Epsilon Scaling - ComfyUI Built-in Node Documentation

Complete documentation for the Epsilon Scaling node in ComfyUI. Learn its inputs, outputs, parameters and usage.

This node implements the Epsilon Scaling method from the research paper “Elucidating the Exposure Bias in Diffusion Models” (arxiv.org/abs/2308.15321v6). It works by scaling the predicted noise during the sampling process to help reduce exposure bias, which can lead to improved quality in the generated images. This implementation uses the “uniform schedule” recommended by the paper for its practicality and effectiveness.

Inputs

Parameter	Description	Data Type	Required	Range
model	The model to which the epsilon scaling patch will be applied.	MODEL	Yes	-
scaling_factor	The factor by which the predicted noise is scaled. A value greater than 1.0 reduces the noise, while a value less than 1.0 increases it (default: 1.005).	FLOAT	No	0.5 - 1.5

Outputs

Output Name	Description	Data Type
model	A patched version of the input model with the epsilon scaling function applied to its sampling process.	MODEL

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/Epsilon Scaling/en.md)

---

Source fingerprint (SHA-256): 245d230ec2db7298493d4fbe58c3d96c15821985da3dc12aaccf64e2bf7922a1