Google DeepMind unveils Genie 4, a world model that talks back and can be steered live

Google DeepMind announced Genie 4 on October 27, a new real-time interactive world model that generates explorable, photorealistic worlds from a text prompt or an image. It is DeepMind's first real-time world model with native audio: music, ambient sound and interactive dialogue are generated simultaneously with the scene, rather than dubbed on afterwards, and characters converse and react to the user's inputs or events in the world in real time.

A world you can rewrite while it runs

Users can steer a running simulation with plain-language prompts — changing the weather, moving the camera, or dropping in a new person or object — without restarting the scene. Gemini Omni anchors how things look and behave in the world, while Gemini 3.8 Flash acts as the guiding agent inside the generation loop, breaking down the user's goals and adjusting the world state frame by frame to keep it coherent. DeepMind says the physics simulation has also improved, covering aerodynamics and collisions, with the biggest visible jump in people: characters that move naturally, react to their surroundings and can be steered while moving.

Two minutes of coherence, still a research preview

Genie 4 can hold a scene together for about two minutes, double the roughly one-minute limit of its predecessor Genie 3 — though DeepMind is candid that visual details and object permanence can still shift or drift over longer runs, after multiple complex prompt edits, or while navigating. For now Genie 4 is a research preview for trusted testers, rolling out first to Google AI Ultra subscribers through Project Genie and to early-access partners of the Gemini API and Google AI Studio, with a waitlist open for others; everything it generates carries SynthID, DeepMind's imperceptible watermark.