In a massive leap toward interactive Artificial Intelligence, Google DeepMind officially launched Project Genie on Thursday, January 29, 2026.
Powered by the groundbreaking Genie 3 world model, this research prototype moves beyond static video generation (like Sora) to create interactive, explorable 3D environments in real time. The tool is currently rolling out to Google AI Ultra subscribers in the United States.
1. What is Project Genie?
Project Genie is a web-based “world-building” experiment. Unlike AI video generators that produce passive clips, Genie 3 simulates the physics and dynamics of a virtual space, allowing you to inhabit it.
- Real-Time Generation: As you move your character, the AI generates the path ahead on-the-fly at 24 frames per second.
- Multimodal Inputs: You can build a world using a simple text prompt or by uploading a single image (sketch, photo, or AI-generated art).
- The “Infinite” Illusion: It maintains surprising visual consistency; if you turn around, the world behind you remains the same—a major technical hurdle for older generative models.
2. Three Core Capabilities
Google has structured the Project Genie interface around three main user actions:
| Feature | Functionality |
| World Sketching | Define your environment, character, and perspective (1st or 3rd person). Uses Nano Banana Pro for high-fidelity previews. |
| World Exploration | Navigate through your creation. You can walk, fly, drive, or ride, with the AI predicting the physics of your movement. |
| World Remixing | Take an existing world (from a curated gallery or your own) and change the “vibe”—switch from a desert to a neon city with one prompt. |
3. Technical Specs and Limitations
While revolutionary, Project Genie is still an “experimental research prototype” with specific constraints:
- The 60-Second Cap: Current sessions are limited to one minute of continuous exploration due to immense computational costs.
- 720p Resolution: Worlds are rendered at 1280×720, prioritising speed and interactivity over 4K photorealism.
- “Vibe” Physics: The environment follows “video game logic” inferred from its training data rather than a rigid, real-world physics engine.
- Memory: The system maintains a “coherent memory” of your world for several minutes, though consistency can degrade over long sessions.
4. Pricing and Availability
Access to Project Genie is a high-tier exclusive, reflecting the massive GPU power required to run world models.
- Subscription: Available only to Google AI Ultra members ($250/month plan).
- Region: Restricted to U.S. users (18+) for the initial rollout phase.
- Device: Web-based via Google Labs, optimized for desktop Chrome browsers.
Conclusion: From Content to Environments
Project Genie marks the official transition of AI from a “content creator” to a “world builder.” While OpenAI’s Sora focused on the cinematic quality of video, Google is betting that the true future of AI lies in agency—the ability for users to not just watch a scene, but to step inside and change it. For game developers, robotics researchers, and storytellers, Project Genie is the first glimpse of a future where virtual worlds are limited only by the speed of our imagination.
