Google DeepMind · 2026-01-29 · major
Project Genie — Google DeepMind's Interactive AI World Generator
Google opens access to Project Genie, powered by Genie 3 world model. Type a text prompt and navigate a generated 3D world in real time at 24fps and 720p. Available to AI Ultra subscribers in the US.
Type what you want to see and walk around inside it — a text-to-explorable-world model running in real time.
Key specs
| Frame rate | 24fps |
|---|---|
| Resolution | 720p |
| Session length | 60 seconds |
What is it?
Project Genie is a consumer-facing research prototype from Google DeepMind that generates interactive 3D environments from text or image prompts. You describe a scene — a forest, a city street, a Mars colony — and the system produces a navigable world you can explore in your browser. It shipped January 29, 2026 for Google AI Ultra subscribers in the US.
How does it work?
It combines three models: Genie 3 (the world model that generates consistent interactive frames), Nano Banana Pro (image generation for initial scene creation), and Gemini (for prompt understanding). Genie 3 renders at 24fps and 720p, maintaining visual consistency for several minutes. It supports three modes: World Sketching (draw and explore), free exploration, and remixing existing worlds.
Why does it matter?
World models have been a research curiosity — interesting papers but nothing you could actually use. Project Genie is the first to ship as a real product people can try. The 60-second session limit shows the compute cost is still high, but the fact that it runs in real time at all is a milestone for interactive generative media.
Who is it for?
Game designers, creative professionals, anyone curious about generative 3D worlds.
Try it
https://labs.google/projectgenie (requires Google AI Ultra, US only)