ComfyUI-Blip

ComfyUI-Blip

A lightweight and high-speed ComfyUI custom node for generating image captions using BLIP models. Optimized for both GPU and CPU environments to deliver fast and efficient caption generation.

Features

• Generate captions for images using BLIP models
• Support for both base and large BLIP models
• Simple and advanced captioning options
• Automatic model downloading and caching
• High performance on both GPU and CPU

Installation

1. Navigate to your ComfyUI custom nodes directory:

cd ComfyUI/custom_nodes/

2. Clone this repository:

git clone https://github.com/1038lab/ComfyUI-Blip.git

3. Install required dependencies:

pip install -r requirements.txt

Manual Model Download

If automatic download fails, you can manually download the models:

1. Base model:

https://huggingface.co/Salesforce/blip-image-captioning-base/tree/main

2. Large model:

https://huggingface.co/Salesforce/blip-image-captioning-large/tree/main

Download the following files and place them in the corresponding directories:

• pytorch_model.bin
• config.json
• preprocessor_config.json
• special_tokens_map.json
• tokenizer_config.json
• tokenizer.json
• vocab.txt

Usage in ComfyUI

Basic Node

1. Add the "Blip Caption" node to your workflow
2. Connect an image input to the node
3. Configure the following parameters:
   • model_name: Choose between base (faster) or large (more detailed) BLIP model
   • max_length: Maximum length of the generated caption (1-100)
   • use_nucleus_sampling: Enable for more creative captions

Advanced Node

1. Add the "Blip Caption (Advanced)" node to your workflow
2. Connect an image input to the node
3. Configure the following parameters:
   • All basic node parameters
   • min_length: Minimum caption length
   • num_beams: Number of beams for beam search
   • top_p: Top-p value for nucleus sampling
   • force_refresh: Force reload model from disk

License

This repository's code is released under the GPL-3.0 License. - see the LICENSE file for details.

Acknowledgments

• BLIP - The original BLIP model
• ComfyUI - The base framework