Skip to content

docs: sync built-in nodes documentation (weekly update 0907) #1620

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
lin-bot23 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

docs: sync built-in nodes documentation (weekly update 0907) #1620

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
50 changes: 50 additions & 0 deletions built-in-nodes/BlockSparseAttention.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,50 @@
---
title: "BlockSparseAttention - ComfyUI Built-in Node Documentation"
description: "Documentation for BlockSparseAttention node."
sidebarTitle: "BlockSparseAttention"
icon: "circle"
mode: wide
---

## Overview

The Block Sparse Attention node applies a block-sparse attention mechanism to a ComfyUI model, reducing attention computation by allowing each query block to attend only a selected subset of key blocks exactly.

## Inputs

| Parameter | Description | Data Type | Required | Range |
|-----------|-------------|-----------|----------|-------|
| `model` | The ComfyUI model to apply the block-sparse attention to. | MODEL | Yes | N/A |
| `selection` | The selection method for choosing which key blocks to attend to. | DYNAMIC_COMBO | Yes | Options: Sol-Attn (adaptive tau), top-k (SLA), VSA (FastVideo) |
| `tau` | The threshold in score-distribution sigmas for Sol-Attn (adaptive tau) selection. | FLOAT | No | default: 1.3, min: 0.0, max: 4.0, step: 0.05 |
| `keep_percent` | The percentage of key blocks each query block keeps exactly for top-k (SLA) selection. | FLOAT | No | default: 10.0, min: 0.5, max: 95.0, step: 0.5 |
| `start_percent` | The percentage of the schedule before which the model uses dense attention. | FLOAT | No | default: 0.2, min: 0.0, max: 1.0, step: 0.01 |
| `end_percent` | The percentage of the schedule after which the model uses dense attention. | FLOAT | No | default: 1.0, min: 0.0, max: 1.0, step: 0.01 |
| `dense_blocks` | A string representing the transformer blocks that always run dense attention. | STRING | No | default: "" |
| `min_tokens` | The minimum number of tokens in a sequence for which the model uses dense attention. | INT | No | default: 12288, min: 0, max: 1 <&lt; 20, step: 512 |
| `extra_tokens` | The number of extra top-scoring tokens each query block attends beyond its selected blocks. | INT | No | default: 256, min: 0, max: 256, step: 64 |
| `sink_conditioning` | The MiniMax-H3 conditioning rows to use for sink conditioning. | COMBO | No | Options: exact_kv, exact_kv_and_rows, off |
| `verbose` | Whether to enable verbose logging. | BOOLEAN | No | default: False |

## Outputs

| Output Name | Description | Data Type |
|-------------|-------------|-----------|
| `model` | The ComfyUI model with block-sparse attention applied. | MODEL |

### Notes

- The `selection` parameter determines how key blocks are chosen. The options are:
- Sol-Attn (adaptive tau): Each query block attends a selected subset of key blocks based on an adaptive threshold.
- top-k (SLA): Each query block keeps a fixed percentage of key blocks exactly.
- VSA (FastVideo): Each query block keeps a fixed percentage of video cubes exactly, using FastH3-VSA's cube tiling and coarse branch.
- The `dense_blocks` parameter allows you to specify transformer blocks that always run dense attention.
- The `min_tokens` parameter sets the minimum number of tokens in a sequence for which the model uses dense attention.
- The `extra_tokens` parameter allows you to specify the number of extra top-scoring tokens each query block attends beyond its selected blocks.
- The `sink_conditioning` parameter determines the MiniMax-H3 conditioning rows to use for sink conditioning.
- The `verbose` parameter enables verbose logging.

> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/BlockSparseAttention/en.md)

---
**Source fingerprint (SHA-256):** `6a27aee45593883f5958ae1aac74a2077362742a0fdf74fc9dbfd68eddc6d259`
46 changes: 29 additions & 17 deletions built-in-nodes/ByteDanceFirstLastFrameNode.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,36 +1,48 @@
---
title: "ByteDanceFirstLastFrameNode - ComfyUI Built-in Node Documentation"
description: "This node generates a video using a text prompt along with first and last frame images."
description: "This node generates a video by using a text prompt along with the first and last frames of an image."
sidebarTitle: "ByteDanceFirstLastFrameNode"
icon: "circle"
mode: wide
---

This node generates a video using a text prompt along with first and last frame images. It takes your description and the two key frames to create a complete video sequence that transitions between them. The node provides various options to control the video's resolution, aspect ratio, duration, and other generation parameters.
This node generates a video by using a text prompt along with the first and last frames of an image. It creates a seamless transition between the two frames, resulting in a complete video sequence. The node offers a range of options to customize the video's resolution, aspect ratio, duration, and additional generation parameters.

## Inputs

| Parameter | Description | Data Type | Required | Range |
| --- | --- | --- | --- | --- |
| `model` | The model to use for video generation (default: `"seedance-1-0-lite-i2v-250428"`). | COMBO | Yes | `"seedance-1-5-pro-251215"`<br />`"seedance-1-0-pro-250528"`<br />`"seedance-1-0-lite-i2v-250428"` |
| `prompt` | The text prompt used to generate the video. | STRING | Yes | - |
| `first_frame` | First frame to be used for the video. Must be between 300x300 and 6000x6000 pixels, with an aspect ratio between 0.4 and 2.5. | IMAGE | Yes | - |
| `last_frame` | Last frame to be used for the video. Must be between 300x300 and 6000x6000 pixels, with an aspect ratio between 0.4 and 2.5. | IMAGE | Yes | - |
| `resolution` | The resolution of the output video. | COMBO | Yes | `"480p"`<br />`"720p"`<br />`"1080p"` |
| `aspect_ratio` | The aspect ratio of the output video (default: `"adaptive"`). | COMBO | Yes | `"adaptive"`<br />`"16:9"`<br />`"4:3"`<br />`"1:1"`<br />`"3:4"`<br />`"9:16"`<br />`"21:9"` |
| `duration` | The duration of the output video in seconds (default: 5). Note: For the `seedance-1-5-pro-251215` model, the minimum supported duration is 4 seconds. | INT | Yes | 3 - 12 |
| `seed` | Seed to use for generation (default: 0). | INT | No | 0 - 2147483647 |
| `camera_fixed` | Specifies whether to fix the camera. The platform appends an instruction to fix the camera to your prompt, but does not guarantee the actual effect (default: False). | BOOLEAN | No | - |
| `watermark` | Whether to add an "AI generated" watermark to the video (default: False). | BOOLEAN | No | - |
| `generate_audio` | This parameter is ignored for any model except `seedance-1-5-pro-251215` (default: False). | BOOLEAN | No | - |
|-----------|-------------|-----------|----------|-------|
| `model` | The model to use for video generation. Select from available options (default: `"seedance-1-5-pro-251215"`). | COMBO | Yes | `"seedance-1-5-pro-251215"`<br />`"seedance-1-0-pro-250528"` |
| `prompt` | The text prompt used to generate the video. This prompt should not contain specific parameters such as resolution, ratio, duration, seed, camerafixed, or watermark. | STRING | Yes | - |
| `first_frame` | The first frame to be used for the video. The image must be between 300x300 and 6000x6000 pixels and have an aspect ratio between 0.4 and 2.5. | IMAGE | Yes | - |
| `last_frame` | The last frame to be used for the video. The image must be between 300x300 and 6000x6000 pixels and have an aspect ratio between 0.4 and 2.5. | IMAGE | Yes | - |
| `resolution` | The resolution of the output video. Choose from the available options (default: `"480p"`). | COMBO | Yes | `"480p"`<br />`"720p"`<br />`"1080p"` |
| `aspect_ratio` | The aspect ratio of the output video. Select from the available options (default: `"adaptive"`). | COMBO | Yes | `"adaptive"`<br />`"16:9"`<br />`"4:3"`<br />`"1:1"`<br />`"3:4"`<br />`"9:16"`<br />`"21:9"` |
| `duration` | The duration of the output video in seconds (default: 5). For the `seedance-1-5-pro-251215` model, the minimum supported duration is 4 seconds. | INT | Yes | 3 - 12 |
| `seed` | The seed to use for generation (default: 0). This parameter is optional. | INT | No | 0 - 2147483647 |
| `camera_fixed` | Specifies whether to fix the camera in the video. The platform appends an instruction to fix the camera to your prompt, but the actual effect is not guaranteed (default: False). | BOOLEAN | No | - |
| `watermark` | Determines whether to add an "AI generated" watermark to the video (default: False). | BOOLEAN | No | - |
| `generate_audio` | This parameter is ignored for all models except `seedance-1-5-pro-251215` (default: False). | BOOLEAN | No | - |

