイアウト、色、スタイルを正確に制御するため -ワークフローにはプロンプト構築テンプレートが含まれており、任意の LLM と組み合わせて JSON プロンプトを生成できます。 +このワークフローには、任意のLLMを使用して一致するJSONプロンプトを生成できるプロンプト構築テンプレートが含まれています。 -ワークフロー内のノートより： -> このモデルは構造化 JSON キャプションでトレーニングされています（シーン要約、スタイルブロック、背景、およびオプションでバウンディングボックスと16進数カラーパレットを含むオブジェクトごとの説明）。公式推論はこのスキーマに対してプロンプトを検証します。 +ワークフロー内の注釈には次のように説明されています： +> モデルは構造化JSONキャプション（シーンの概要、スタイルブロック、背景、および境界ボックスと16進数のカラーパレットを含むオプションのオブジェクトごとの説明）でトレーニングされています。公式の推論では、そのスキーマに対してプロンプトを検証します。 ### Ideogram 4.0 モデルのダウンロード -Hugging Face の [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4) ですべての再パッケージ化されたモデルファイルを見つけることができます。 +再パッケージ化されたすべてのモデルファイルは、Hugging Faceの[Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogram-4)にあります。 - Ideogram 4.0 拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 + Ideogram 4.0 の拡散モデル（約13.8 GB）。models/diffusion_models/ に配置します。 - Ideogram 4.0 条件なし拡散モデル（~13.8 GB）。models/diffusion_models/ に配置 + Ideogram 4.0 の無条件拡散モデル（約13.8 GB）。models/diffusion_models/ に配置します。 - Ideogram 4.0 テキストエンコーダー（~8 GB）。models/text_encoders/ に配置 + Ideogram 4.0 のテキストエンコーダ（約8 GB）。models/text_encoders/ に配置します。 - Ideogram 4.0 テキストエンコーダー（~2 GB）。models/text_encoders/ に配置 + Ideogram 4.0 のテキストエンコーダ（約2 GB）。models/text_encoders/ に配置します。 - Ideogram 4.0 VAE（~335 MB）。models/vae/ に配置 + Ideogram 4.0 のVAE（約335 MB）。models/vae/ に配置します。 @@ -85,34 +87,34 @@ Hugging Face の [Comfy-Org/Ideogram-4](https://huggingface.co/Comfy-Org/Ideogra ``` - - このワークフローは Subgraph ノードを使用しています。Subgraph ドキュメントでカスタマイズ方法を確認してください。 + + このワークフローは、モジュール処理にサブグラフノードを使用しています。サブグラフのドキュメントを参照して、カスタマイズと拡張方法を学んでください。 ### ワークフローの手順 -1. モデルをダウンロードし、正しいディレクトリに配置（上表参照） -2. ワークフローファイルをダウンロードして ComfyUI にドラッグ -3. Ideogram4 サブグラフノードにプロンプトを入力（自然言語または構造化 JSON） -4. （オプション）`ResolutionSelector` ノードで解像度を調整 -5. `Run` ボタンをクリックするか、ショートカット `Ctrl(cmd) + Enter` で画像を生成 -6. `Save Image` ノードで結果を確認 — 出力ファイルは `ComfyUI/output/` に保存 +1. モデルをダウンロードし、正しいディレクトリに配置します（上記の表を参照） +2. ワークフローファイルをダウンロードし、ComfyUI にドラッグします +3. Ideogram4 サブグラフノードにプロンプト（自然言語または構造化JSON）を入力します +4. （オプション）`ResolutionSelector` ノードを使用して解像度を調整します +5. `Run` をクリックするか、ショートカット `Ctrl(cmd) + Enter` を使用して画像を生成します +6. `Save Image` ノードで結果を表示します。出力ファイルは `ComfyUI/output/` に保存されます -### 安全フィルターに関する注意 +### セーフティフィルタに関する注意 -「Image blocked by safety filter」と表示された場合、それは Ideogram 4.0 の組み込み安全フィルターです。JSON 以外（プレーンテキスト）のプロンプトは誤検出率が高くなります。構造化 JSON プロンプトを使用すると、プロンプトがブロックされる可能性が低減します。 +「イメージがセーフティフィルタによってブロックされました」というメッセージが表示された場合、これはIdeogram 4.0の組み込みセーフティフィルタです。JSON以外（プレーンテキスト）のプロンプトは偽陽性率が高くなります。構造化JSONプロンプトを使用すると、プロンプトがブロックされる可能性が低くなります。 -詳細については、[Ideogram 4 の公式プロンプトガイド](https://github.com/ideogram-oss/ideogram4/blob/main/docs/prompting.md#safety-filter)をご覧ください。 +詳細については、[Ideogram 4の公式プロンプトガイド](https://github.com/ideogram-oss/ideogram4/blob/main/docs/prompting.md#safety-filter)を参照してください。 -## Ideogram & ComfyOrg 創業者対談 +## Ideogram & ComfyOrg によるライブ対談 -Mohammad Norouzi（Ideogram CEO）と Yoland Yan（ComfyOrg CEO）をゲストに迎えた特別なライブ対談。司会は Purz と Rob。 +Mohammad Norouzi（Ideogram CEO）と Yoland Yan（ComfyOrg CEO）による特別ライブ対談。