ComfyUI-animetimm

中文介绍

ComfyUI custom node for tagging anime images using models from animetimm.

Features

• Performs anime image tag prediction using various TIMM models from animetimm
• Supports classification of general tags, character tags, artist tags, and rating tags
• Supports batch image processing

Installation

Recommended

• Install via ComfyUI-Manager.

Manual

• Navigate to ComfyUI/custom_nodes in your terminal (cmd).
• Clone the repository into the custom_nodes directory using the following command:

  git clone https://github.com/MakkiShizu/ComfyUI-animetimm
  cd ComfyUI-animetimm
  

• Install the dependencies in your Python environment.
  • For Windows portable version, run the following command in ComfyUI\custom_nodes\ComfyUI-animetimm:

    ..\..\..\python_embeded\python.exe -m pip install -r requirements.txt
    

  • If using venv or conda, activate your Python environment first, then run:

    pip install -r requirements.txt
    

Usage

Anime TIMM Classifier

The node provides the following parameters:

• image: Input image
• threshold: Confidence threshold (default 0.35), only tags with confidence above this value will be output
• model_repo: Select model repository (default animetimm/caformer_s36.dbv4-full)
• include_general: Whether to include general tags (default True)
• include_character: Whether to include character tags (default True)
• include_artist: Whether to include artist tags (default False)
• include_rating: Whether to include rating tags (default True)

Output

• tags: Comma-separated tag string
• confidence_scores: Corresponding confidence score list
• raw_output: Raw output with categories and confidence scores
• general_tags: General tags
• character_tags: Character tags
• artist_tags: Artist tags
• rating_tags: Rating tags

Notes

• Model files need to be downloaded on first run, which may take some time
• Models will be cached in the ComfyUI/models/animetimm directory after first use
• GPU is recommended for inference to achieve better performance
• Different models have trade-offs between accuracy and speed, choose according to your needs
• Some models (e.g., eva02_large_patch14_448.dbv4-full) require larger VRAM

Performance

Ranklist: dbv4-full-ranklist

Troubleshooting

Model download fails

The official repository is publicly accessible, but you must log in and accept the terms to access its files and content. If download from the official repository fails, the node will automatically try to download from a backup repository.