Google’s Project Genie Lets People Create and Explore AI-Generated Worlds

Google has launched an experimental new tool called Project Genie that lets users make and explore AI-generated interactive worlds using text and images. This is part of Google’s research into advanced artificial intelligence and “world models,” which could help computers understand and simulate environments like real places.

What Project Genie Is

Project Genie is a research prototype that’s now available to people in the United States who subscribe to Google AI Ultra. With it, users can:

• Create worlds by typing descriptions or uploading images
• Explore the worlds in real time just like walking through a scene
• Remix worlds made by others and make them your own

The tool is powered by Genie 3, a new AI world model developed by Google DeepMind. This system builds environments that respond as you move and interact with them — like a primitive video game you describe with words and pictures.

How Project Genie Works

Project Genie lets users design and explore interactive 3D environments using three main features:

1. World sketching: Users type a text prompt or upload an image to start building a digital environment.
2. World exploration: Once a world is made, you can walk around it and see what it’s like.
3. World remixing: You can take someone else’s world and change it to make something new.

These worlds are generated in real time as you explore, so they can adapt to your actions — making each visit feel different.

Why Google Is Building It

Google says Project Genie is part of its research into how AI can understand and simulate environments, a step toward more advanced AI systems that could learn and make decisions like humans. World models — AI systems that can predict how an environment works and how actions affect it — are a key part of this research.

These technologies may someday help:

• Train robots by having them practice tasks in simulated spaces
• Build games, education tools, or simulations where people can learn by exploring virtual worlds
• Advance AI that can reason and plan ahead by understanding how actions change a scene

Limitations and Future Plans

Project Genie is still early and experimental. Some limitations include:

• Generated worlds don’t always look perfectly real
• Characters may be hard to control
• Worlds are currently limited in how long they can last without resetting

Google says it hopes to improve these features in future updates and make the technology available to more people over time.

Bottom Line

Project Genie is a prototype from Google DeepMind that lets people use AI to build and explore interactive worlds with text and images. It shows how artificial intelligence is evolving beyond simple text or image generation into systems that can simulate entire environments — a potential step toward more powerful, embodied AI in the future.