Atlas
Atlas is an omni world model for spatial intelligence, built as a multimodal autoregressive diffusion transformer. It natively operates on text, images, camera poses and 3D depth maps, and can generate, reconstruct and simulate worlds. Outputs include images and video up to one minute at 1440p, novel views from specified camera positions, 360 degree panoramas, depth maps, and worlds as point clouds or 3D Gaussian splats. It can reconstruct a scene from as few as one to three input images.
- Type
- 3D
- Weights
- Closed
More from World Labs
7
-
Flex4DHuman
Multimodal
Flex4DHuman is a multi-view video diffusion model that turns monocular or sparse multi-view videos of dynamic…
-
Marble
Multimodal
Marble is World Labs’ product for creating, editing, and sharing persistent high-fidelity 3D worlds from prom…
-
Marble 1.1
Multimodal
Marble 1.1 is World Labs’ fixed-cost world-generation model for creating higher-quality 3D worlds. It is the…
-
Marble 1.1 Plus
Multimodal
Marble 1.1 Plus is World Labs’ most advanced model for creating the largest 3D worlds. It automatically expan…
About World Labs
1
World Labs develops spatially intelligent world models that perceive, generate, reason about, and interact with three-dimensional environments.