ComfyUI-AutomaticCFG
If your image was a pizza and the CFG the temperature of your oven: this is a thermostat that ensures it is always cooked like you want. Also adds a 30% speed i
What is ComfyUI-AutomaticCFG?
If your image was a pizza and the CFG the temperature of your oven: this is a thermostat that ensures it is always cooked like you want. Also adds a 30% speed increase. For ComfyUI / StableDiffusion
Overview of ComfyUI-AutomaticCFG
What it is
ComfyUI-AutomaticCFG is a custom node for ComfyUI and StableDiffusion that implements a self-rescaling CFG (Classifier-Free Guidance) mechanism to prevent image burning and artifacts. It also includes a warp drive node that can increase generation speed by up to 160% when used with the AYS scheduler. The node modifies the CFG behavior during sampling by rescaling it at each step based on the average min/max values, aiming for a desired output intensity. It is licensed under Apache-2.0 and is open source.
Key features
- Self-rescaling CFG that prevents burns and artifacts by automatically adjusting the CFG scale at each step.
- Warp drive node that uses the previous step to determine a negative, cutting generation time by half for approximately 3/4 of the steps, achieving up to 160% speed increase.
- Boost toggle that turns off negative guidance when sigmas are near 1, doubling inference speed.
- Negative strength lerp that interpolates between conditioned and unconditioned outputs.
- Custom attention modifiers and presets (loaded from JSON files) that can modify attention mechanisms, with several built-in presets for different styles.
- Works with SDXL, SD 1.5, PixArt Sigma, and SD3 (without boost features for SD3).
Use cases
- Generating higher quality images with fewer artifacts and less burning compared to standard CFG.
- Enabling more creative and varied outputs from prompts, especially at lower CFG scales (e.g., 4).
- Speeding up inference significantly while maintaining or improving image quality.
- Experimenting with different attention modifications and CFG behaviors through presets.
Getting started
See the repository README for installation and usage instructions.
License and pricing
This project is licensed under Apache-2.0 and is open source. No pricing information is provided beyond the open source license.
Based on the project's public README and repository metadata.