World Labs Atlas — A Single Model for 3D Generation, Reconstruction, and Simulation
World Labs (founded by Fei-Fei Li and funded with $1B) has published Atlas, a new model for creating/reconstructing 3D scenes from 1–3 images that outperforms specialty models in 75–94% of user trials.
World Labs, founded by Fei-Fei Li (creator of ImageNet) and funded with $1 billion, has announced Atlas, a new model for creating/reconstructing 3D scenes from 1–3 images that outperforms specialty models in 75–94% of user trials. The model can generate 1440p videos up to a minute long with full camera control and is better than specialty models like MiniMax H3, Gemini Omni Flash, and Seedance 2.5.
Up until this point, we had separate tools for video synthesis, 3D reconstruction, and robot simulation which required separate experts. With Atlas, we have a unified model that can do all of these things and outputs directly to point cloud and 3D Gaussian splat formats. Now, we can raise the bar for games, movies, and robot training by rapidly creating three dimensional worlds based on real life.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles