Skip to main content
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
The VAEDecodeHunyuan3D node converts latent representations into 3D voxel data using a VAE decoder. It processes the latent samples through the VAE model with configurable chunking and resolution settings to generate volumetric data suitable for 3D applications.

Inputs

ParameterData TypeRequiredRangeDescription
samplesLATENTYes-The latent representation to be decoded into 3D voxel data
vaeVAEYes-The VAE model used for decoding the latent samples
num_chunksINTYes1000-500000The number of chunks to split the processing into for memory management (default: 8000)
octree_resolutionINTYes16-512The resolution of the octree structure used for 3D voxel generation (default: 256)

Outputs

Output NameData TypeDescription
voxelsVOXELThe generated 3D voxel data from the decoded latent representation