Genie 4: Google DeepMind's Real-Time World Model

The short answer

Genie 4 is Google DeepMind's next step in "world models" — AI systems that simulate how a scene looks, moves, and sounds. You set an event in motion and the model predicts what happens next, then lets you steer it in real time with natural language while native audio (ambient sound, music, and even spoken dialogue) is generated alongside the world^[1][2].

What Genie 4 is

Genie 4 is a natively multimodal world model. That means it doesn't just render pixels — it simultaneously generates the audio that belongs to each scene, from wind in trees and waves at a beach to music and interactive dialogue^[2]. Google DeepMind frames world simulation as a step toward AGI, useful both for building safe training environments for robots and agents and for creative tools in filmmaking, journalism, gaming, and manufacturing^[1].

Built on the Gemini family

Genie 4 incorporates and builds on capabilities from the Gemini family. Gemini Omni anchors the simulations in real-world appearance and behavior, while other Gemini models supply the complex reasoning needed to keep a simulation coherent and continuous as it runs^[1][2].

Steering scenes in real time

Instead of regenerating from scratch, you can modify a running simulation in natural language — change the weather, move the camera, or introduce a new person or object — and Gemini adapts your prompt to match the surrounding scene. Between your inputs, Gemini also monitors and steers the scene so the world keeps progressing cohesively^[2].

Native audio and interactive dialogue

Genie 4 is Google DeepMind's first real-time world model to generate audio natively. Every world carries the ambient sound of its setting, and characters can converse and react to your inputs or events in the world in real time^[2]. Google DeepMind says all accompanying sound in its Genie 4 demos is produced directly by the model^[2].

A new level of realism and physics

Genie 4 is built on an advanced physical understanding of the world, so it simulates scenarios like aerodynamics and collisions. Google DeepMind calls the biggest leap how it simulates people — characters that move naturally, respond organically to their environment, and can be steered interactively in real time^[2].

Rollout and availability

Genie 4 is launching as a research preview for trusted testers first. Google plans to open it next to Google AI Ultra subscribers in Project Genie and to early-access partners of the Gemini API and Google AI Studio, with a waitlist available for Project Genie^[1][2].

Should you care?

Verdict: Genie 4 is a research preview, not something you can buy today. For most readers it's a signal of where Google's video and agent tooling is heading — toward real-time, steerable, audio-capable simulation. If you build with the Gemini API or hold a Google AI Ultra subscription, it's worth joining the waitlist; everyone else can watch the demos and wait for a consumer release^[1].

Frequently asked questions

Is Genie 4 available now?

Only as a research preview for trusted testers. Google plans to open it to Google AI Ultra subscribers in Project Genie and Gemini API early-access partners next^[1].

What makes Genie 4 different from video generators?

It's interactive: you steer a live simulation with natural language in real time, and audio — including dialogue — is generated natively along with the world^[2].

Does Genie 4 run on the Gemini models?

Yes. It incorporates Gemini Omni for real-world appearance and behavior, plus other Gemini models for the reasoning that keeps simulations coherent^[1].