OPEN_SOURCE ↗
YT · YOUTUBE// 25d agoRESEARCH PAPER
RFGCA pushes full-body relighting toward realtime
Relightable Full-Body Gaussian Codec Avatars is a 2025 research system that models full-body humans with realistic skin, hair, and specular effects under changing light and unseen poses. It combines local and non-local light transport modeling to keep quality high while making relighting more computationally practical.
// ANALYSIS
This is one of the clearest signs that codec avatars are moving from “great demo” to “near-deployable graphics tech” for telepresence and immersive apps.
- –Replacing spherical harmonics-heavy relighting with learnable zonal harmonics improves rotational efficiency under articulation-heavy body motion.
- –A dedicated shadow network handles non-local self-occlusion effects between body parts, which is a major failure point in many avatar pipelines.
- –Deferred shading for specular transport helps preserve high-frequency cues like highlights and eye glints that sell realism.
- –The project is positioned as a SIGGRAPH 2025 conference-track paper, signaling strong technical credibility in graphics research circles.
// TAGS
rfgcarelightable-full-body-gaussian-codec-avatarsresearchavatarsrelightinggaussian-splattingcomputer-graphics
DISCOVERED
25d ago
2026-03-17
PUBLISHED
25d ago
2026-03-17
RELEVANCE
8/ 10
AUTHOR
Two Minute Papers