ホストは Purz & Rob です。 diff --git a/ja/tutorials/image/krea/krea-2.mdx b/ja/tutorials/image/krea/krea-2.mdx index 084c7da79..35d134a65 100644 --- a/ja/tutorials/image/krea/krea-2.mdx +++ b/ja/tutorials/image/krea/krea-2.mdx @@ -14,7 +14,7 @@ translationBlockHashes: import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' - + **Krea 2** は、[Krea AI](https://www.krea.ai) がゼロから学習した画像生成モデルで、クリエイティブでスタイルの探求に重点を置いています。 @@ -22,7 +22,7 @@ Krea 2は、連携して動作するように設計された2つのモデルと - **Krea 2 RAW**：フルステップサンプリング（52ステップ）のベースモデル。蒸留なしで多様性が高く柔軟性に富み、ファインチューニングやLoRAトレーニングに最適です。 - **Krea 2 Turbo**：8ステップの蒸留チェックポイントで、高速かつ高品質な生成を実現。RAWで学習したLoRAをシームレスにTurboに適用できます。 -**ライセンス**: [Krea AI Community License](https://www.krea.ai/krea-2-licensing) +**ライセンス**: [Krea AI コミュニティライセンス](https://www.krea.ai/krea-2-licensing) ## ローンチイベントを見る @@ -225,3 +225,190 @@ The workflow is organized into a few parts: │ └── 📂 loras/ │ └── krea2_style_reference.safetensors ``` + +## Krea-2 Turbo text-to-image workflow + +### Krea-2: Text to Image (`image_krea2_turbo_t2i`) + +Generate images from text prompts using Krea 2, a foundation model built for aesthetic quality and creative control. It focuses on rendering expressive, stylistically diverse images. + +Krea-2 Turbo text-to-image workflow preview + + + + Comfy Cloud で開く + + + JSON をダウンロード、またはテンプレートライブラリで "Krea-2" を検索 + + + + + +The workflow is organized into a few parts: + +1. **Text to Image (Krea-2 Turbo) subgraph**: the core generation pipeline, containing model loading, prompt handling, sampling, and VAE decode +2. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K — set the megapixels value to 2.0 to get 2K resolution. +3. **CustomCombo (LoRA selector)**: a pre-built trigger word selector for the available style LoRAs. If you download additional LoRAs, you can customize this selector and pair them with the corresponding LoRA files accordingly. +4. **SaveImage**: saves the generated image + + + This workflow uses Subgraph nodes for modular processing. Check out the Subgraph documentation to learn how to customize and extend the workflow. + + +### Single-click generation + +At its simplest, just enter a text prompt in the subgraph, select a resolution, and click **Queue Prompt**. The defaults (8 steps, prompt enhancement enabled, no LoRA) produce a high-quality image with minimal configuration. + +### Workflow controls + +The **Text to Image (Krea-2 Turbo)** subgraph exposes the following controls: + +| Control | Description | +|---------|-------------| +| **Text String (User Prompt)** | The text prompt describing the image you want to generate | +| **prompt_enhance** | Toggle LLM-powered prompt expansion on/off | +| **LLM_max_token** | Maximum token length for prompt enhancement | +| **Width / Height** | Output resolution (controlled by ResolutionSelector) | +| **Seed** | Random seed for reproducibility | +| **enable_lora?** | Enable or disable style LoRA | +| **LoRA Strength** | Strength of the applied style LoRA | +| **LoRA Trigger Word** | Auto-populated trigger word for the selected LoRA | + +### Style LoRAs + +Krea also released a collection of style LoRAs for Krea 2. Select one in the **CustomCombo** node and enable `enable_lora?` in the subgraph to apply the style: + +| LoRA | Trigger Word | Recommended Strength | Download | +|------|-------------|:-------------------:|----------| +| krea2_darkbrush | monochrome ink wash style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_darkbrush.safetensors) | +| krea2_dotmatrix | monochrome stippling style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_dotmatrix.safetensors) | +| krea2_kidsdrawing | naive expressive sketch style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_kidsdrawing.safetensors) | +| krea2_neondrip | textured abstract style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_neondrip.safetensors) | +| krea2_rainywindow | rainy window style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_rainywindow.safetensors) | +| krea2_retroanime | purple retro anime style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_retroanime.safetensors) | +| krea2_softwatercolor | art deco watercolor style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_softwatercolor.safetensors) | +| krea2_sunsetblur | ethereal motion blur style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_sunsetblur.safetensors) | +| krea2_vintagetarot | vintage tarot style | 1.0 | [Download](https://huggingface.co/Comfy-Org/Krea-2/blob/main/loras/krea2_vintagetarot.safetensors) | + +Place the `.safetensors` files in `ComfyUI/models/loras/`. + +### Model downloads + +For local use, download the ComfyUI-optimized model files from [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2). + + + + krea2_turbo_fp8_scaled.safetensors: Turbo FP8 (recommended for most users) + + + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B text encoder + + + qwen_image_vae.safetensors + + + View all style LoRAs on Hugging Face + + + +Other model variants (BF16, NVFP4, MXFP8) are also available for users with higher-end hardware. + +#### Model storage location + +``` +📂 ComfyUI/ +├── 📂 models/ +│ ├── 📂 diffusion_models/ +│ │ └── krea2_turbo_fp8_scaled.safetensors +│ ├── 📂 text_encoders/ +│ │ └── qwen3vl_4b_fp8_scaled.safetensors +│ ├── 📂 vae/ +│ │ └── qwen_image_vae.safetensors +│ └── 📂 loras/ +│ └── krea2_softwatercolor.safetensors (and other style LoRAs) +``` + +## Krea-2 Turbo style reference workflow + +### Krea-2 Int8: Image Style Reference (`image_krea2_turbo_int8_image_style_reference`) + +Generate images with the Krea-2 Turbo model while referencing the style of 1–2 uploaded images, using the high-performance Int8 Convrot format for fast inference. + +Krea-2 Turbo style reference workflow preview + + + + Comfy Cloud で開く + + + JSON をダウンロード、またはテンプレートライブラリで "Krea-2 Style Reference" を検索 + + + +**Input materials** + +Upload this file to the matching `LoadImage` node: + + + + `LoadImage` node 69 · `krea2_reference_image.png` + + + +
