Google introduces Genie 4, its interactive virtual world
Google introduces Genie 4, an interactive simulation model that creates worlds with images, sound and dialogue generated instantly. Its scenes can be modified with natural-language instructions, although initial access is limited to testers and partners.
Google has introduced Genie 4, a model capable of creating real-time interactive simulations with realistic images, instantly generated sound and scenes you can control using natural language.
Worlds that respond to your instructions
Rather than simply generating an image or video, Genie 4 builds an environment that maintains some continuity as you interact with it. You can start a scene and modify it on the fly with instructions such as moving the camera, changing the weather or adding new characters.
The system combines capabilities from the Gemini family. Gemini Omni provides a representation closer to the appearance and behavior of the real world, while other Gemini models handle the reasoning needed to keep the simulation coherent.
The practical difference matters: you are not only describing how a scene should look. You can intervene in it and watch how it responds. For example, you could ask it to start raining, change the viewpoint and add a character without having to rebuild the entire environment from scratch.
Sound generated inside the simulation
Genie 4 also creates audio natively, meaning as part of the simulation itself. It can produce:
- Environmental sounds, such as rain, traffic or footsteps.
- Music adapted to what is happening in the scene.
- Interactive dialogue for characters.
This makes it possible to create more complete experiences for video games, audiovisual prototypes or training environments. Audio is not added only at the end: it responds to changes that occur as you interact.
What it can be used for
Google considers simulating the world an important step toward more capable artificial intelligence systems. A controllable virtual environment could be used to train robots and AI agents without directly exposing them to real-world risks.
It could also expand the tools available across several sectors:
- Film and audiovisual production, to explore scenes and camera movements.
- Video games, to create worlds that react to players' actions.
- Journalism, to explain events through interactive environments.
- Manufacturing, to test processes before taking them to a real plant.
This does not mean Genie 4 is available to everyone or that it can simulate any situation without errors. Google presents it as research and has already begun testing it with a limited group of trusted testers.
Access will arrive soon for Google AI Ultra subscribers through Project Genie. An early access phase is also being prepared for some partners through the Gemini API and AI Studio.
What matters to watch now is whether Genie 4 can keep scenes consistent during long interactions and whether its simulations are reliable enough to train agents and robots. The central idea is already clear: AI is starting to move from generating isolated content to building spaces you can interact with in real time.