SyncLight accepted at NeurIPS 2026: relight multi-view scenes with one click, no camera poses
anand_bhattad · x · 2026-10-13
SyncLight has been accepted at NeurIPS 2026. Users click a light in one image, pick its color and intensity, and the system relights every other view of a multi-view scene to match.
- Built by repurposing an image generative model for cross-view light propagation
- Requires no camera poses, no 3DGS, and no distillation
- Aims at consistent lighting control for multi-view / novel-view settings
Related event: SyncLight Accepted to NeurIPS 2026: Single-Click Multi-View Relighting(2 posts)→
More from Multimodal
- Making an AI Music Video With Neural Frames: 'Castles Made of Sound' — markjeffrey · 2026-10-13
- Muse Wonder Video Model Tested: 'Pretty Rough' Output, Every Narration in Australian Accent — danshipper · 2026-10-13
- Dev who ran a fake AI Instagram model to 40k followers open-sources a ComfyUI photoshoot node — Prestigious_Bill221 · 2026-10-13
- Director made an $80M-looking holiday film with AI for just $1.5M — lmoroney · 2026-10-13
- Vidu Q4 video model debuts to strong first impressions — TomLikesRobots · 2026-10-13
- 7-minute AI film rendered as one sequence in 3 weeks, toolchain open-sourced — Super_Range45 · 2026-10-13