Genie 4 simulates worlds in real time, and it is not a game yet

Google DeepMind announced Genie 4 on October 27, 2026. It is a world model: instead of handing over a finished video, it simulates a scene that keeps going in real time. You can change the weather, move the camera, or add a person or object with natural language while the scene continues. Sound is generated with the picture. Ambience, music, and dialogue come from the model itself. In the published examples, navigation and the camera also use a keyboard.

For people making games, DeepMind describes what testers are trying: going from an idea to an explorable prototype, and testing look, materials, and mechanics, before writing engine code. That does not make Genie 4 a game, a replacement for a game engine, or a tool the public can open. The research preview is with trusted testers. DeepMind says that, with no date set, the model later reaches Google AI Ultra subscribers in Project Genie and early-access partners on the Gemini API and Google AI Studio. There is no price and no window for Brazil.

The official post treats the model as experimental. A scene holds together better in short sessions. Visual detail and object permanence can still shift when a simulation runs longer, after several prompts, or as you move through the scene. More complex physics, audio accuracy, readable text, spatial accuracy, and long-term consistency remain research. Generated video and audio carry the SynthID watermark.

Our take: a world model is a simulation generated on the fly, useful for picturing a scene. For someone playing games today, the catalog does not change. When the preview opens to more people, the useful question is how long the scene stays coherent, not only whether the first frame looks striking.