Games · Model Watch
Genie 4 lets creators direct a world while it is still happening
Google DeepMind's new world model generates interactive environments, sounds and characters that respond as people explore—and lets them change the scene along the way.
Reporting updated Nov 1, 2026
THE TAKEAWAY
Google DeepMind's Genie 4 introduces a striking idea: a generated world doesn't have to stop at the moment you press Enter. You can move through it, hear its surroundings and ask it to change while the simulation continues. It's still an early research preview, but its approach could give game developers and filmmakers a very different way to explore ideas.
WHAT HAPPENED
Imagine entering a street that doesn't exist yet. You move the camera, follow a character around a corner, and decide the scene would work better in a rainstorm.
Instead of starting again with a new prompt, you ask for rain. The world changes around you.
That's the creative promise behind Genie 4, introduced by Google DeepMind on October 27.
Unlike a conventional AI video generator, Genie 4 produces an interactive simulation. Keyboard controls let a person navigate while text instructions can introduce objects, people, weather changes and camera directions.
The world generates audio alongside its imagery, including environmental sound, music and dialogue. Google says the system draws on Gemini-family capabilities to interpret instructions and help the simulated environment respond coherently.
The demonstrations are impressive because they move beyond presenting a beautiful scene. They invite someone to interact with it.
For a game developer, that could mean rapidly exploring an environment, testing a mood or staging a moment before building production assets. For a filmmaker, it could become a way to investigate camera placement, atmosphere and character movement inside an evolving scene.
Those are possibilities, not yet proven production workflows.
DeepMind is initially providing access to trusted research testers. Broader access is planned through Project Genie for Google AI Ultra subscribers, along with developer access through Google AI Studio and the Gemini API.
WHY IT MATTERS
Games are built around what happens when players do something unexpected. Films are built around carefully chosen moments, performances and camera positions.
Genie 4 starts to bring those two ways of thinking closer together.
A creator can experiment with a scene by moving through it rather than only describing it. They can discover an interesting angle, change the atmosphere and see how the environment responds.
The limitation is continuity. DeepMind acknowledges that visual details and object permanence can drift. Complex physics, spatial accuracy, readable text and long sessions remain challenging.
That makes Genie 4 more useful to think about as an experimental directing and world-exploration tool than as a finished game engine.
How it works
Explore
Navigate generated environments with live controls.
Direct
Use language to change weather, characters, objects and scene conditions.
Listen
Hear audio generated alongside the evolving world.
Experiment
Test creative possibilities while recognizing that continuity and physics still need work.
DeepMind describes the system as a research preview and explicitly acknowledges that longer sessions can lose consistency.
Source excerpt
VEYR ANGLE
The most exciting question isn't whether AI can generate a beautiful world. We've seen plenty of those.
It's what happens when a creator can enter that world, change their mind and keep directing.
