Google DeepMind unveils Genie 4, a real-time world model with native audio

On October 27, 2026, Google DeepMind announced Genie 4, its new world model for real-time interactive simulation. Users can change a scene while it runs by typing instructions, such as moving the camera, changing the weather, or adding people and objects. Each world now generates its own audio, including ambient sound, music, and characters that hold spoken dialogue in response to the user.

Google says Genie 4 builds on the Gemini family: Gemini Omni anchors how things look and behave, and other Gemini models handle reasoning to keep scenes coherent. In one demo, Gemini 3.8 Flash acts as a guiding agent that plans steps and adjusts when the user steps in. The post gives no resolution, frame rate, session length, or benchmark numbers.

Access is limited for now. Genie 4 is a research preview for trusted testers, and Google says it is coming soon to Google AI Ultra subscribers in Project Genie and to early-access partners in the Gemini API and Google AI Studio. No pricing or regions were announced. Google says generated video and audio will carry SynthID watermarks, and it lists object permanence over long sessions, legible text, and audio accuracy as known limits.