Fei-Fei Li’s $1.2B startup World Labs released Atlas on September 1. It’s an “omni world model”: feed it one photo (or text, or video) and it generates up to a minute of 1440p video with fully controllable camera trajectories — or reconstructs the same scene as geometry-consistent 3D assets. Video generation and 3D reconstruction in a single model is the pitch. Nobody else ships both.
What it’s actually for
Not filmmaking. Robotics. Scan a room with your phone, Atlas rebuilds it as a simulation robots can train in. Sora and Veo make pretty pixels; Atlas claims its pixels obey geometry. World Labs reports 81–93% user preference over MiniMax H3, Gemini Omni Flash, and FLUX 3 on camera-controlled generation.
The catch
Every number is self-reported. No paper, no model card, no pricing, and the “select partners” in early access are anonymous — entry is a request form. For a company positioning itself as the lab of spatial intelligence, “trust us” is a strange launch. The demos are stunning. Right now, that’s the entire body of evidence.
You Might Also Like
- World Labs Marble fei fei li Just Raised 1 Billion to Build 3d Worlds From a Single Photo
- Mistrals Next Open Weight Model Opens Early Access in July
- Starnus Just hit 1 on Product Hunt and Yeah its Worth the Hype
- Lovon Just Topped Product Hunt on Valentines day and its not a Dating app
- Zenmux Just hit 1 on Product Hunt Heres why Everyones Paying Attention
Top AI Product 
Leave a comment