San Francisco (Web Desk): Google DeepMind has introduced its new world model, Genie 4, which uses artificial intelligence to create realistic virtual worlds where users can move around as they would in a video game.
One of the model’s key features is its ability to generate audio alongside visuals. Sounds such as rustling wind, music, ambient noise and character dialogue are produced as the virtual world is created.
According to DeepMind, Genie 4 is its first real-time world model capable of generating audio live. Environments do not need to be pre-built or programmed. Instead, users simply describe their desired world through text, and the model continuously generates each frame in response to their movements or camera rotations.
Genie 4 has been developed using the capabilities of Google’s Gemini models. DeepMind says Gemini Omni gives virtual worlds more realistic appearances and behaviours, while other Gemini models provide the reasoning needed to maintain continuity and consistency within a scene.
Genie 4 has initially been released to selected testers as a research preview. The company says it will become available in the coming days to Google AI Ultra subscribers through Project Genie, as well as to early partners using the Gemini API and Google AI Studio. A waitlist has also been opened for interested users.