+ krea2_reference_image.png +
+ +**Example output** + +![Krea-2 style reference example output](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_krea2_turbo_int8_image_style_reference.png) + +The style reference workflow builds on the Krea-2 Turbo pipeline by adding reference image conditioning. Upload one or more reference images to influence the aesthetic style, mood, and visual direction of the generated output. + +The workflow is organized into a few parts: + +1. **Image Style Reference (Krea-2 Turbo) subgraph**: the core generation pipeline with style reference support, containing model loading, reference image conditioning, prompt handling, and sampling +2. **LoadImage**: upload your style reference images +3. **ResolutionSelector**: choose your desired output resolution. Krea 2 supports outputs from 1K to 2K. +4. **SaveImage**: saves the generated image + +### Using style reference + +1. **Upload reference images** — use the **LoadImage** node to load one or more style reference images +2. **Enter your prompt** — type a text description in the subgraph's user prompt input +3. **Select a model variant** — choose the model checkpoint in the subgraph's model selector +4. **Adjust settings** — configure resolution, seed, and optionally enable prompt enhancement +5. **Click Queue** — press `Ctrl+Enter` to generate + +This style reference workflow uses a dedicated diffusion model and LoRA. The text encoder and VAE are shared with the text-to-image workflow. + + + + krea2_turbo_int8_convrot.safetensors + + + qwen3vl_4b_fp8_scaled.safetensors: Qwen3VL-4B text encoder + + + qwen_image_vae.safetensors + + + krea2_style_reference.safetensors + + + +Place the files in the following directories: + +``` +📂 ComfyUI/ +├── 📂 models/ +│ ├── 📂 diffusion_models/ +│ │ └── krea2_turbo_int8_convrot.safetensors +│ ├── 📂 text_encoders/ +│ │ └── qwen3vl_4b_fp8_scaled.safetensors +│ ├── 📂 vae/ +│ │ └── qwen_image_vae.safetensors +│ └── 📂 loras/ +│ └── krea2_style_reference.safetensors +``` diff --git a/ja/tutorials/image/lens/lens.mdx b/ja/tutorials/image/lens/lens.mdx index e63737a63..2e4a90314 100644 --- a/ja/tutorials/image/lens/lens.mdx +++ b/ja/tutorials/image/lens/lens.mdx @@ -2,15 +2,16 @@ title: "Lens ComfyUI ワークフロー例" description: "Lens は Microsoft による 3.8B パラメータのテキスト画像生成モデルです。デュアルストリーム MMDiT、GPT-OSS-20B テキスト特徴量、FLUX.2 VAE を採用し、効率的な高解像度生成を実現します。" sidebarTitle: "Lens" -translationSourceHash: cb4c08ff +translationSourceHash: de8cb82f translationFrom: tutorials/image/lens/lens.mdx translationBlockHashes: "_intro": b0a3e175 - "Lens text-to-image workflow": a45ab498 + "Lens text-to-image workflow": e082a32d "Lens model downloads": 5055ce84 "Available models": 5876b860 --- + import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' **Lens** は **Microsoft** によるオープンなテキスト画像生成モデルで、MIT ライセンスで提供されています。**38億** パラメータを持ち、**デュアルストリーム MMDiT** アーキテクチャに **GPT-OSS-20B** テキストエンコーダーの特徴量と **FLUX.2 セマンティック VAE** を組み合わせ、より大規模なモデルよりも少ない学習計算量で競争力のある画質を実現します。 @@ -36,15 +37,15 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' このワークフローは Subgraph ノードを使用したモジュール処理を採用しています。Subgraph のドキュメントを参照して、ワークフローのカスタマイズと拡張方法を学んでください。 -### Lens: Text to Image (`image_lens_t2i`) +### Lens: テキストから画像へ (`image_lens_t2i`) -Input a text prompt and select resolution and aspect ratio. Generate a high-quality image using the efficient Lens text-to-image model. +テキストプロンプトを入力し、解像度とアスペクト比を選択します。効率的な Lens テキスト画像生成モデルを使用して、高品質な画像を生成します。 -Lens text-to-image ワークフロープレビュー +Lens テキスト画像生成ワークフロープレビュー - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Lens」を検索 {/* TODO: Enable Cloud template when Lens is available on Comfy Cloud */} {/**/} @@ -54,27 +55,27 @@ Input a text prompt and select resolution and aspect ratio. Generate a high-qual **出力例** -![Lens text-to-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_t2i.png) +![Lens テキスト画像生成出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_t2i.png) -#### Get started +#### はじめる -1. Update ComfyUI to the latest version +1. ComfyUI を最新バージョンにアップデート {/* TODO: Add Cloud option when template is available */} -2. Go to **Template** and search for **Lens** -3. Select the **Lens** workflow -4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** +2. **テンプレート**に移動し、「Lens」を検索 +3. **Lens**ワークフローを選択 +4. 不足しているモデルをダウンロード（[モデルダウンロード](#lens-model-downloads)を参照）、プロンプトを入力し、**キュー**をクリック -### Lens Turbo: Text to Image (`image_lens_turbo_t2i`) +### Lens Turbo: テキストから画像へ (`image_lens_turbo_t2i`) -Input a text prompt and select resolution, aspect ratio, and inference steps. Generate a high-quality image using the Lens text-to-image model. +テキストプロンプトを入力し、解像度、アスペクト比、推論ステップ数を選択します。Lens テキスト画像生成モデルを使用して、高品質な画像を生成します。 -Lens Turbo text-to-image ワークフロープレビュー +Lens Turbo テキスト画像生成ワークフロープレビュー - JSON をダウンロード、またはテンプレートライブラリで「\1」を検索 + JSON をダウンロード、またはテンプレートライブラリで「Lens Turbo」を検索 {/* TODO: Enable Cloud template when Lens Turbo is available on Comfy Cloud */} {/**/} @@ -84,15 +85,15 @@ Input a text prompt and select resolution, aspect ratio, and inference steps. Ge **出力例** -![Lens Turbo text-to-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_turbo_t2i.png) +![Lens Turbo テキスト画像生成出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_lens_turbo_t2i.png) -#### Get started +#### はじめる -1. Update ComfyUI to the latest version +1. ComfyUI を最新バージョンにアップデート {/* TODO: Add Cloud option when template is available */} -2. Go to **Template** and search for **Lens Turbo** -3. Select the **Lens Turbo** workflow -4. Download any missing models (see [model downloads](#lens-model-downloads)), enter your prompt, and click **Queue** +2. **テンプレート**に移動し、「Lens Turbo」を検索 +3. **Lens Turbo**ワークフローを選択 +4. 