DualCFGGuider - ComfyUI Built-in Node Documentation

The Dual CFG Guider node creates a guidance system for sampling that uses two conditioning inputs together with one negative conditioning input. It applies two separate guidance scales to control how strongly each conditioning influences the generated result, and it supports two ways of combining those scales: “regular” and “nested”.

Inputs

Parameter	Description	Data Type	Required	Range
`model`	The model to use for guidance	MODEL	Yes	-
`cond1`	The first positive conditioning input	CONDITIONING	Yes	-
`cond2`	The second conditioning input, used as the reference between the first positive conditioning and the negative conditioning	CONDITIONING	Yes	-
`negative`	The negative conditioning input	CONDITIONING	Yes	-
`cfg_conds`	Guidance scale applied to the first positive conditioning (default: 8.0)	FLOAT	Yes	0.0 - 100.0
`cfg_cond2_negative`	Guidance scale applied between the second conditioning and the negative conditioning (default: 8.0)	FLOAT	Yes	0.0 - 100.0
`style`	The guidance style to apply (default: “regular”). When set to “nested”, the guidance is applied in a nested fashion	COMBO	Yes	”regular" "nested”

Note: In regular style, cfg_cond2_negative is applied between cond2 and negative, and cfg_conds is applied between cond1 and cond2. In nested style, cfg_conds is applied first between cond1 and cond2, and the resulting prediction is then guided away from negative using cfg_cond2_negative. Note: In regular style, when cfg_cond2_negative is equal to 1.0 the negative conditioning is skipped, and when cfg_conds is also equal to 1.0 the second conditioning is skipped as well. This reduces the number of model evaluations performed.

Outputs

Output Name	Description	Data Type
`GUIDER`	A configured guidance system ready for use with sampling	GUIDER

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): ef28d109149cb545bcd76215fd99535905c1d395222149a029c1c210e2912e97

​Inputs

​Outputs

Inputs

Outputs