NVIDIA Lyra 2.0 turns a single image into an explorable 3D world
Wccftech reports that NVIDIA's Spatial Intelligence Lab has showcased Lyra 2.0, which combines AI video generation with 3D reconstruction to turn a photograph into a navigable, persistent environment. The system uses geometry-guided spatial memory and training on its own outputs to limit spatial forgetting and temporal drift, while its output can be converted into 3D Gaussian Splats or polygonal meshes for real-time rendering and export to NVIDIA Isaac Sim. NVIDIA has published the implementation on GitHub and research-use model weights on Hugging Face, although the report says dynamic scenes and occasional visual inconsistencies remain limitations.
Lyra 2.0 provides an implementable route from one image to persistent 3D environments that can be rendered in real time and used in simulation, while its research-use weights and stated limitations define the current scope of the release.
NVIDIA’s Lyra 2.0 Turns a Single Image Into an Explorable 3D World, With Persistent Environments and Real-Time Rendering
Sebastian Castellanos•Nov 9, 2026 at 08:10am EST
Add to Preferred Sources
Some source content is omitted. Continue reading on the original site.
#NVIDIA#NVIDIA Lyra 2.0#NVIDIA Spatial Intelligence Lab#NVIDIA Isaac Sim#NVIDIA GB200