FreeU_V2 - ComfyUI Built-in Node Documentation

FreeUV2 enhances image generation quality by applying frequency-based modifications to a diffusion model’s U-Net architecture.

FreeU_V2 enhances image generation quality by applying frequency-based modifications to a diffusion model’s U-Net architecture. It uses configurable scaling factors to adjust feature channels in different blocks, improving output without requiring additional training. The node works by patching the model’s output blocks and filtering high-frequency components of the hidden states.

Inputs

Parameter	Description	Data Type	Required	Range
model	The diffusion model to apply FreeU enhancement to	MODEL	Yes	-
b1	Backbone feature scaling factor for the first block (default: 1.3)	FLOAT	Yes	0.0 - 10.0
b2	Backbone feature scaling factor for the second block (default: 1.4)	FLOAT	Yes	0.0 - 10.0
s1	Skip feature scaling factor for the first block (default: 0.9)	FLOAT	Yes	0.0 - 10.0
s2	Skip feature scaling factor for the second block (default: 0.2)	FLOAT	Yes	0.0 - 10.0

Note: b1, b2, s1, and s2 are advanced parameters hidden by default in the node’s UI. They can be set in steps of 0.01 within the 0.0 - 10.0 range. b1 and s1 control the U-Net block with the most channels (four times the model’s base channel count), while b2 and s2 control the block with half as many channels (twice the base channel count).

Outputs

Output Name	Description	Data Type
model	The enhanced diffusion model with FreeU modifications applied	MODEL

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

---

Source fingerprint (SHA-256): 4cef2af9b04164a8ead25bea9c9bb3311be9224f2539a5cc6edbe97ad8465d65