Google DeepMind launches Genie 4, a world model with live editing

Google DeepMind has launched Genie 4, its latest world model, now in research preview. It can create the most realistic interactive virtual worlds yet, joining a race with rivals like Runway.

The model supports live editing. Users can change weather, camera angles or add people and objects with prompts, and it generates audio synced to the scene. A scene can stay consistent for up to two minutes, twice as long as Genie 3, with better physics and more lifelike characters.

Genie 4 works more closely with Gemini, which plans steps and checks frames while Genie 4 creates the visuals. It targets gaming, interactive learning and robotics, and trusted testers found it useful for film prototyping.

The model is now open to trusted testers and will roll out to Google AI Ultra subscribers and partners in coming weeks. The team admits it is still early: details drift in long simulations, and audio and text rendering need work.