Atlas by World Labs

Omni world model generating camera-controlled 1440p video

Visit website →

About

Atlas is World Labs' omni world model for spatial intelligence. It takes text, images, video, and 3D as input and generates camera-controlled 1440p video up to a minute long. Reconstruct real scenes from a few photos, generate entirely new environments from text, and simulate space-time for robotics applications.

The key capability: Atlas understands geometry, not just pixels. Given two phone videos of a scene, it generates the parts nobody filmed, freezes time, and lets you reframe from angles that were never shot. Reconstruct a childhood home from three Google Street View images, or build a film sequence from references nobody captured.

Atlas is the model that will power Marble, World Labs' consumer product for creating and navigating 3D worlds. Early access is open for developers and researchers. Founded by Fei-Fei Li, the tool targets creators, game developers, architects, and robotics teams.

AI-assisted draft, human-reviewed before publishing — see how we choose & review tools.

Why we picked it

Atlas earns a spot beyond typical video generators for understanding geometry, not just pixels — reconstructing real scenes from a few photos and generating camera-controlled 1440p video, with an eye toward robotics simulation, not just content.