Google launches Genie 4 world model for realistic interactive simulations

Google announced the Genie 4 world model, a system capable of generating more realistic environments than previous iterations. Users can modify an experience in real time by providing a natural language prompt.
The model also produces audio and utilizes Gemini 3.8 Flash to power its simulations. Genie 4 creates experiences lasting approximately 2 minutes, an increase from the 1-minute duration of the Genie 3 model.
Technical limitations remain, as details and object permanence can shift as simulations run for longer periods. Filmmakers might use the technology for pre-visualization, while game designers could use it to prototype ideas.
Robotics engineers could also employ the model to build training scenarios. A research preview of Genie 4 is launching first to trusted testers.
Google stated the model will arrive soon for Google AI Ultra subscribers. A waitlist for Project Genie has been opened by the company.








































