Google DeepMind has announced Genie 4, its newest world model for real-time interactive simulation. Unlike a traditional video generator, Genie 4 builds an explorable world frame by frame as the user moves through it, and it can now generate the soundtrack alongside the scene.
The headline change is native audio. Music, ambient sound, and interactive dialogue are produced alongside the visuals, so a lakeside scene carries the sound of water and characters can answer in their own voices. Steering has also improved: users can change the weather, move the camera, or introduce new people and objects mid-simulation using plain language, while Gemini adapts the prompt to fit what is already on screen.
Behind the scenes, Gemini 3.8 Flash acts as a guiding agent, breaking goals into steps and checking frames as the world evolves. DeepMind says Genie 4 holds a coherent scene for about two minutes, roughly double its predecessor, with more natural human movement and better physics for collisions and aerodynamics.
Beyond games, filmmakers can run multiple live takes of a scene, developers can prototype an explorable idea in seconds, and robotics teams can test scenarios in simulation first. Tools like Runway and Synthesia show how fast AI video is maturing, but Genie 4 is aimed at interactive worlds rather than finished clips. Our Runway vs Pika comparison explains how today’s leading video generators differ.
Limits remain: visual details and object permanence can still drift in longer sessions or after many edits, and complex physics, legible text, and spatial accuracy are still research problems. All Genie 4 video and audio carries SynthID watermarking.
Genie 4 is in research preview for trusted testers, with wider access coming to Google AI Ultra subscribers in Project Genie and early access partners in the Gemini API and AI Studio.
Source: Google DeepMind
