TripoSplatConditioning - ComfyUI Built-in Node Documentation - ComfyUI

This node encodes an input image using the DINOv3 image encoder and the Flux2 VAE to create positive and negative conditioning data for the TripoSplat model. It also generates a fixed-size noise target (latent plus camera data) that serves as the starting point for the KSampler.

Inputs

Parameter	Description	Data Type	Required	Range
`clip_vision`	DINOv3 ViT-H/16+ image encoder	CLIP_VISION	Yes	-
`vae`	Flux2 VAE	VAE	Yes	-
`image`	The input image to encode	IMAGE	Yes	-

Outputs

Output Name	Description	Data Type
`positive`	Positive conditioning data containing DINOv3 features and Flux2 VAE latent	CONDITIONING
`negative`	Negative conditioning data containing zero-filled DINOv3 features and zero-filled Flux2 VAE latent	CONDITIONING
`latent`	The fixed size noise target (latent + camera) for the KSampler	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): 59ebeef272d125a2cc2045f4ff54f99268b1273d0a1fd46f7462e6d312f3a805