Google has unveiled Project Genie, an experimental AI system that generates fully playable video game worlds in real-time based on text prompts, images, or real-world locations from Google Maps Street View. The prototype, housed under Google Labs, represents a significant leap from passive AI content generation like images and videos to interactive, dynamic environments that respond to player input. Users can explore infinitely expanding worlds for up to 60 seconds per session, with the AI synthesizing new frames and physics on the fly rather than pulling from pre-made game assets.

What Powers Project Genie's Real-Time World Generation?

Project Genie operates as a "world model" rather than a traditional game engine, combining three core AI systems to create its interactive experiences. The foundation rests on Genie 3, a world model trained on massive amounts of video data including internet gameplay footage and real-world video. This training allows the system to understand physics, character movement, and environmental logic without explicit programming. Gemini, Google's flagship multimodal AI model, interprets your text and image prompts and translates human intent into parameters the world model can process. A newly introduced internal technology called Nano Banana Pro optimizes real-time frame generation to keep latency low enough for responsive, interactive gameplay.

When you press a jump button or move your character forward, the AI predicts what should happen in the context of your created world and renders those frames instantly. The environment doesn't have a predetermined end; instead, it expands infinitely based on your actions in the present moment, creating a genuinely dynamic experience.

How to Create and Explore Your First Project Genie World

• Text and Image Prompts: Type detailed descriptions of your desired world, such as "A cyberpunk city bathed in neon pink rain, played as a robotic cat." You can also upload your own images to serve as the aesthetic foundation for generation.
• Google Maps Street View Integration: Select any real-world location via Street View, apply an artistic style to it, and the AI transforms it into a playable level. Users in Mumbai, Delhi, or Bangalore could theoretically turn their local neighborhoods into platformer levels.
• Randomizer Option: Click the randomizer button to let the AI surprise you with a completely unique landscape and character, perfect for spontaneous exploration.

Once you generate a world, you're dropped directly into the environment with familiar gaming controls. Use WASD keys to move, the spacebar to jump or fly, and arrow keys to adjust your camera orientation. You have 60 seconds to explore before the session ends, though this timeframe is sufficient to experience the core breakthrough of AI generating a video game frame-by-frame beneath your digital feet.

Who Can Access Project Genie Right Now?

Google is taking a cautious, phased approach to rolling out Project Genie due to its substantial computing requirements. Access currently requires a Google AI Ultra subscription, as the free tier doesn't support world generation. You must be at least 18 years old, and the rollout is currently limited to users in the United States, though Google has stated it will expand to additional territories "in due course." Indian users with US Google accounts may be able to access it through early-access workarounds, but official support for India is anticipated.

To access the prototype, navigate to labs.google/projectgenie, sign in with your Google Account tied to your AI Ultra membership, and click the Create with Project Genie button. You may be placed in a queue due to computational demands; keep the tab open and monitor the progress bar. Once it's your turn, you have 30 seconds to enter your prompts and choose between first-person or third-person perspective before generation begins.

What Happens After Your 60-Second Exploration?

When your exploration time expires, Project Genie offers several options for what to do next. You can revisit your world using the exact same prompts, which will generate a slightly different, unique result each time. Alternatively, you can reuse your prompts but tweak them to explore variations. The system also allows you to download a video file of your 60-second gameplay to share on social media platforms like X or YouTube, or with friends. If you want to start completely fresh, you can scrap your creation and build an entirely new world from scratch.

Google has also built straightforward data management tools into the prototype to address privacy concerns. You can delete your creations at any time by clicking the "More" icon in the top right of the app and selecting Delete Data. If you want to back up your worlds and data, Google allows you to request a full export by contacting their support team.

Why Does Project Genie Matter for the Future of AI and Gaming?

Project Genie marks a fundamental shift in what generative AI can accomplish. Rather than producing static images or passive videos, it creates interactive, responsive environments that adapt to user input in real-time. This capability has profound implications for gaming, simulation, education, and entertainment. Developers could potentially use similar technology to accelerate game creation, reducing the years typically required to build complex worlds. The system's ability to understand physics and environmental logic without explicit programming suggests that AI-generated interactive content could eventually rival hand-crafted game design in complexity and responsiveness.

The Gallery feature adds a collaborative dimension, allowing users to remix worlds created by others. If you discover a gorgeous alien landscape or eerie environment created by another user, you can click Remix to borrow their base prompts, tweak the text or character, and spawn a completely new interpretation of their idea. This community-driven approach could foster creative experimentation and accelerate the discovery of compelling world designs.

Google's cautious rollout strategy reflects the significant computational resources required to generate worlds in real-time. By limiting access to AI Ultra subscribers and capping exploration time at 60 seconds per session, Google manages server load while gathering user feedback and refining the technology. As the system matures and computational efficiency improves, broader access and longer exploration sessions could become feasible, potentially opening Project Genie to millions of users worldwide.