Advanced Sampling & Model Control EnhancementsThis release delivers significant improvements to sampling algorithms and model control systems, particularly benefiting advanced AI researchers and workflow creators:
These improvements make ComfyUI more robust for production workflows while expanding creative possibilities for AI artists working with advanced sampling techniques and model control systems.
Enhanced Model Support & Workflow ReliabilityThis release brings significant improvements to model compatibility and workflow stability:
Expanded Model Documentation: Added comprehensive support documentation for Flux Kontext and Omnigen 2 models, making it easier for creators to integrate these powerful models into their workflows
VAE Encoding Improvements: Removed unnecessary random noise injection during VAE encoding, resulting in more consistent and predictable outputs across workflow runs
Memory Management Fix: Resolved a critical memory estimation bug specifically affecting Kontext model usage, preventing out-of-memory errors and improving workflow stability
These changes enhance the reliability of advanced model workflows while maintaining ComfyUI’s flexibility for AI artists and researchers working with cutting-edge generative models.
Cosmos Predict2 Support: Full implementation for both text-to-image (2B and 14B models) and image-to-video generation workflows, expanding video creation capabilities
Enhanced Flux Compatibility: Chroma Text Encoder now works seamlessly with regular Flux models, improving text conditioning quality
LoRA Training Integration: New native LoRA training node using weight adapter scheme, enabling direct model fine-tuning within ComfyUI workflows
Performance & Hardware Optimizations
AMD GPU Enhancements: Enabled FP8 operations and PyTorch attention on GFX1201 and other compatible AMD GPUs for faster inference
Apple Silicon Fixes: Addressed long-standing FP16 attention issues on Apple devices, improving stability for Mac users
Flux Model Stability: Resolved black image generation issues with certain Flux models in FP16 precision
Advanced Sampling Improvements
Rectified Flow (RF) Samplers: Added SEEDS and multistep DPM++ SDE samplers with RF support, providing more sampling options for cutting-edge models
ModelSamplingContinuousEDM: New cosmos_rflow option for enhanced sampling control with Cosmos models
Memory Optimization: Improved memory estimation for Cosmos models with uncapped resolution support
Developer & Integration Features
SQLite Database Support: Enhanced data management capabilities for custom nodes and workflow storage
PyProject.toml Integration: Automatic web folder registration and settings configuration from pyproject files
Frontend Flexibility: Support for semver suffixes and prerelease frontend versions for custom deployments
Tokenizer Enhancements: Configurable min_length settings with tokenizer_data for better text processing
Quality of Life Improvements
Kontext Aspect Ratio Fix: Resolved widget-only limitation, now works properly in all connection modes
SaveLora Consistency: Standardized filename format across all save nodes for better file organization
Python Version Warnings: Added alerts for outdated Python installations to prevent compatibility issues
WebcamCapture Fixes: Corrected IS_CHANGED signature for reliable live input workflows
This release significantly expands ComfyUI’s model ecosystem support while delivering crucial stability improvements and enhanced hardware compatibility across different platforms.
Pillow Compatibility: Updated deprecated API calls to maintain compatibility with latest image processing libraries
ROCm Support: Improved version detection for AMD GPU users
Template Updates: Enhanced project templates for custom node development
These updates strengthen ComfyUI’s foundation for complex AI workflows while making the platform more accessible to new users through improved documentation and helper tools.