Google’s Project Genie Lets Users Explore AI-Generated Worlds

What would it look like to explore a virtual world built instantly from a text prompt? 

Google is giving select subscribers a new way to step inside AI-generated environments with Project Genie, an experimental prototype. The tool allows users to create and navigate interactive worlds that are generated in real-time.

The prototype is available first to Google AI Ultra subscribers in the US who are aged 18 and above and is powered by DeepMind’s Genie 3 world model. The launch highlights Google’s growing focus on “world models,” AI systems designed to simulate dynamic environments, with potential uses ranging from creative media to research and robotics. 

Google expands access to world model research 

Google said the prototype web app combines Genie 3, Nano Banana Pro, and Gemini, giving users a direct way to experience the capabilities of the company’s world model. 

The tech titan positioned the tool as part of its broader work on “world models,” AI systems that simulate how environments evolve and respond to user actions.

Project Genie is being released through Google Labs after limited testing with trusted users. The prototype builds on Google’s earlier preview of Genie 3, which demonstrated interactive environments rather than static 3D scenes. 

What users can do inside Project Genie

Google said Project Genie centers on three core capabilities: world sketching, world exploration, and world remixing. 

With world sketching, users can prompt with text and a generated or uploaded image to construct personalized worlds. Users can define their character, environment, and exploration style, from traditional movement like walking and driving to more imaginative options. 

Nano Banana Pro is integrated for more precise control, allowing users to preview and refine their worlds before entering them. Users can also choose their perspective, either first-person or third-person, to customize how they experience the environment. 

World exploration allows users to navigate environments as they are generated in real time. As people move, Project Genie predicts and creates the path ahead based on their actions. Meanwhile, world remixing lets users build on existing worlds by modifying prompts, exploring curated examples in a gallery, and downloading videos of their creations. 

Limitations and responsible AI development

Google emphasized that Project Genie remains an early research prototype with known limitations. “Generated worlds might not look completely true-to-life, always adhere closely to prompts or images, or follow real-world physics,” the company noted.

It also noted that characters in Project Genie can sometimes be less controllable, experience latency in controls, and are limited to 60-second generations.

Some Genie 3 capabilities previewed previously, such as prompt-driven events that dynamically change the world during exploration, are not yet included. According to Google, the rollout is intended to better understand how people will use world models across AI research and generative media while continuing to build responsibly. 

Access is currently limited to AI Ultra subscribers in the US, with expansion to more regions planned, but no timeline announced.

To learn more about how users can experiment with Nano Banana Pro, read our list of the best Nano Banana prompts to try.

The post Google’s Project Genie Lets Users Explore AI-Generated Worlds appeared first on eWEEK.