GET3D

GET3D is a generative model developed by NVIDIA's Toronto AI Lab that directly generates explicit textured 3D meshes with complex topology, rich geometric details, and high-fidelity textures. The model learns from 2D image collections and bridges advances in differentiable surface modeling, differentiable rendering, and 2D generative adversarial networks. It can synthesize textured meshes for various object categories including cars, chairs, animals, motorbikes, human characters, and buildings. The model generates 3D content by creating a 3D signed distance field (SDF) and texture field via latent codes, then extracting a 3D surface mesh and querying the texture field for surface colors. GET3D is specifically designed to produce meshes that can be directly consumed by 3D rendering engines and common 3D software, making the output immediately usable in downstream applications. A key limitation is that it requires training on specific image categories rather than offering a unified model for all 3D object types.