Project Genie Street View Integration

Google DeepMind has integrated Google Street View imagery into Project Genie, enabling the general-purpose world model to generate interactive environments anchored in real-world locations. This update allows AI agents and robots to navigate and interact with real-world complexities within a simulated space, while providing users with tools to reimagine physical locations through generative styles.

Real-World Grounding via Street View

Project Genie now utilizes Maps Imagery Grounding technology to base generated worlds on actual physical locations. Users can select a location in the United States via a Maps pin and combine it with a specific style—such as "Desert Sands," "Stone Age," or "B&W film"—to transform a real-world site into an imaginative environment.

Key capabilities of this grounding feature include:

• Location-Based Generation: The model uses Street View imagery as the starting point for creating interactive worlds.
• Stylized Environments: Users can apply thematic overlays to real locations, such as visualizing the Golden Gate Bridge in an "Ocean World" style or the Fort Worth Stockyards as they might have appeared in the 1920s using "B&W film" style.
• Character Integration: Users can describe a character (e.g., a comic book hero or claymation monster) to inhabit these grounded environments.

Currently, Street View imagery integration is available for locations within the U.S., with plans for future expansion to additional regions.

Project Genie as a World Model

Genie serves as a foundational tool for research into general-purpose world models. Its ability to generate diverse, interactive environments has already been applied to several high-impact use cases:

• Agent Training: Enabling AI agents to learn and reason within complex virtual settings (as seen in SIMA 2).
• Autonomous Driving: Assisting Waymo in the simulation of hyper-realistic road environments.

Availability and Access

Project Genie, including the new Street View grounding capability, is an experimental research prototype hosted in Google Labs. It is currently rolling out globally to eligible Google AI Ultra subscribers (aged 18+).

Because it remains a prototype, Google DeepMind continues to refine the model to improve the sharpness and accuracy of the generated environments.