World Labs Unveils Atlas World Model
aiinsider ·
World Labs introduced Atlas, a new 'spatial intelligence' world model designed to generate, reconstruct, and simulate 3D environments. It supports Real-to-Sim workflows, reconstructing physical spaces from a small number of images or video frames and generating the RGB images and depth data a robot would encounter during navigation. For manipulation tasks, Atlas can recreate interactions with rigid, articulated, and deformable objects, allowing developers to vary objects, positions, robot motions, lighting, and backgrounds for training and testing scenarios. The model is described as a multimodal autoregressive diffusion transformer that combines text, images, video, and 3D information within a shared spatial context, grounding images and depth maps to specific camera positions. Atlas can also produce explicit 3D outputs such as point clouds and 3D Gaussian splats from images or video. Beyond robotics, it supports camera-controlled video generation, 3D reconstruction, and image generation. The model is entering early access with select partners and will power future versions of World Labs' Marble world-generation product.
aiinsider