The Omni World Model
Meet Atlas
Explore camera-controlled video, 3D reconstruction, and spatial simulation. Join the waitlist for future hands-on tools.
What is Atlas?
Atlas is an omni world model that reads text, images, video, and spatial data inside one shared 3D context.
Text
Prompts and instructions
Images
Reference views
Video
Temporal sequences
3D Space
Camera poses and depth
What can Atlas do?
Select a capability to see how Atlas generates, reconstructs, and simulates spatial experiences.
Generate new views from reference images while specifying the camera position, angle, and motion.
Camera-Controlled Video
How Atlas works
Atlas grounds visual inputs at camera positions, builds a spatial context, and generates each new output from what came before.
Multimodal inputs
- Text and reference images
- Video as image sequences
- Camera poses and depth maps
Shared spatial context
Every visual input is tied to a position in 3D space, giving Atlas a consistent frame for camera control and scene structure.
Generated outputs
- New views and long videos
- Point clouds and 3D splats
- Images and simulations
What could Atlas make possible?
Explore where camera-aware generation and spatial reconstruction could change creative and technical workflows.
Direct shots from new camera positions
Generate controlled camera moves, reframe existing footage, and explore new perspectives for previs, visual effects, and scene design.
Atlas Updates
Be ready for Atlas
Join the waitlist. We will email you when there is something new to try on Atlas Model Hub.