TRELLIS.2 is a generative image-to-3D system that creates high-resolution 3D assets with physically based rendering materials from 2D images. It utilizes a sparse voxel representation to handle complex topologies and internal structures without relying on iso-surface fields.
The project features a structured latent space representation that maps geometry and texture attributes to maintain visual fidelity. It employs an optimization-free geometry reconstruction process to decode latent representations directly into voxel grids and includes a PBR texture generator for synthesizing base color, roughness, and metallic maps.
The system covers 3D content generation and pipeline automation, including bidirectional conversion between textured surface meshes and sparse volumetric grids. It supports the export of generated assets into industry-standard GLB files.
The framework provides capabilities for training generative models from scratch or fine-tuning them on custom datasets.