Google DeepMind's Genie 4 builds a world you can explore in real time, and for the first time it makes its own sound and lets the people in it talk.

Google DeepMind unveiled Genie 4 on Tuesday, calling it "our most realistic and interactive world model yet" (Google DeepMind on X).

A world model generates an interactive scene as you move through it, a bit like a video game that's made up on the fly instead of built in advance. DeepMind says Genie 4 is the first real-time world model to generate audio natively, including music, ambient noise and dialogue, so people in a scene can react and speak. You can also change the scene with plain words while you're in it, such as changing the weather, moving the camera or adding a new person or object (Google DeepMind).

Scenes stay consistent for up to about two minutes, double Genie 3's limit (TechCrunch, The Verge). DeepMind says early testers have used it for film previsualization and for game prototypes made "before writing a single line of engine code." It also admits a limit: visual details and objects can still "shift or drift over time." For now it's a research preview for trusted testers. Sundar Pichai says it's "coming soon to Google AI Ultra subscribers in Project Genie and early access partners via the Gemini API and AI Studio" (Sundar Pichai on X), and TechCrunch reports that rollout is planned "in the coming weeks."

When you'll feel it: not today unless you're a tester. If you have Google AI Ultra, watch for Genie 4 in Project Genie over the next few weeks. Developers will get it through the Gemini API and AI Studio later, starting with early access partners, and Google hasn't announced API pricing yet.