Skip to content

Latest commit

 

History

16 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ComfyUI Cliption Integrated Nodes thumbnail

ComfyUI-cliption-integrated

A streamlined fork of pharmapsychotic/comfy-cliption for ComfyUI, providing a single, integrated CLIPtion Beam Search node.

CLIPtion is a fast, lightweight image captioning model built on top of CLIP. This fork updates it for ComfyUI's V3 node scheme and consolidates the node pack into a single, integrated node. Processing has been sped up for GPU inference, and a VRAM offload option has been added.

Workflow

Single Image

single image sample workflow

Multiple Images

multiple images sample workflow

Loads the CLIPtion decoder on demand at caption time (rather than at workflow-load time) and runs beam search to caption an image, ranking the resulting candidates by CLIP similarity to the input image.

Inputs

Name Type Description
clip CLIP CLIP text encoder (must include CLIP-L).
clip_vision CLIP_VISION CLIP vision encoder (must be CLIP-L).
image IMAGE Input Image (s).

Settings

Name Type Description
beam_width INT Number of beams to maintain during search (default: 9).
unload_after_run BOOLEAN Unload the CLIPtion decoder from VRAM after captioning (default: False).
device (optional) ["default", "cpu"] Inference device override.

Output

Name Type Description
STRING STRING (list) Generated caption (s).

CLIPtion Model

The CLIPtion decoder weights (CLIPtion_20241219_fp16.safetensors) are downloaded automatically from easygoing0114/ComfyUI-use-models on Hugging Face the first time the node runs, and are saved to ComfyUI/models/cliption for reuse on subsequent runs. You can also place a copy of the file in that folder yourself ahead of time to skip the automatic download.

Recommended CLIP-SAE-ViT-L-14 Model

For the clip and clip_vision inputs, we recommend zer0int/CLIP-SAE-ViT-L-14, a free, high-accuracy fine-tune of CLIP ViT-L/14 released by zer0int. This model has also been mirrored to the easygoing0114/ComfyUI-use-models repository above.

Download CLIP-SAE-ViT-L-14_FP32.safetensors and place it in both of the following folders:

  • ComfyUI/models/text_encoders
  • ComfyUI/models/clip_vision

Installation

cd ComfyUI/custom_nodes
git clone https://github.com/easygoing0114/ComfyUI-cliption-integrated.git

Restart ComfyUI. The node should now appear in the node search as CLIPtion Beam Search (Integrated).

Credits

License

This project is licensed under the MIT License.

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages