Google DeepMind’s Project Genie enables creating and exploring virtual worlds in real time

Project Genie is a research prototype that Google DeepMind has begun distributing among Google AI Ultra subscribers in the United States. The tool allows users to create navigable environments using text descriptions and images, which can then be explored in real time.

The technology powering this prototype is Genie 3, a world model that simulates the dynamics of an environment and predicts how elements evolve according to user actions. Unlike static 3D experiences, Genie 3 generates the path ahead in real time as the user moves and interacts. The model simulates physics and interactions to create dynamic worlds and enables the simulation of various scenarios, from robotics to fiction or exploration of historical locations.

Project Genie's functionality is based on three capabilities. The first is world sketching, which allows users to create an environment using text and images. Users define their character, the world and how to explore it, whether walking, flying or driving. Nano Banana Pro has been integrated to preview the world and adjust the image before entering, as well as to define the character's perspective.

The second functionality is world exploration. Once created, the environment becomes a navigable space where Project Genie generates the path in real time according to user actions, with the ability to adjust the camera during the journey.

The third capability is world remixing. Users can modify existing worlds to create new versions or explore worlds from the gallery. Once exploration is complete, users can download videos of the created worlds.

As an experimental prototype, Project Genie has limitations. Generated worlds may not appear completely realistic or adhere to real-world prompts or physics. Characters can be less controllable and generations are limited to 60 seconds.

Access has begun rolling out to Google AI Ultra subscribers in the United States aged 18 and over, with plans to expand to more territories soon.

Key points

• Google DeepMind launches Project Genie, a prototype for creating and exploring interactive worlds with AI
• Powered by Genie 3, a model that generates environments in real time as users interact
• Allows creating worlds with text and images, exploring them navigably and remixing existing creations
• Integrates Nano Banana Pro to preview and adjust worlds before accessing them
• Worlds are generated dynamically according to user actions during exploration
• Initially available to Google AI Ultra subscribers in the United States aged 18 and over
• Has limitations such as visual realism, character control and maximum duration of 60 seconds
• Google plans to expand access to more territories in the future