Documentation Updates:
• Add comprehensive VAE Color Corrector section with features and usage
• Document four correction methods: luminance_zones, histogram_matching, statistical_matching, advanced_3d_lut
• Explain VAE-aware analysis and automatic bias detection capabilities
• Add workflow diagram showing original → processed → corrected flow
• Update installation dependencies for deep learning features
Feature Documentation:
• Document smart preservation (auto-detect vs mask-based)
• Explain performance features (lock input, smart caching, safe clamping)
• Add usage instructions and recommended settings
• Update basic workflow section with VAE Color Corrector steps
Enhanced Content:
• Add new nodes overview including Film Emulation and Color Palette Extractor
• Update installation dependencies with HuggingFace Hub and related packages
• Add FAQ entries for VAE Color Corrector usage and black squares fix
• Expand feature descriptions for better user understanding
Technical Details:
• Document quantization artifact fixes
• Explain upstream processing control benefits
• Add professional workflow integration guidance
• Include correction method recommendations and best practices
Major Features:
• New VAEColorCorrector node for fixing VAE-induced color shifts in inpainting/img2img workflows
• Four correction methods: luminance_zones, histogram_matching, statistical_matching, advanced_3d_lut
• VAE-aware analysis with automatic bias detection and compensation
• Upstream processing control (lock_input_image) to prevent unnecessary reprocessing
• Enhanced DDColor integration for deep learning colorization with fallback support
Color Correction Improvements:
• Fix quantization artifacts (black squares) with safe color clamping using gradient-preserving sigmoid curves
• Implement maximum shift limiting to prevent extreme color corrections
• Add comprehensive diagnostic warnings for extreme color values
• Balance correction strength across different methods for consistent results
• Enhanced mask-based preservation with edge feathering
Technical Enhancements:
• Semantic selective color adjustments for skin tones, sky, and foliage
• Advanced 3D LUT color mapping with k-means clustering
• Improved face detection parameters for better AI-generated content handling
• VAE bias analysis across luminance zones (shadows/midtones/highlights)
• Smart caching system for expensive computations when input is locked
Node Architecture:
• Modular structure with organized utils and nodes directories
• Enhanced Film Emulation capabilities
• Expanded gitignore for model downloads and cache files
• Updated dependencies for deep learning features
Quality & Stability:
• Graceful fallbacks when advanced libraries unavailable
• Comprehensive error handling and recovery mechanisms
• Performance optimizations for real-time parameter adjustments
• Enhanced debugging output and progress reporting
- New standalone node extracts color palettes from images
- Outputs palette data string, histogram visualization, and palette image
- Compatible with main EasyColorCorrection node reference input
- Configurable palette size (3-16 colors) with extract toggle
- Uses same color extraction algorithms as main node for consistency
- Proper error handling and fallback outputs for missing dependencies
- Enhanced GPU processing with proper tensor device assignment and debugging output
- Completely rewrote reference color matching to use LUT-style luminance zone processing (highlights/midtones/shadows)
- Unified white balance behavior to work consistently with/without AI analysis (-1 cool, +1 warm)
- Fixed lift/gamma/gain controls in Manual mode by adding missing HSV-to-RGB conversion
- Added highlight protection to prevent green artifacts when using skin tone adjustment with high enhancement strength
- Increased lift/gamma/gain strength values for more visible effects (0.3→0.8, 0.6→1.2)
- Removed incorrect enhancement_strength widget hiding in JavaScript frontend
- Added warning messages for enhancement_strength > 1.0 to prevent color artifacts
- Fixed reference_strength to work smoothly from 0.0 to 1.0
- Cleaned up duplicate JavaScript handlers that were causing UI conflicts
- Renamed confusing realtime_preview parameter to lock_input_image
- Updated tooltip to clarify it prevents upstream node reprocessing
- Changed default value from true to false
- Updated README documentation for new parameter
- Standardized naming from "Color Correction" to "Color Corrector"
- Updated JavaScript frontend and all UI references
- Fix UnboundLocalError for 'subfolder' variable in ColorCorrectionViewer
- Add missing parameters (white_balance_strength, enhancement_strength, adjust_for_skin_tone) to batch processing
- Fix 'lighting' variable undefined error in batch GPU processing
- Refactor ColorCorrectionViewer to use VideoHelperSuite patterns for proper video display
- Replace console.log with ComfyUI toast notifications for batch size messages
- Add animated GIF creation for multi-frame sequences with proper metadata
- Implement proper ComfyUI OUTPUT_NODE format with gifs UI structure
- Add file location display and cleanup guidance functionality
- Fix JSON serialization errors by using ComfyUI's output directory structure
- Enhance notification system to respect user settings from settings panel
Corrections to quick install with old naming of the node. The node should be ComfyUI-EasyColorCorrector. The "Correction" name is an old name from when I first started the project. Comfy Registry is making it a pain to update everywhere.
## Core Improvements
- **Professional Manual Mode**: Added separate temperature and tint controls using LAB color space
- **Balanced Parameter Ranges**: Fixed slider ranges for realistic control (contrast: -0.5 to 0.5, brightness: -0.5 to 0.5)
- **Live Preset Updates**: Sliders now automatically update when presets change for full transparency
- **Smart UI Behavior**: AI analysis auto-disabled in Manual mode for pure manual control
## Technical Fixes
- **White Balance Redesign**: Proper -1.0 to 1.0 temperature range with LAB color space processing
- **Preset Mode**: Fixed broken functionality, presets now properly apply visual changes
- **Manual Mode**: Fixed parameter conflicts, lift/gamma/gain now work without AI dependency
- **Skin Tone Processing**: Eliminated green tint on light skin by removing excessive green reduction
- **Enhancement Strength**: Now hidden when AI analysis disabled to reduce UI confusion
## Professional Features
- **Temperature Control**: Real temperature adjustment (blue↔orange) in LAB space, not hue shift
- **Tint Control**: Separate green↔magenta control for professional 2-axis color grading
- **3-Way Color Correction**: Working lift/gamma/gain with balanced strength values
- **Improved Defaults**: More subtle enhancement_strength (0.2) and effect_strength (0.6)
## Updated Documentation
- **README**: Added v1.0.1 features section and updated all technical details
- **Examples**: Updated console output examples and configuration snippets
- **Pro Tips**: Added guidance for new temperature/tint controls and UI behavior
- Enhanced skin tone detection ranges to better handle dark and light skin
- Implemented heavy feathering system with multi-stage Gaussian blurring
- Added distance-based falloff masks for seamless face enhancement blending
- Fixed green cast issues on lighter skin tones with balanced adjustments
- Removed detailed comments throughout codebase while preserving headers
- Added screenshot to README and updated disclaimer formatting
This commit message captures the main improvements:
1. Skin tone processing fixes - Better detection ranges and green cast reduction
2. Face feathering improvements - Heavy multi-stage blurring for seamless blending
3. Code cleanup - Comment removal while keeping structure
