World Labs unveils Atlas, a spatial world model that generates 1440p videos up to one minute
- World Labs introduced Atlas, an omni world model pretrained from scratch to natively process text, images, video, and 3D in a shared spatial context.
- Atlas generates camera-controlled videos from one to six reference images, with manually designed camera paths, at up to one minute in 1440p.
- The model takes precise camera geometry as a native input, generating views at specified positions and angles while maintaining the input scene's content and geometry.
- For spatial reconstruction, Atlas accepts one to dozens of images, produces novel-view frames and explicit 3D outputs, and World Labs claims it surpasses specialized 3D reconstruction models.
- Atlas also models space and time from video for reframing and robotics-oriented Real-to-Sim workflows; World Labs plans to use it in future Marble versions and is taking early-access requests.