## Outputs

| Output Name | Description | Data Type |
| --- | --- | --- |
| `output` | The generated video file | VIDEO |
|-------------|-------------|-----------|
| `output` | The generated video file. | VIDEO |

## Notes

- The `model` parameter determines the capabilities and limitations of the video generation process.
- The `prompt` should be creative and clear, as it will guide the video generation.
- The `first_frame` and `last_frame` images should be representative of the desired video content.
- The `resolution` and `aspect_ratio` parameters will affect the final output video's quality and dimensions.
- The `duration` parameter sets the length of the video, with a minimum of 3 seconds and a maximum of 12 seconds.
- The `seed` parameter is optional and can be used for reproducibility of the video generation process.
- The `camera_fixed` parameter is an advanced option that may not always result in the expected effect.
- The `watermark` parameter can be used to add a watermark to the video, indicating that it was generated by an AI.
- The `generate_audio` parameter is currently ignored for all models except `seedance-1-5-pro-251215`.

> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/ByteDanceFirstLastFrameNode/en.md)

---
**Source fingerprint (SHA-256):** `56b619d191c0a78ac51eaddf93c86f6ec85a486cc4e6d0e021c1e58b4df29189`
**Source fingerprint (SHA-256):** `ae0f3a34a21baad7f04f6917e98d16dc64496479a050896869ec6693a9a9ebaf`
28 changes: 16 additions & 12 deletions built-in-nodes/ByteDanceImageToVideoNode.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,27 +1,29 @@
---
title: "ByteDanceImageToVideoNode - ComfyUI Built-in Node Documentation"
description: "The ByteDance Image to Video node generates videos using ByteDance models through an API based on an input image and text prompt."
description: "The ByteDance Image to Video node generates a video from an input image and a text prompt using ByteDance's API."
sidebarTitle: "ByteDanceImageToVideoNode"
icon: "circle"
mode: wide
---

The ByteDance Image to Video node generates videos using ByteDance models through an API based on an input image and text prompt. It takes a starting image frame and creates a video sequence that follows the provided description. The node offers various customization options for video resolution, aspect ratio, duration, and other generation parameters.
# ByteDance Image to Video Node

The ByteDance Image to Video node generates a video from an input image and a text prompt using ByteDance's API. It creates a video sequence that visually represents the provided description, with options to customize the output's resolution, aspect ratio, duration, and other parameters.

