Report
Odyssey-3 reportedly builds interactive worlds from prompts in real time
A post says users can walk, jump and move the camera, with the generated environment responding as they move.
TLDR
A post describes Odyssey-3 as a world model that builds an environment from a prompt and generates it in real time as users move through it. It gives examples including ripples when someone steps into water, a car rolling through a crosswalk and a beaver wading out of a lake. The post says Odyssey is adapting the same model for robot arms, humanoids and driving, and that a research preview was open on October 8, 2026.
Combined views
15.4K
6 Sources, first seen ago
