World Labs unveils Atlas, a single AI model that generates, reconstructs, and simulates 3D worlds from just a few photos
decoder ·
World Labs, co-founded by AI researcher Fei-Fei Li, has unveiled Atlas, an AI model that can generate, reconstruct, and simulate 3D worlds from just a few photos. The company describes Atlas as an "omni-model" trained on text, images, video, and 3D data, with every input anchored to a specific position in 3D space. Atlas can generate camera-controlled video up to one minute at 1440p, reconstruct real scenes from one to over a hundred input images, and output explicit 3D data such as point clouds and 3D Gaussian splats. It also functions as a simulator for robotics, turning a few photos of a room into varied training data for simulated sensors and control models. According to World Labs, human evaluators preferred Atlas over specialized models in most comparisons, and it achieved lower reconstruction error than competing systems. The model will power future versions of the company's Marble product and is currently available through an early-access program for select partners. World Labs was founded in 2024 and has raised funding from investors including Andreessen Horowitz, Nvidia, and AMD.
decoder