The paper presents a method for building high-fidelity relightable head avatars that can be animated to generate novel expressions, using a geometry model based on 3D Gaussians and a relightable appearance model based on learnable radiance transfer. The method achieves real-time relighting with spatially all-frequency reflections and improves the fidelity of eye reflections.

https://arxiv.org/abs//2312.03704

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers