Google DeepMind announced Genie 4 on October 27, 2026, introducing a real-time world model for interactive simulation that is starting as a research preview for trusted testers.
In the Google DeepMind announcement, the company says Genie 4 generates audio alongside visuals, including ambient sound, music, and interactive dialogue inside a scene.
DeepMind is also pitching live control as a core change. Users can steer a running simulation with natural-language edits, such as changing the weather, moving the camera, or adding people and objects, instead of regenerating a scene from scratch.
The company says Genie 4 also improves realism for physics and human motion, including aerodynamics and collisions. That points to a more useful testbed for interactive agents and robotics-style evaluation than video generation alone.
Availability is still narrow. Google says Genie 4 is rolling out first to trusted testers, with broader access planned later for Google AI Ultra users in Project Genie and selected early partners using the Gemini API and Google AI Studio.