Google announces Genie 4 with audio and real-time world editing

Google announced Genie 4 on October 27, describing it as a new world model for real-time interactive simulation. In its announcement on The Keyword, the company highlighted greater realism and native audio.

The model follows Genie 3, which Google DeepMind revealed last year. Ars Technica reported that world models have since become a major focus for AI labs. The central idea is an AI-generated environment that users can interact with, rather than simply watch.

The October 27 announcement put more lifelike environments at the centre of the update. Ars Technica described promises of more realistic physics and humans. Those are claims about the model's capabilities, not independent findings established by the supplied coverage.

Audio is another addition. Digital Trends reported that Genie 4 supports sound and interactive dialogue, bringing conversations into the generated environments alongside improvements to physics.

Users can also change a simulation while it runs. According to TechCrunch, Genie 4 accepts text prompts mid-simulation, giving users a way to steer the experience without starting over.

The interaction window has grown, too. 3DNews reported that the virtual worlds can now support interaction for up to two minutes.

Genie 4 is available in a research preview for trusted testers, TechCrunch reported. That limits who can try the announced features. The supplied coverage does not establish a date for wider public access or provide an independent assessment of how reliably the simulations work.

The immediate questions are whether Google expands access beyond trusted testers and how the model performs under broader testing. Its realism, dialogue and ability to respond to changes during a simulation remain the main capabilities to assess against the announcement.