Google DeepMind unveils Genie 4 world model with real-time audio
Google said Genie 4 lets users steer simulated worlds with natural language and is rolling out first to a limited group of trusted testers.
AI illustrationGoogle DeepMind on 27 October announced research on Genie 4, a world model for interactive simulation in real time. The company said the model brings a level of realism it called unprecedented and generates its own audio.
According to Google, Genie 4 draws on and extends capabilities from its Gemini family of models. Gemini Omni grounds the simulations in how things look and behave in the real world, while other Gemini models supply the reasoning that helps keep a simulation coherent and continuous.
Users can start an event and steer a scene as it runs by giving instructions in natural language, the company said, for example to move the camera, change the weather or add new characters. Each world also produces sound as it runs, including ambient noise, music and interactive dialogue.
Google described simulating the world as a crucial step toward AGI. It said such simulation can provide safe training environments for robots and agents and strengthen creative tools in industries such as filmmaking, journalism, gaming and manufacturing.
Genie 4 has begun rolling out to a limited group of trusted testers, according to the company. Google said the model is coming soon to Google AI Ultra subscribers in Project Genie and to early access partners in the Gemini API and AI Studio.