Google DeepMind announces Genie 4, a real-time world model that generates its own sound, as a research preview

Google DeepMind announced Genie 4 on 27 October. It generates interactive environments in real time, which users steer with keyboard controls and text prompts while the scene runs. DeepMind says: “Genie 4 is our first real-time world model to generate audio natively”, covering ambient sound, music and dialogue. It draws on Gemini models to ground what scenes look like and keep them coherent.

It is a research preview for trusted testers. DeepMind says it is coming to “Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio”, with a waitlist and no date for wider release.

The post gives no figures for resolution, frame rate or session length, and no benchmark results. It also admits limits: “visual details and object permanence can still shift or drift over time—especially as a simulation runs longer”.

World models matter beyond games. Labs use them to train agents and robots in simulated places before they act in real ones, and DeepMind's post shows a simulated Apptronik humanoid among its examples. For now Genie 4 is a preview without published measurements, so what it can do outside DeepMind's own examples is not yet visible from here.

• Confirmed Genie 4 is announced as a research preview and is DeepMind's first real-time world model to generate audio natively. Google DeepMind
• Confirmed Access goes to trusted testers first, then Google AI Ultra subscribers in Project Genie and early-access partners of the Gemini API and Google AI Studio. Google DeepMind