✨ Enhance your ComfyUI workflows with Divergent Nodes – a growing collection of custom nodes designed to expand your creative possibilities!
This repository offers a curated set of custom nodes that introduce new functionalities and streamline your ComfyUI experience.
For detailed documentation, visit the Divergent Nodes Wiki.
Nodes Currently Included:
- ✨ Divergent CLIP Token Counter: Precisely count CLIP tokens in your text prompts.
- 🐬 DolphinVision Node: Generate text descriptions of images using the DolphinVision 7b multimodal model.
-
Clone the Repository:
git clone https://github.com/thedivergentai/divergent_nodes.git custom_nodes/divergent_nodes
(Or download as a ZIP and extract to custom_nodes.)
-
Install Dependencies:
cd custom_nodes/divergent_nodes pip install -r requirements.txt -
Restart ComfyUI: Restart to load the new nodes.
Description: Accurately count CLIP tokens in a text string. Essential for workflows with token limits.
Inputs: text (STRING, Multiline)
Outputs: token_count (INT)
Key Features: Zero token handling, robust text processing, CLIP standard compliance.
Description: Generate text descriptions of images using the DolphinVision 7b model (hosted on Hugging Face).
Inputs: image (IMAGE TENSOR), prompt (STRING)
Outputs: text (STRING)
Model Loading:
load_model(cache=False): Loads the model.cache: Uses a cached version if available.
IS_CHANGED(image, prompt, **kwargs): Optimizes performance by checking for input changes.unload(): Unloads the model from memory.
Key Features: Multimodal generation, flexible prompting, automatic model download.
See the Troubleshooting/FAQ page on the Wiki.
- Additional Nodes
- User Interface Enhancements
- Community Contributions (see the Contributing page on the Wiki)
- ComfyUI
- PyTorch
- Transformers
- Hugging Face Hub
- GitPython
Designed for the latest version of ComfyUI. Report compatibility issues on the GitHub repository.
Divergent Nodes for ComfyUI is released under the MIT License. See the LICENSE file for details.