Google DeepMind’s Genie 4: interactive worlds with real-time sound
The model simulates a world in response to your actions and prompts, and now creates music, sound and dialogue. It’s a preview for trusted testers.
On October 27, Google DeepMind introduced Genie 4, a world model that simulates what happens next: how the environment responds and how people move and speak. It isn’t finished video, but a scene you steer while it’s being generated.
What’s new
- Text control: change the weather, move the camera, or add a person or object. Gemini adapts these prompts to fit the scene.
- Movement: navigate with standard keyboard controls for movement and camera.
- Sound: Google says it’s its first real-time world model to generate audio natively: music, ambient sound and dialogue that responds to your actions.
- Physics: simulation of aerodynamics and collisions, plus more natural character movement.
Testers are exploring Genie 4 for previsualization, game and scenario prototyping, and turning archival photos into simulations.
Who gets it
It’s a research preview for trusted testers. It’s coming next to Google AI Ultra subscribers in Project Genie, which has a waitlist, and to early-access partners in the Gemini API and AI Studio.
Limitations
Visual details and object permanence can drift, especially over longer sessions or after many prompt edits. Google also lists complex physics, audio accuracy, legible text and spatial accuracy as unfinished. All generated video and audio carries a SynthID watermark.
Prepared with AI from public sources.