Google DeepMind has announced Genie 4, a world model that generates an interactive scene and simulates how it responds as a user moves through it. Users steer with keyboard controls and can type prompts mid-session to change the weather, move the camera or add people and objects. DeepMind says it is its first real-time world model to generate audio natively, including music, ambient sound and dialogue, alongside the visuals.
Announced on 27 October, Genie 4 is a research preview for trusted testers. Google says it is coming soon to Google AI Ultra subscribers through Project Genie and to early access partners in the Gemini API and Google AI Studio, without giving dates. A Gemini model acts as a guide in the loop, adapting prompts to the scene. All generated video and audio carries Google’s SynthID watermark.
Why it matters: world models are one route to training robots and AI agents in simulation before they act in the real world, and Google lists that among Genie 4’s uses. DeepMind is frank about the limits: scenes stay coherent only over short sessions, and visual details and object permanence can drift over time.