Researchers have developed Generative Relightable Avatars (GRA), a novel method for creating photorealistic, person-specific avatars that can be rendered from any viewpoint and relit under different environment maps. GRA employs a hybrid approach, combining physics-based relighting with generative refinement. It optimizes material parameters and renders a coarse appearance, then refines textures using a feed-forward model to capture complex illumination effects. Finally, a diffusion model transforms these renderings into temporally coherent, high-detail videos while maintaining 3D control. AI
IMPACT This research could lead to more realistic virtual characters in gaming, film, and metaverse applications, enhancing immersion and creative possibilities.
RANK_REASON This is a research paper detailing a new method for generating 3D avatars. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Generative Relightable Avatars
- Gotit.pub
- Hugging Face
- Kunwar Maheep Singh
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
