RODIN Diffusion

Generates detailed 3D avatars from portraits or text prompts using diffusion models. Rodin: A generative model for sculpting 3d digital avatars using diffusion. It takes a single portrait photo for guided generation or natural language text prompts for creation and editing. The current demo shows pre-generated examples with interactive 360 views, not user-upload generation. As a research demo it appears free for viewing examples, though full generation may require research access or compute resources. It specializes in high-fidelity avatars using diffusion on neural radiance fields with strong text-editing control and multi-view realism. The focus stays on head and upper-body detailed avatars, not complete rigged bodies in the main demos. Original RODIN paper came out in 2023, with follow-ups like RodinHD in 2024 improving portrait fidelity.