不足しているモデルをダウンロード（[モデルダウンロード](#lens-model-downloads)を参照）、プロンプトを入力し、**キュー**をクリック ## モデルダウンロード diff --git a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx index 8666d7611..ad37db5f6 100644 --- a/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx +++ b/ja/tutorials/image/newbie-image/newbie-image-exp-0-1.mdx @@ -2,25 +2,25 @@ title: "ComfyUI NewBie-image-Exp0.1 ワークフロー例" description: "NewBie-image-Exp0.1 は、Next-DiT アーキテクチャを基盤とする 35 億パラメータのアニメスタイル文生成画像（text-to-image）モデルであり、XML 構造化プロンプトに対応し、高品質なアニメ画像生成に最適化されています。" sidebarTitle: "NewBie-image-Exp0.1" -translationSourceHash: 22db4899 +translationSourceHash: ec9941c7 translationFrom: tutorials/image/newbie-image/newbie-image-exp-0-1.mdx translationBlockHashes: "_intro": 7f6f1284 - "NewBie-image text-to-image workflow": e7fbe7f5 + "NewBie-image text-to-image workflow": 8828eb35 "Model links": cd75bb7e "Prompt format": 5b819c6c --- import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' -**NewBie-image-Exp0.1** は、NewBieAI Lab が開発した 35 億パラメータの DiT（Diffusion Transformer）モデルで、アニメスタイルの文生成画像タスク専用に設計されています。Next-DiT アーキテクチャを採用しており、非常に詳細で視覚的に印象的なアニメ画像を生成できます。 +**NewBie-image-Exp0.1** は、NewBieAI Lab が開発した 35 億パラメータの DiT モデルで、アニメスタイルのテキストから画像生成に特化しています。Next-DiT アーキテクチャを採用しており、非常に詳細で視覚的に印象的なアニメ画像を生成できます。 **主な特徴**： - **35 億パラメータモデル**：高品質なアニメ画像生成に十分な性能を備えながらも効率的なモデルサイズ -- **Next-DiT アーキテクチャ**：Lumina アーキテクチャの研究に基づき、新たに設計された NewBie 固有のアーキテクチャを採用 +- **Next-DiT アーキテクチャ**：Lumina アーキテクチャの研究に基づき、新たに設計された NewBie アーキテクチャを採用 - **二重テキストエンコーダー**：メインエンコーダーとして Gemma3-4B-it を使用し、Jina CLIP v2 を補助エンコーダーとして活用することで、プロンプトの理解精度を向上 - **FLUX VAE**：FLUX.1-dev の 16 チャネル VAE を採用し、より豊かな色表現と精細なテクスチャディテールを実現 -- **XML 構造化プロンプト**：注意機構の正確なバインディングおよび属性の分離（disentanglement）を可能にする XML 形式をサポート +- **XML 構造化プロンプト**：より良いアテンションバインディングと属性の分離（disentanglement）を実現するため、XML 形式をサポート **関連リンク**： - [GitHub](https://github.com/NewBieAI-Lab/NewBie-image-Exp0.1) @@ -29,7 +29,7 @@ import UpdateReminder from '/snippets/ja/tutorials/update-reminder.mdx' ### NewBie Exp0.1: Anime Generation (`image_newbieimage_exp0_1-t2i`) -Generate detailed anime-style images with NewBie Exp0.1's Next-DiT architecture. Supports XML structured prompts for better multi-character scenes and attribute binding. +NewBie Exp0.1 の Next-DiT アーキテクチャを使用して詳細なアニメスタイル画像を生成します。XML 構造化プロンプトをサポートし、より優れたマルチキャラクターシーンと属性バインディングを実現します。 NewBie-image text-to-image ワークフロープレビュー @@ -48,34 +48,28 @@ Generate detailed anime-style images with NewBie Exp0.1's Next-DiT architecture. -## モデルのダウンロードリンク +## NewBie-image テキストから画像へのワークフロー -**text_encoders** +### NewBie Exp0.1: アニメ生成 (`image_newbieimage_exp0_1-t2i`) -- [gemma_3_4b_it_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/gemma_3_4b_it_bf16.safetensors) -- [jina_clip_v2_bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/text_encoders/jina_clip_v2_bf16.safetensors) +NewBie Exp0.1 の Next-DiT アーキテクチャで詳細なアニメスタイル画像を生成します。マルチキャラクターシーンや属性バインディングを改善するXML構造化プロンプトをサポートします。 -**diffusion_models** +NewBie-image テキストから画像へのワークフロープレビュー -- [NewBie-Image-Exp0.1-bf16.safetensors](https://huggingface.co/Comfy-Org/NewBie-image-Exp0.1_repackaged/blob/main/split_files/diffusion_models/NewBie-Image-Exp0.1-bf16.safetensors) - -**vae** + + + Comfy Cloudで開く + + + JSONをダウンロードするか、テンプレートライブラリで"NewBie-image"を検索してください + + -- [ae.safetensors](https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors) +**出力例** -**モデルの保存場所** +![NewBie-image 出力例](https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/image_newbieimage_exp0_1-t2i.png) -``` -ComfyUI/ -├── models/ -│ ├── text_encoders/ -│ │ ├── gemma_3_4b_it_bf16.safetensors -│ │ └── jina_clip_v2_bf16.safetensors -│ ├── diffusion_models/ -│ │ └── NewBie-Image-Exp0.1-bf16.safetensors -│ └── vae/ -│ └── ae.safetensors -``` + ## プロンプト形式 @@ -142,3 +136,69 @@ NewBie-image-Exp0.1 は以下の 3 種類のプロンプト形式をサポート | `` | 画質に関するタグ | | `` | シーン内に存在する物品 | | `` | その他の追加タグ | + +## プロンプト形式 + +NewBie-imageは、キャラクター生成に最適化されたアニメ画像生成モデルです。XML構造化プロンプトを使用してトレーニングされており、`<>`タグがカテゴリ（``、``など）を定義し、``で閉じます。内部のタグはスタンダードなDanbooruタグです。この構造により、マルチキャラクターシーンにおける属性のバインディングを向上させ、より正確な制御が可能になります。 + +完全なプロンプト作成ガイドについては、[公式ドキュメント](https://ai.feishu.cn/wiki/NZl9wm7V1iuNzmkRKCUcb1USnsh)を参照してください。 + +NewBie-image-Exp0.1は、次の3つのプロンプト形式をサポートしています： +- **自然言語**: スタンダードなテキスト記述 +- **タグ**: Danbooruスタイルのタグ +- **XML構造化形式**: マルチキャラクターシーンに推奨 + +### XML構造化プロンプト + +マルチキャラクターシーンでは、XML構造化プロンプトを使用することで、より正確な画像生成結果が得られ、アテンションの結合と属性の分離が向上します。 + +```xml + +$character_1$ +1girl +chibi, red_eyes, blue_hair, long_hair, hair_between_eyes, head_tilt, tareme, closed_mouth +school_uniform, serafuku, white_sailor_collar, white_shirt, short_sleeves, red_neckerchief, bow, blue_skirt, miniskirt, pleated_skirt, blue_hat, mini_hat, thighhighs, grey_thighhighs, black_shoes, mary_janes +happy, smile +standing, holding, holding_briefcase +center_left + + + +$character_2$ +1girl +chibi, red_eyes, pink_hair, long_hair, very_long_hair, multi-tied_hair, open_mouth +school_uniform, serafuku, white_sailor_collar, white_shirt, short_sleeves, red_neckerchief, bow, red_skirt, miniskirt, pleated_skirt, hair_bow, multiple_hair_bows, white_bow, ribbon_trim, ribbon-trimmed_bow, white_thighhighs, black_shoes, mary_janes, bow_legwear, bare_arms +happy, smile +standing, holding, holding_briefcase, waving +center_right + + + +2girls, multiple_girls + +white_background, simple_background +cheerful +high_resolution, detailed +briefcase +alternate_costume + +``` + +### XMLタグリファレンス + +| タグ | 説明 | +|-----|-------------| +| `` | キャラクター名/識別子 | +| `` | キャラクターの性別（1girl、1boyなど） | +| `` | 外観（髪、目、体型） | +| `` | 服装とアクセサリー | +| `` | 表情 | +| `` | ポーズとアクション | +| `` | 画像内の位置 | +| `` | キャラクター数 | +| ` +white_background, simple_background +cheerful +high_resolution, detailed +briefcase +alternate_costume + +``` + +### XML 태그 참조 + +| 태그 | 설명 | +|-----|-------------| +| `` | 캐릭터 이름/식별자 | +| `` | 캐릭터 성별 (1girl, 1boy 등) | +| `` | 신체적 특징 (머리, 눈, 체형) | +| `` | 의상 및 액세서리 | +| `` | 얼굴 표정 | +| `` | 포즈 및 동작 | +| `` | 이미지 내 위치 | +| `` | 캐릭터 수 | +| ` +white_background, simple_background +cheerful +high_resolution, detailed +briefcase +alternate_costume + +``` + +### XML 标签参考 + +| 标签 | 描述 | +|------|------| +| `` | 角色名称/标识符 | +| `` | 角色性别（1girl、1boy 等） | +| `` | 外观特征（发型、眼睛、体型等） | +| `` | 服装与配饰 | +| `` | 面部表情 | +| `` | 姿势与动作 | +| `` | 在图像中的位置 | +| `` | 角色数量 | +| `