Google Genie 4 brings sound and interaction to virtual worlds

Google DeepMind introduced Genie 4 on October 27, 2026, as a world model for generating interactive virtual environments through natural language instructions. Unlike fixed video generators, it lets users explore scenes, introduce objects and characters, and change conditions while a simulation runs. Its capabilities include improved physics, more natural character movement, and native generation of environmental sounds, music, and dialogue.

Integration with Gemini models adds reasoning and planning to the simulations. Google describes Gemini 3.8 Flash as a guiding agent that observes generated frames, plans actions, and checks task completion. Potential applications include game prototyping, filmmaking, educational experiences, and simulated environments where robots and AI agents can practice tasks before real-world testing. These uses remain research opportunities rather than established commercial solutions.

The technology is experimental and does not replace professional game engines. Limitations include shifting visual details, imperfect physics, difficulties generating readable text, and audio quality and synchronization issues. Continuous interactions are limited to a few minutes. Access is restricted to trusted testers, with expansion planned through Project Genie for Google AI Ultra subscribers and selected partners using the Gemini API and Google AI Studio. Google has not confirmed public pricing or a general free-access offer.