HunyuanRefinerLatent - ComfyUI Built-in Node Documentation - ComfyUI

The HunyuanRefinerLatent node prepares conditioning and latent data for the Hunyuan video refinement process. It attaches the input latent image data to both the positive and negative conditioning, applies a noise augmentation value to them, and creates a new zero-filled latent with 32 channels for further processing.

Inputs

Parameter	Description	Data Type	Required	Range
`positive`	The positive conditioning input to be processed	CONDITIONING	Yes	-
`negative`	The negative conditioning input to be processed	CONDITIONING	Yes	-
`latent`	The latent representation input, used as the latent image data for the conditioning and to define the dimensions of the output latent	LATENT	Yes	-
`noise_augmentation`	The amount of noise augmentation to apply (default: 0.10). This parameter is shown in the advanced section of the node.	FLOAT	Yes	0.0 - 1.0

Outputs

Output Name	Description	Data Type
`positive`	The processed positive conditioning with the latent image data attached and the noise augmentation applied	CONDITIONING
`negative`	The processed negative conditioning with the latent image data attached and the noise augmentation applied	CONDITIONING
`latent`	A new latent filled with zeros, with the same batch size and the same last three dimensions as the input latent, and 32 channels	LATENT

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): 4c5669cf2ad5ba00e176876741b7d8d3f092cc58d2163871a10fd769ee4ff84c