PiDConditioning - ComfyUI Built-in Node Documentation - ComfyUI

Attaches a latent and a degrade_sigma value to a CONDITIONING so it can be used for PiD decoding or upscaling. This lets you control how much the latent is degraded before it is processed.

Inputs

Parameter	Description	Data Type	Required	Range
`positive`	The conditioning data to attach the latent and degrade sigma to.	CONDITIONING	Yes	-
`latent`	The latent (from VAEEncode or a KSampler) to attach to the conditioning.	LATENT	Yes	-
`latent_format`	The format of the latent. Flux1 (16-ch) and Flux2 (128-ch) latents are auto-detected from the channel dimension under “flux”. For SD3 (16-ch), SDXL (4-ch), or QwenImage (16-ch), select manually (default: “flux”).	COMBO	Yes	`"flux"` `"sd3"` `"sdxl"` `"qwenimage"`
`degrade_sigma`	The amount of degradation to apply. 0 means a clean latent. Increase this value to denoise corrupted latent outputs (default: 0.0).	FLOAT	Yes	0.0 to 1.0 (step: 0.01)

Note: When latent_format is set to "flux", the node automatically detects the latent type from the channel dimension: 128 channels are treated as Flux2 latents, while 16 channels are treated as Flux1 latents. Note: An unsupported latent_format value raises an error, but all available options are handled by the node.

Outputs

Output Name	Description	Data Type
`CONDITIONING`	The original conditioning data with the latent and the degrade sigma value attached.	CONDITIONING

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): 24f613b33e7872cb35f458aa5794a3cc4d37ceaecd43725039edef75f4a512bc