TRELLIS 3D AI Giới thiệu

At its core, TRELLIS uses a sophisticated SLAT representation system that integrates a sparsely-populated 3D grid with dense multiview visual features extracted from powerful vision foundation models. This unique approach allows the system to comprehensively capture both structural (geometry) and textural (appearance) information while maintaining flexibility during decoding. The system employs rectified flow transformers specifically tailored for SLAT as its 3D generation models. When processing an image, TRELLIS analyzes it through advanced neural networks, converting the visual information into a detailed 3D representation that can be output in various formats including Radiance Fields, 3D Gaussians, and meshes. The system also supports local editing capabilities, allowing users to modify specific regions of generated 3D assets without regenerating the entire model.