Google DeepMind's Genie 4 Generates Interactive Worlds You Can Edit With Words

The new world model produces real-time, up-to-two-minute interactive scenes with visuals, ambient audio, and dialogue — and lets users change elements with natural language.

Google DeepMind introduced Genie 4, a real-time world simulation model that pushes past content generation toward interactive, navigable environments. According to @FoundOnArdy, the model generates interactive scenes complete with visuals, ambient audio, and dialogue, and allows users to change elements within a scene using natural language. That last capability — editing a living world by describing the change you want — is the part that separates a world model from a video generator.

Unlock the full briefing

Get every story in today's briefing, the full archive, and the daily AI intelligence brief.

All stories today

Full archive

Daily brief

Cancel anytime. Payments powered by Stripe.