DeepMind Previews Genie 4 World Model That Generates Audio in Real Time

Google DeepMind says Genie 4 is its first real-time world model to generate audio natively, producing music, ambient noise and interactive dialogue alongside the scene. Users can steer a running simulation in natural language, such as changing the weather, moving the camera or adding a new person or object. Gemini 3.8 Flash runs as a guiding agent that watches the scene as it generates, while Gemini Omni anchors appearance and behavior. The release is a research preview, gated to Google AI Ultra subscribers and early-access partners of the Gemini API and Google AI Studio. DeepMind's own write-up says visual details and object permanence can drift over longer runs or after several prompt edits. Resolution, frame rate, inference cost and maximum session length were not disclosed. For builders, real-time steerable simulation is a plausible training ground for embodied agents and interactive products. Drift and unknown cost are the open questions, so teams should test session length and pricing before designing a product around it.