The dream of the Star Trek holodeck just took a massive leap forward on a standard web browser. Google has officially opened the doors to Project Genie, an experimental prototype that allows users to generate and explore infinite interactive worlds using nothing but text and image prompts. This release puts the power of the Genie 3 world model into the hands of Google AI Ultra subscribers in the United States and marks a significant milestone in the company’s quest to build general-purpose artificial intelligence that understands the physics of our reality.
• Project Genie is now available to US AI Ultra subscribers
• The tool allows users to generate infinite interactive worlds
• It runs on the advanced Genie 3 world model
Traditional video games rely on pre-rendered assets and scripted events, yet Genie 3 operates on an entirely different principle. The system acts as a real-time simulator that predicts how an environment should evolve frame by frame as the user interacts with it. This capability moves beyond static 3D snapshots to create dynamic pathways that generate instantly as you walk, drive, or fly through a landscape that did not exist mere seconds ago.
• The model generates environments in real-time
• It predicts physics and interactions frame by frame
• Users can explore worlds that are created instantly
Creation begins with a feature called World Sketching which integrates the Nano Banana Pro image model and Gemini’s language capabilities. Users can define the visual style of their environment and choose their perspective before stepping into the simulation. Fine-tuning tools allow creators to upload their own reference images or modify generated previews to ensure the digital realm matches their specific vision before the exploration begins.
• World Sketching uses Nano Banana Pro and Gemini
• Users can define visual styles and perspectives
• Reference images can be used to fine-tune the world
Collaboration drives the platform through a robust remixing system that encourages users to build upon existing foundations. Curated worlds in the public gallery serve as jumping-off points where prompts can be altered to create entirely new interpretations of a scene. The prototype essentially functions as an infinite sandbox where a single idea can fracture into thousands of unique timelines that can be recorded and shared as video clips.
• Users can remix existing worlds from a gallery
• Prompts can be altered to create new variations
• Explorations can be recorded and downloaded
Access remains exclusive to a specific tier of early adopters as the technology is still very much in its infancy. Google AI Ultra subscribers over the age of eighteen can start experimenting today, though they will face limitations including a sixty-second generation cap and occasional lapses in the laws of physics. Engineers view this rollout as a critical research phase to gather data on how humans interact with generative environments before expanding availability to the rest of the globe.
• Access is limited to US AI Ultra subscribers (18+)
• Generations are currently capped at 60 seconds
• The rollout serves as a research phase for Google
Via: Google Blog
