Google DeepMind Genie 4: playable AI worlds now come with sound and live text control

Google DeepMind has announced Genie 4, the next version of its "world model": an AI system that generates an explorable 3D scene you can walk through and steer in real time. As someone who spends too much time with games, I think this one is worth watching more closely than the usual chatbot update.

What Genie 4 adds

DeepMind lists four main upgrades:

• Real-time steering with plain language. While you move around with keyboard controls, you can type changes such as new weather, a different camera angle, or an extra character, and the scene adapts.
• Physics and people. DeepMind says collisions, aerodynamics, and especially human movement look much more natural than before.
• Native audio. This is its first real-time world model that generates sound along with the image: ambient noise, music, and even dialogue from characters who react to what you do.
• Gemini under the hood. Gemini Omni keeps scenes grounded in real-world appearance, and other Gemini models help keep the simulation consistent.

Who can use it

For now it's a research preview for trusted testers only. Google says it will come to Google AI Ultra subscribers through Project Genie, and to early access partners in the Gemini API and Google AI Studio, with a waitlist already open. There's no public pricing yet.

Why games fans should care

Google pitches this for training robots and AI agents, but it also names game prototyping and filmmaking as uses. Being able to rough out a level, change the weather mid-walk, and hear the space react is a genuinely new kind of tool. DeepMind also admits it's still early research with clear limitations. I see it as a sketchpad for ideas, not something that will replace hand-built game worlds or animation any time soon. I'm curious whether stylized looks, not just photorealism, will be possible once it opens up.

Source: Google DeepMind blog.