How Project Genie uses AI
Project Genie is the public prototype of Genie 3, the world model Google DeepMind first previewed in 2025. Unlike a video generator, it renders the path ahead in real time in response to your movement, so the world reacts as you explore it rather than playing back a fixed clip.
A session has three parts. In world sketching you write a prompt for the environment and your character, or upload an image, and Nano Banana Pro draws a preview you can refine before entering. You then step into the world and move through it as Genie 3 generates each moment, working together with Gemini. Remixing lets you start from an existing world and change it into something new.
Google launched Project Genie on January 29, 2026 for Google AI Ultra subscribers in the United States aged 18 and over, and in May 2026 opened it to Ultra subscribers worldwide and added a Google Street View integration that turns real streets into explorable worlds. Sessions are capped at 60 seconds, at about 720p and 20 to 24 frames per second, and you can download videos of your worlds and explorations. Google lists realism and character control among the current limits of this research prototype.
