Details
- Google DeepMind is adding a Street View–grounding capability to Project Genie, its experimental web-based prototype powered by the Genie general-purpose world model.
- The feature lets users anchor interactive, AI-generated worlds to real U.S. locations by selecting a Maps pin, choosing a style (e.g., “Desert Sands,” “Stone Age,” “Ocean World,” “B&W film”) and describing a character.
- Genie uses Google Maps Imagery Grounding and Street View imagery to generate navigable environments that creatively reimagine real places, such as exploring the Golden Gate Bridge underwater or Fort Worth Stockyards as a 1920s film scene.
- Street View-based worlds currently work for U.S. locations, with Google planning to expand coverage to more geographies over time as it refines realism, sharpness and physical consistency.
- Access to Project Genie, including the new Street View capability, is now gradually rolling out globally to all eligible Google AI Ultra $200 subscribers aged 18 and over, extending an earlier U.S.-only release to a broader user base.
Impact
Grounding Genie in Street View imagery moves world models closer to real-world deployment for robotics, navigation and simulation-heavy applications, aligning with a broader industry push toward physically grounded AI. By tying the feature to the Google AI Ultra subscription, Google deepens its premium AI value proposition and seeds a developer and creator ecosystem around world-model-based experiences over the next 12–24 months.
