Skip to main content

TripoSplat Conditioning

This node encodes an input image using DINOv3 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

ParameterDescriptionData TypeRequiredRange
clip_visionDINOv3 ViT-H/16+ image encoderCLIP_VISIONYes-
vaeFlux2 VAEVAEYes-
imageThe input image to encodeIMAGEYes-

Outputs

Output NameDescriptionData Type
positivePositive conditioning data containing DINOv3 features and Flux2 VAE latentCONDITIONING
negativeNegative conditioning data containing zero-filled DINOv3 features and zero-filled Flux2 VAE latentCONDITIONING
latentThe fixed size noise target (latent sequence plus camera token) for the KSamplerLATENT
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): 9187a4a020818b9adc762eb41e913086b59d62c47abe92d4bafdb14bc8779f51