NVIDIA unveils Lyra 2.0, turning a single image into a consistent explorable 3D world
NVIDIAAI · x · 2026-11-07
NVIDIA previewed research for SIGGRAPH Asia 2026, headlined by Lyra 2.0: it combines video generation with 3D reconstruction to turn one image into an explorable 3D world along a user-chosen camera path, drawing on earlier views to correct accumulating visual errors. Paper, code and demos are out. Sessions also cover Isaac Sim/Lab, Newton, Cosmos 3, Gaussian splatting and neural rendering.
Related event: NVIDIA Unveils Lyra 2.0: Explorable 3D Worlds from a Single Image(4 posts)→
More from Multimodal
- Phonic launches Phonic-1, tops Speech-to-Speech Index, raises $27.5M Series A — amos_gyamfi · 2026-11-07
- Kimi K3.1 Pro turns a photo into a 3D Three.js scene, with open-sourced template — demian_ai · 2026-11-07
- TischEins adds 8,800+ Midjourney SREF style codes in November, free to browse — tisch_eins · 2026-11-07
- ComfyUI-TabCapture: live tab capture node for video, stills and audio — Sad_Berry_4621 · 2026-11-07
- 20-minute deep dive on AI world models: Genie 3, Cosmos, and the synthetic data revolution — bilawalsidhu · 2026-11-07
- Cinematic AI-Generated Character Portraits Show Photo-Real Film-Quality Results — gen_ericai · 2026-11-07