Google Project Genie explained: It turns text into playable 3D worlds

Imagine describing a place in words and instantly being able to walk around inside it. That’s the basic idea behind Google’s new AI experiment called Project Genie. It’s not a game engine and not exactly a video game either. Think of it more as a sandbox where you type a prompt and an explorable world appears in front of you. Also Read: Global smartphone processors may see a dip in 2026, says Counterpoint

Right now, this tool is rolling out as a research prototype for Google AI Ultra subscribers in the US, but what it shows is a glimpse of how AI could one day build interactive spaces. Also Read: Gemini and Google AI Mode to help you prepare for JEE Main exam: Here's how

Project Genie

The tech giant explained project Genie is built on Google’s “Genie 3” world model. Instead of generating just images or short videos, this model creates a full environment that reacts to what you do. You can start by typing a prompt or even uploading an image. Want a desert with floating islands and a character that can fly? Type it. Prefer a small town you can walk through? That works too. Also Read: How to turn your photo into vector image using Gemini Nano Banana; Check prompt here

You also pick how you want to see the world: first-person like a camera on your eyes, or third-person like a game character you control with arrow keys.

Google breaks the experience into three simple ideas:

• First is world sketching. This is where you define the environment using text and images. You can tweak the look before entering, thanks to an image preview system.
• Next is world exploration. As you move forward, the system generates the world in real time. It predicts what should appear next based on your direction and actions. You can change camera angles while moving around.
• Then comes world remixing. You can take an existing world, change the original prompt, and turn it into something new. There’s also a gallery of ready-made worlds you can open, play with, and modify. When you’re done, you can export your journey as a short video.

Not a finished product (yet)

Note that this is still an experiment. Worlds currently last about 60 seconds before you need to regenerate. Physics and character controls can feel imperfect, and what you get may not always match your prompt exactly.

Add Techlusive as a Preferred Source

Some advanced ideas Google showed earlier, like dynamic events that change the world mid-exploration, aren’t available here yet.