LTXVDualCFGGuider - ComfyUI Built-in Node Documentation

This node creates a guided sampling object (CFG guider) for LTXV-AV models. It applies a separate guidance scale to the video part and the audio part of a packed LTXV-AV latent, letting you control the influence of conditioning on each modality independently. If the two scales are equal, or if the latent does not contain separate video and audio components, a single overall scale is used instead.

Inputs

Parameter	Description	Data Type	Required	Range
`model`	The model to use when sampling.	MODEL	Yes	-
`positive`	Positive conditioning to guide generation toward.	CONDITIONING	Yes	-
`negative`	Negative conditioning to steer generation away from.	CONDITIONING	Yes	-
`video_cfg`	Guidance strength applied to the video modality of the latent (default: 3.0).	FLOAT	Yes	0.0 to 100.0
`audio_cfg`	Guidance strength applied to the audio modality of the latent (default: 7.0).	FLOAT	Yes	0.0 to 100.0

Note: When video_cfg and audio_cfg are equal (or very close in value), the guider uses that value as a single CFG scale for the whole latent. If the latent is not a packed LTXV-AV latent, only the video_cfg value is used.

Outputs

Output Name	Description	Data Type
`guider`	The configured CFG guider to be passed to a sampler node.	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): 8b5ea32d0e73ab4f9b9f053ac7513d621fcc047e1ff468b6d0b5dd2aa3ff791a

​Inputs

​Outputs

Inputs

Outputs