Fei-Fei Li's WorldLabs Unveils Atlas Multimodal World Model
机器之心 · wechat · 2026-09-02
WorldLabs, founded by Fei-Fei Li, released Atlas, a new 'omnimodel' that natively processes text, images, video, and 3D data. Its core innovation is 'spatial context,' which treats camera poses as a native input, enabling pixel-perfect controlled camera generation and spatial reconstruction from sparse views. Atlas uses a multi-modal autoregressive diffusion Transformer architecture. Benchmarks show it outperforms models like MiniMax and Gemini in camera-controlled generation tasks.
Related event: World Labs Launches Atlas, an Omni World Model for Spatial Intelligence(64 posts)→
More from Multimodal
- Higgsfield playbook: scene templates, paid creator network, $20-to-enterprise funnel — sujingshen · 2026-10-18
- Flux 3 Text-to-Image Can't Make Realistic Faces at All, Reddit User Reports — EstablishmentFun6629 · 2026-10-18
- MiniMax H3 Ref-to-Video: 15s Shot from 3 Images + Vocals, 1088p Latent Refinement — michael_goldenhour · 2026-10-18
- Dev builds fully local AI tool generating HD 3D textures with Qwen, may sell for $5 — TinfoilTricorn · 2026-10-18
- 14-Agent AI Film Pipeline on One RTX 5090: Fixing Cut Defects Between Clips — developervkmp · 2026-10-18
- 4MP Text-to-Image Comparison: Qwen Image 2.1 Turbo vs MOSS vs Krea 2 vs Flux 3 Klein — FORNAX_460 · 2026-10-18