❯ Google DeepMind introduces Genie 4 with native audio and real-time world control

Explorable worlds gain synchronized soundGoogle DeepMind introduced Genie 4 on October 27, a world model that generates explorable scenes and simulates what follows users’ actions. Native audio adds environmental sounds, music and character dialogue alongside visuals. Access currently takes the form of a research preview for trusted testers.

Prompts change a running sceneUsers can navigate and control the camera with a keyboard or type instructions to change weather and introduce people or objects. Characters can adapt their dialogue to inputs and events. Early tests include film previsualization and interactive concepts, allowing creators to try camera angles, lighting, dialogue and action within the same evolving scene.

Gemini observes and plansGoogle says Gemini Omni contributes realistic appearance and behavior, while Gemini 3.8 Flash observes frames, breaks goals into steps, checks progress and replans after intervention. It also demonstrates Gemini 4 Argon and Genie 4 simulating an Apollo humanoid playing checkers. The combination explores environments for agents; the example does not establish equivalent capabilities in a real robot.

Broader access remains forthcomingGoogle’s announcement says access will follow for Google AI Ultra subscribers in Project Genie and early Gemini API and Google AI Studio partners. General availability has not been announced. DeepMind notes that longer sessions, complex edits and navigation can cause visual and object drift. Physics, spatial accuracy and long-term consistency remain research challenges.

▮ SIGNALWorld models are connecting scene generation, sound and goal planning within one interactive loop, creating environments in which creators and agents can experiment repeatedly.