PiDConditioning - ComfyUI Built-in Node Documentation - ComfyUI

Attaches a latent image and a degrade sigma value to a CONDITIONING data. This is used for PiD (Pixel-in-Detail) decoding or upscaling, allowing you to control how much the latent is degraded before processing.

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.

Outputs

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