Google unveils Genie 4 AI model for richer interactive simulations

Google announced the Genie 4 world model, the latest iteration of its AI-driven interactive simulation platform. Compared with Genie 3, the new system can generate audio, produce scenes up to two minutes long, and respond to live natural-language edits, all powered by the Gemini 3.8 Flash engine. The company positions Genie 4 as a tool for pre-visualization in film, rapid game prototyping, and creating training environments for robotics.

However, it notes that extended simulations may experience changes in detail and object permanence over time. Genie 4 is initially released as a research preview for a limited group of trusted testers, with a public waitlist opened for Project Genie participants. Google plans to roll the model out to Google AI Ultra subscribers later.