## Inputs

| Parameter | Description | Data Type | Required | Range |
|-----------|-------------|-----------|----------|-------|
| `model` | The ByteDance model to use for video generation (default: `"seedance-1-0-pro-fast-251015"`). | STRING | Yes | `"seedance-1-5-pro-251215"`<br />`"seedance-1-0-pro-250528"`<br />`"seedance-1-0-lite-i2v-250428"`<br />`"seedance-1-0-pro-fast-251015"` |
| `model` | The ByteDance model to use for video generation. Available options are: <br />`"seedance-1-5-pro-251215"`<br />`"seedance-1-0-pro-250528"`<br />`"seedance-1-0-pro-fast-251015"` | STRING | Yes | As listed above |
| `prompt` | The text prompt used to generate the video. Must be at least 1 character long after trimming whitespace. | STRING | Yes | - |
| `image` | First frame to be used for the video. Must be between 300x300 and 6000x6000 pixels, with an aspect ratio between 0.4 and 2.5. | IMAGE | Yes | - |
| `resolution` | The resolution of the output video. | STRING | Yes | `"480p"`<br />`"720p"`<br />`"1080p"` |
| `aspect_ratio` | The aspect ratio of the output video. | STRING | Yes | `"adaptive"`<br />`"16:9"`<br />`"4:3"`<br />`"1:1"`<br />`"3:4"`<br />`"9:16"`<br />`"21:9"` |
| `duration` | The duration of the output video in seconds (default: 5). For the `seedance-1-5-pro-251215` model, the minimum supported duration is 4 seconds. | INT | Yes | 3 - 12 |
| `seed` | Seed to use for generation (default: 0). | INT | No | 0 - 2147483647 |
| `camera_fixed` | Specifies whether to fix the camera. The platform appends an instruction to fix the camera to your prompt, but does not guarantee the actual effect (default: False). | BOOLEAN | No | - |
| `watermark` | Whether to add an "AI generated" watermark to the video (default: False). | BOOLEAN | No | - |
| `generate_audio` | This parameter is ignored for any model except `seedance-1-5-pro-251215` (default: False). | BOOLEAN | No | - |
| `image` | The first frame to be used for the video. The image must be between 300x300 and 6000x6000 pixels, with an aspect ratio between 0.4 and 2.5. | IMAGE | Yes | - |
| `resolution` | The resolution of the output video. Available options are: <br />`"480p"`<br />`"720p"`<br />`"1080p"` | STRING | Yes | As listed above |
| `aspect_ratio` | The aspect ratio of the output video. Available options are: <br />`"adaptive"`<br />`"16:9"`<br />`"4:3"`<br />`"1:1"`<br />`"3:4"`<br />`"9:16"`<br />`"21:9"` | STRING | Yes | As listed above |
| `duration` | The duration of the output video in seconds. For the `seedance-1-5-pro-251215` model, the minimum supported duration is 4 seconds. | INT | Yes | 3 - 12 |
| `seed` | Seed to use for generation. Optional, with a default value of 0. | INT | No | 0 - 2147483647 |
| `camera_fixed` | Specifies whether to fix the camera. The platform appends an instruction to fix the camera to your prompt, but does not guarantee the actual effect. Optional, with a default value of False. | BOOLEAN | No | - |
| `watermark` | Whether to add an "AI generated" watermark to the video. Optional, with a default value of False. | BOOLEAN | No | - |
| `generate_audio` | This parameter is ignored for any model except `seedance-1-5-pro-251215`. Optional, with a default value of False. | BOOLEAN | No | - |

**Note:** The prompt must not contain the following words (case-insensitive): `resolution`, `ratio`, `duration`, `seed`, `camerafixed`, `watermark`. These parameters are set via their dedicated inputs.

Expand All @@ -31,7 +33,9 @@ The ByteDance Image to Video node generates videos using ByteDance models throug
|-------------|-------------|-----------|
| `output` | The generated video file based on the input image and prompt parameters. | VIDEO |

The documentation provided above reflects the information extracted from the source code and adheres to the guidelines specified for the structure and content of the documentation.

> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/ByteDanceImageToVideoNode/en.md)

---
**Source fingerprint (SHA-256):** `3f95a381c8df083ec4cb417de0146efa68c0251b14697f6cc727be66c87cc761`
**Source fingerprint (SHA-256):** `9f4ad18642533689f9c664f2ca6a4ce8e92c8698754cdf9b6bb2d2735bc80415`
Loading
Loading