Archive · 30 Jan 2026
Google DeepMind Launches Project Genie
Google DeepMind has introduced Project Genie, a new AI tool that can create playable virtual worlds from simple text or image prompts.
Imagine typing 'a magical forest' and being able to walk, fly, or drive through it in real-time.
Project Genie uses an advanced AI model called Genie 3 to build these interactive environments.
Right now, it's only available to certain Google AI Ultra subscribers in the US, but Google plans to expand it later.
The AI can create natural settings like forests or deserts, as well as fantasy worlds.
As you explore, the world changes in real-time based on your actions.
You can also modify existing worlds or combine them to create new experiences.
However, there are some limitations, like the worlds not always matching the prompts perfectly or having some delays in character control.
Google says they will improve these features based on user feedback.
Google DeepMind launched Project Genie, an AI prototype that creates interactive virtual worlds from text and image prompts.
Project Genie uses Genie 3, an advanced AI model that builds dynamic environments in real-time.
The tool is currently available to Google AI Ultra subscribers in the US aged 18 and above.
Key features include World Sketching, World Exploration, and World Remixing, allowing users to create, explore, and modify virtual worlds.
Project Genie has some limitations, such as inaccuracies in matching prompts, latency in character control, and a generation duration of up to 60 seconds.
- Who
- Google DeepMind
- What
- Launched Project Genie, an AI prototype that creates interactive virtual worlds from text and image prompts
- When
- January 29, 2026
- Where
- United States (initially)
- Why
- To explore advanced world modeling and interactive AI experiences
- Project Name
- Project Genie
- Developer
- Google DeepMind
- AI Model
- Genie 3
- Availability
- US (Google AI Ultra subscribers, aged 18+)
- Key Features
- World Sketching, World Exploration, World Remixing
- Generation Duration
- Up to 60 seconds
- Limitations
- Inaccuracies, latency, and missing advanced features