Google DeepMind has expanded its Genie world model to transform real-world locations from Google Maps into interactive, navigable AI environments. The update, announced at Google I/O and now rolling out to subscribers, lets users select any geolocated place on Earth, apply a creative style, and watch Genie generate a fully interactive 360-degree world built on top of that location's actual Street View imagery. This represents a significant shift in how generative world models can leverage real-world data.

What Makes This Different From Previous World Models?

Until this update, Project Genie had been available to Google AI Ultra subscribers since January 2026, but it operated purely in the realm of imagination. Users would provide a text prompt, and the model would invent an entirely synthetic environment from scratch. The new Street View grounding feature flips this approach completely. Instead of starting from pure algorithmic imagination, Genie now begins with a real, photographed location and builds a stylized variant on top of it.

The distinction matters because it demonstrates a fundamental capability of world models: they don't just generate static images or video clips. A world model is designed to simulate environments that respond to user actions. Rather than producing a fixed sequence of frames, Genie predicts what the world looks like next based on what you do within it. The Street View integration operates at 20 to 24 frames per second, creating fluid, real-time interaction within the generated environment.

How Does the Street View Integration Actually Work?

The technical foundation relies on something called Maps Imagery Grounding, the same technology developers use to create AI visuals with Street View. The workflow is straightforward: you pick a real location from Google Maps, select a creative style you want applied, and Genie generates an interactive world based on that place's actual imagery. What makes this possible is the sheer scale of Google's underlying dataset.

Project Genie can now draw on more than 280 billion images captured across 110 countries and all seven continents. This archive represents a competitive advantage that few other AI labs can easily replicate. The integration of this massive real-world dataset with generative world models shows how foundation models trained on diverse, global imagery can ground abstract generation in concrete reality.

Steps to Understanding World Models vs. Video Generators

• Video Generators: These systems produce fixed sequences of frames based on a prompt or initial image. Once generated, the output cannot be changed or interacted with in real time. They are fundamentally passive.
• World Models: These systems simulate environments that respond to user input. They predict the next frame based on both the world description and the user's actions, enabling real-time, interactive exploration. Genie operates as an auto-regressive system, building frames one at a time based on context and user behavior.
• Grounded World Models: By anchoring generation to real Street View data, Genie combines the responsiveness of world models with the authenticity of real-world locations, creating environments that feel both interactive and geographically plausible.

This capability is particularly significant because it bridges two previously separate domains: generative AI and real-world geographic data. Users can now explore stylized versions of actual places they know, rather than purely fictional environments. A user could select their hometown, apply a cyberpunk aesthetic, and navigate through a transformed version of their own streets.

What Does This Mean for the Future of Interactive AI?

The Street View grounding feature is not a new product launch; it is a meaningful expansion of an already-live research prototype. Project Genie has been accessible to Google AI Ultra subscribers since January 2026, but this update significantly broadens its practical applications. The ability to ground world models in real geographic data opens possibilities for virtual tourism, urban planning visualization, creative storytelling, and interactive design tools.

The timing coincides with broader industry momentum around spatial AI and immersive experiences. At SIGGRAPH 2026, the world's premier conference on computer graphics and interactive techniques, the focus on spatial storytelling and immersive experiences underscores how the field is moving toward environments where users can step inside narratives and interact with digital worlds in embodied ways. Genie's Street View integration fits into this larger shift toward interactive, responsive digital environments that feel grounded in real places.

The convergence of world models, real-world data, and interactive design represents a meaningful step forward in how AI can augment human creativity and exploration. Rather than replacing real places, these tools allow people to reimagine and interact with the world in new ways, grounded in the geographic authenticity that Google's 280 billion-image archive provides.