Google launches Project Genie for AI Ultra subscribers to explore infinite, interactive virtual worlds

Google DeepMind Launches Project Genie: Infinite AI Worlds for Ultra Subscribers

Google DeepMind has unveiled Project Genie, a groundbreaking web app prototype powered by the Genie 3 world model, enabling users to create and explore interactive virtual environments in real time. Available exclusively to Google AI Ultra subscribers in the US (18+), it marks a major step toward accessible AI-driven world simulation.

What is Project Genie?

Project Genie builds on Genie 3, a general-purpose world model announced in August 2025 that generates diverse, photorealistic interactive environments from text prompts at 720p resolution and 24 FPS. The prototype integrates Genie 3 with Nano Banana Pro and Gemini for immersive experiences like world sketching, exploration, and remixing. Users prompt with text or images to spawn living worlds, defining characters and movement modes such as walking, flying, or driving.

Core Features

Project Genie centers on three key capabilities for hands-on creation.

• World Sketching: Generate environments from text/images, preview with Nano Banana Pro for tweaks, and set first- or third-person views before entry.

• World Exploration: Navigate dynamically generated paths in real time, with adjustable cameras and simulated physics/interactions.

• World Remixing: Build on gallery worlds or randomize for inspiration, then download exploration videos.

​These features allow sustained interaction with consistent details, simulating natural phenomena, ecosystems, and fantastical scenarios.

Genie 3: The Power Behind It

Genie 3 advances world models by predicting environment evolution and action impacts, crucial for AGI and agent training in unlimited simulations. It excels in multi-modal synthesis (visual, auditory, textual), hierarchical representations, and real-time consistency without degradation. Applications span robotics, animation, fiction, and historical recreations, with photorealistic outputs modeling physics like water and lighting.

​

Access and Availability

Rollout starts for US Google AI Ultra subscribers, Google's premium tier with advanced features like Gemini 3 Pro, Veo 3.1 video generation, and 30 TB storage. Expansion to more countries is planned, positioning it as an experimental tool to discover novel use cases in AI research and media.

Limitations and Responsible Development

As an early prototype, Project Genie has constraints: worlds may not fully match prompts/physics, characters can lag, and sessions limit to 60 seconds. Google emphasizes responsible AI, gathering feedback from trusted testers to refine toward broader access and real-world applications like robotics.