Google's Genie 4 Builds Living Worlds With Sound — But You Can't Build On It Yet

Google DeepMind introduced Genie 4 on October 27, 2026, and the most significant development is that the simulation now comes with sound. For the first time in DeepMind's real-time world-model line, the simulation and audio are generated together, including ambient noise, music, and interactive dialogue. This advancement marks a significant step forward in creating more convincing and realistic virtual environments.

However, the tool is currently only available through a waitlist, not yet accessible for immediate use. The capabilities of Genie 4 are described by DeepMind as enhanced photorealism, more fluid movement, and better physics, with a particular focus on simulated people. Users can interact with the generated world by using natural language prompts to change the scene, such as moving the camera or introducing new elements.

The generation process is facilitated by two Gemini pieces: Gemini Omni, which provides grounding in the real world, and Gemini 3.8 Flash, which ensures coherence throughout the scene. Despite the promising advancements, Genie 4 is still in the research preview phase, available only to trusted testers and select partners. Currently, there is no public API endpoint or pricing information, and developers should approach the tool with curiosity and patience, understanding that it is best suited for training environments, previsualisation, and game prototyping rather than immediate production use.