ETH and Disney Research's Neural Render Proxies Enable Interactive, Differentiable Relighting
ssh4net · x · 2026-11-14
A paper from ETH Zurich, DisneyResearch|Studios and others introduces Neural Render Proxies (NRP) for interactive and differentiable relighting of static scenes with fixed camera and materials. In CG production, lighting tweaks normally require re-rendering full scenes with slow offline renderers, costing minutes to hours per frame. NRP's key insight is decoupling traditional rendering into path sampling and emission computation, letting artists adjust lighting at interactive rates.
More from Multimodal
- HeyGen's video model impresses in first test: faster than Grok, strong anime results — azed_ai · 2026-11-14
- End-to-end Japanese multi-speaker conversation model trained on 130k+ hours runs on consumer GPUs — kastnerkyle · 2026-11-14
- 'Best AI video I've ever seen': AI video hits Hollywood-director level — kimmonismus · 2026-11-14
- MiniMax H3 Image lands on PixVerse with 16 reference images and up to 4K output — aziz4ai · 2026-11-14
- MiniMax H3 image model tests show striking collage, woodcut and felt aesthetics — LudovicCreator · 2026-11-14
- Sentinel: a real-time GPU exploration demo from PurzBeats — PurzBeats · 2026-11-14