Skip to content

DataCTE/ComfyUI-DataVoid-nodes

Repository files navigation

ComfyUI-DataVoid-nodes

A collection of custom nodes for ComfyUI focused on model merging and style adaptation.

Nodes

IPAdapter (Same.Energy)

A node that automatically fetches reference images from Same.Energy based on style prompts and applies them using IPAdapter.

Features:

  • Automatically searches and retrieves reference images based on style descriptions
  • Integrates with IPAdapter for style transfer
  • Handles cleanup of temporary images
  • Supports negative images and attention masks
  • Built-in error handling for API requests

Parameters:

  • model: The base model to apply the style to
  • ipadapter: IPAdapter model
  • clip_vision: CLIP Vision model
  • style_prompt: Description of the desired style (e.g., "studio ghibli style")
  • weight: Strength of the style application (0.0 to 5.0)
  • start_at: Starting point for style application (0.0 to 1.0)
  • end_at: Ending point for style application (0.0 to 1.0)
  • image_negative (optional): Negative reference image
  • attn_mask (optional): Attention mask for localized application

Mega Merge SDXL

A sophisticated model merging node designed specifically for SDXL models, using block-wise merging strategies.

Features:

  • Intelligent block-based model merging
  • Preserves non-UNet components
  • Supports complex merging patterns
  • Memory efficient processing

Parameters:

  • model1: First SDXL model
  • model2: Second SDXL model

Merging Strategy:

The node uses a predefined range pattern that alternates between models across different blocks:

ranges = [
(0, 16, 1.0, 0.0), # Model 1 only
(8, 24, 0.0, 1.0), # Model 2 only
(17, 32, 1.0, 0.0), # Model 1 only
# ... and so on
]

Installation

  1. Clone this repository into your ComfyUI custom nodes directory:
cd ComfyUI/custom_nodes
git clone https://github.com/datacte/ComfyUI-DataVoid-nodes.git

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages