Google DeepMind has unveiled Genie 4, a model that generates and runs interactive virtual environments as the user watches them unfold. It is not a chatbot or an image generator: the result is a world you can explore, one that responds to your commands while it is being created.
What sets it apart is control. Users describe a scene in natural language, with no technical commands, and can intervene mid-simulation to move the camera, change the weather or introduce new characters. Audio is generated alongside the visuals, with ambient sound, music and dialogue that adapt to what happens on screen rather than playing back a pre-recorded track.
Image: genie4.space — https://genie4.space/
Behind the model is not a single system. Genie 4 draws on several members of the Gemini family: Gemini Omni anchors the simulations to real-world appearance and behaviour, while other Gemini models handle the reasoning needed to keep scenes coherent over time. It is an orchestration of models working together.
Google frames the technology as a step toward artificial general intelligence. Such simulations could train robots and AI agents safely, without risks in the physical world. Applications cited range from film and journalism to gaming and manufacturing.
Image: StartupHub AI — https://www.startuphub.ai/ai-news/artificial-intelligence/2026/google-deepmind-unveils-gemma-4-ai-models
For now access is restricted to a small group of testers. Wider availability is announced for Google AI Ultra subscribers through Project Genie, and for early-access partners via the Gemini API and AI Studio. Demos and technical details are published on Google's blog.
The launch fits into a broader push by Google into generative worlds, starting with tools that let people build video games without writing code. The open question is timing: how much of this will reach everyday products, and how soon.
No comments yet. Be the first!