Google DeepMind announced Genie 4, a new world model designed for real-time interactive simulation with unprecedented realism and native audio. The company says the model can generate music, ambient sound, and dialogue alongside visual scenes, and that users can steer events in real time with prompts such as changing the weather or introducing characters.
The release matters because simulation is becoming a practical frontier for training agents and exploring embodied AI. Google says Genie 4 is already available as a research preview for trusted testers, with access planned for Google AI Ultra subscribers in Project Genie and early-access partners in the Gemini API and AI Studio. For developers and researchers, the immediate takeaway is that world-model tooling is moving from demo territory toward usable preview access.