mcmonkeyprojects/sd-dynamic-thresholding
A Stable Diffusion extension enabling higher CFG scales without color issues by clamping latents during the diffusion sampling process.
Not currently ranked — collecting fresh signals.
star history
This extension implements the Dynamic Thresholding (CFG Scale Fix) technique for Stable Diffusion image generation. It modifies the latent clamping behavior during sampling steps, addressing the color artifacts that typically appear when using high Classifier-Free Guidance scales. The technique was originally developed by Birch-san and has been ported across multiple Stable Diffusion interfaces including SwarmUI, ComfyUI, and Automatic1111 WebUI, making it accessible to users working with diffusion models.
Frequently asked
- What is mcmonkeyprojects/sd-dynamic-thresholding?
- A Stable Diffusion extension enabling higher CFG scales without color issues by clamping latents during the diffusion sampling process.
- Is sd-dynamic-thresholding open source?
- Yes — mcmonkeyprojects/sd-dynamic-thresholding is open source, released under the MIT license.
- What language is sd-dynamic-thresholding written in?
- mcmonkeyprojects/sd-dynamic-thresholding is primarily written in Python.
- How popular is sd-dynamic-thresholding?
- mcmonkeyprojects/sd-dynamic-thresholding has 1.2k stars on GitHub.
- Where can I find sd-dynamic-thresholding?
- mcmonkeyprojects/sd-dynamic-thresholding is on GitHub at https://github.com/mcmonkeyprojects/sd-dynamic-thresholding.