Trellis2UpsampleStage - ComfyUI Built-in Node Documentation - ComfyUI

This node takes the 512-resolution shape latent produced by the first shape-stage sampling pass, upscales it to a higher target resolution, and prepares the conditioning and latent needed for the second shape-stage sampling pass. It attaches per-stage metadata to the conditioning so the model can use it during generation.

Inputs

Parameter	Description	Data Type	Required	Range
`positive`	The positive conditioning to which the upsampling-stage shape metadata is attached.	CONDITIONING	Yes
`negative`	The negative conditioning to which the upsampling-stage shape metadata is attached.	CONDITIONING	Yes
`shape_latent`	The 512-resolution shape latent output from the first shape-stage KSampler.	LATENT	Yes
`vae`	The Trellis2 VAE used to decode the shape latent into high-resolution sparse coordinates.	VAE	Yes
`target_resolution`	Voxel resolution of the upsampled shape. Higher = more detail, more VRAM. Default: 1024.	INT	Yes	1024 - 2048 (step 128)

Outputs

Output Name	Description	Data Type
`positive`	Positive conditioning with the upsampling-stage shape metadata attached.	CONDITIONING
`negative`	Negative conditioning with the upsampling-stage shape metadata attached.	CONDITIONING
`latent`	Zero-filled latent prepared for the second shape-stage sampling pass at the target resolution, carrying the upsampled coordinates and resolution metadata.	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): 0582579bfab487718d69789de508a5ec243d98a0e06ad7165c406154a64677d6