ComfyUI Any Color Correction Node

This is a fork of ComfyUI-ColorCorrection by zaheenrahman

A flexible custom node for ComfyUI that performs color correction on any masked region (e.g., clothing, background, skin) based on a reference image. Originally designed to fix color shifts during face-swapping. It now supports broader use cases with batch processing, optional masks, and multiple correction methods.

🔄 Changelog

(2025-06-21)

• ✅ Added batch dimension support (works with 1-to-N and N-to-N inputs)
• ✅ Fully optional mask_ref and mask_target inputs
• ✅ Safer handling of shape mismatches (with proper error messages)
• ✅ Better normalized color difference metric (for thresholding)
• ✅ Ensured single-image input returns a single tensor output

Features

• Multiple Color Correction Methods: Choose from statistical, histogram, or LAB color space transfers
• Fine Control: Adjust the strength and threshold of the color correction
• Preserve Luminosity: Option to maintain the original brightness while correcting colors
• Masked Processing: Only applies corrections to specified regions

Installation

1. Navigate to your ComfyUI custom nodes directory:

   cd ComfyUI/custom_nodes/

2. Clone this repository:

   git clone https://github.com/yourusername/ComfyUI_AnyColorCorrection

3. Install requirements:

   pip install -r ComfyUI_AnyColorCorrection/requirements.txt

4. Restart ComfyUI

Usage

The node requires the following inputs:

⚙️ Node Inputs

Input Name	Type	Required	Default	Description
image_ref	IMAGE	✅	—	Reference image to draw color information from.
image_target	IMAGE	✅	—	Target image where color correction will be applied.
mask_ref	MASK	❌	None	Optional mask for the reference image to localize color sampling.
mask_target	MASK	❌	None	Optional mask for the target image to localize correction.
color_threshold	FLOAT	❌	0.15	Minimum color difference to trigger correction. Range: 0.01 to 1.0.
strength	FLOAT	❌	0.8	Strength of color correction. Range: 0.0 (off) to 1.0 (full correction).
preserve_luminosity	BOOLEAN	❌	True	Whether to preserve original brightness/luminosity in correction.
method	ENUM	❌	lab_transfer	Method to use: statistical, histogram, or lab_transfer.

Example Workflow

1. Load your reference and target images
2. Create masks for the areas in both images (using segmentation or manual masking)
3. Connect all inputs to the Color Correction node
4. Adjust parameters as needed

Advanced Options

Color Correction Methods

• Statistical: Uses mean and standard deviation to transfer color characteristics
• Histogram: Matches color distributions using histogram equalization
• LAB Transfer: Works in LAB color space for more perceptually accurate corrections

Acknowledgements

This node was originally created by zaheenrahman to solve color shifts that commonly occur during face swapping, particularly when the original and swapped faces have different skin tones or when lighting conditions change. Extended to support more general, region-specific color transfer.

License

MIT License