MarigoldV2PostProcess - ComfyUI Built-in Node Documentation - ComfyUI

This node converts a decoded Marigold V2 prediction into a viewable image. It takes the raw prediction tensor and formats it depending on which type of prediction it is: normalized depth (near objects appear bright), unit-length surface normals, or sRGB albedo.

Inputs

Parameter	Description	Data Type	Required	Range
`image`	The decoded Marigold V2 prediction to be converted into a displayable image.	IMAGE	Yes	-
`prediction`	The type of prediction contained in the input, which determines how the data is converted. `"depth"` normalizes the depth values so that near surfaces are bright and far surfaces are dark, then copies the result into all three color channels. `"normals"` rescales the values to the -1..1 range, normalizes them into unit surface normals, and maps them back to 0..1. `"albedo"` applies a linear-to-sRGB conversion to the values.	COMBO	Yes	`"depth"` `"normals"` `"albedo"`

Outputs

Output Name	Description	Data Type
`IMAGE`	The processed image in the selected representation: normalized depth (grayscale repeated across three channels), unit surface normals, or sRGB albedo.	IMAGE

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): 848b29e2dfcd34c44cf9707b9b26cb13c18e82bb16618a15afbe477a1d620e1e