What DreamFusion 3D does
DreamFusion is a 2022 method for text-to-3D generation that combines neural rendering with a pretrained text-to-2D image diffusion model. The page says it uses Imagen as the diffusion prior and synthesizes diverse 3D objects from text. Named components include Score Distillation Sampling and a NeRF-based 3D scene representation, plus regularizers that improve geometry. Outputs include normals, depth and surface geometry, relightable with a Lambertian shading model, and NeRF models can be exported to meshes with marching cubes. The authors listed are Ben Poole, Ajay Jain, Jonathan T. Barron and Ben Mildenhall. The page links a paper, gallery and project placeholder but no code repository or license.
How DreamFusion 3D compares with its closest alternatives
All 5 closest alternatives have a free plan, as DreamFusion 3D does. 4 of the 5 offer an API, as DreamFusion 3D does. DreamFusion 3D offers a self-hosted version. Of the 4 alternatives with platforms on record, 2 offer a self-hosted version. DreamFusion 3D was founded in 2022; the alternatives were founded between 2022 and 2024. The alternatives with a known headquarters are in the United Kingdom and the United States.