Fei-Fei Li's World Labs unveils Atlas: an omni world model with 1-min 1440p video and 3D reconstruction
theworldlabs · x · 2026-09-09
World Labs introduced Atlas, a next-generation omni world model pretrained from scratch to natively operate on text, images, video, and 3D. It's a multimodal autoregressive diffusion transformer that merges all inputs into a shared spatial context, staying 3D-consistent and scaling with compute.
- Camera-controlled generation: pixel-perfect camera control from one or more images, up to 1 minute of 1440p video
- Spatial reconstruction: from 1 to dozens of images, outperforming specialized 3D reconstruction SOTA with novel-view frames and explicit 3D outputs
- Space-time simulation: video reframing and Real-to-Sim workflows for robotics
- Image generation: text-to-image and 360 panoramas with complex prompts and text rendering
Atlas will power future versions of Marble; early access is now open.
More from Embodied
- Bedrock Robotics raises $600M as its fully autonomous excavators do paid work on live job sites — Sethwinterroth · 2026-10-27
- AW-0 hits 80.8% on RoboTwin tasks, beating strongest baseline by 16 points — trevordarrell · 2026-10-27
- AW-0's code stream fixes world model prediction: picks mango where baseline grabs apple — trevordarrell · 2026-10-27
- AW-0 runs at 230ms on a single RTX 5090, 2.4x faster than autoregressive code planning — trevordarrell · 2026-10-27
- AW-0 debuts: one generative world action model plans in code, predicts, and controls robots — trevordarrell · 2026-10-27
- Skyfall AI's SKY-001 world model turns sketches into CAD builds in 3.9s, vs 199s for GPT-6 Astra — rohanpaul_ai · 2026-10-27