Google’s Genie world model now simulates real streets with Street View

Most of us have used Google Street View to show a friend our old neighborhood or dropped the little person icon into Paris to check if our hotel is in a trendy area. Now, imagine experiencing that with full immersion and interactivity—simulating the street and its surroundings, adjusting weather conditions, or even visualizing extreme scenarios like those in “The Day After Tomorrow.”

This is the vision behind Google’s latest integration. Starting today, Google DeepMind links Street View with Project Genie, its general-purpose world model capable of generating diverse, interactive environments. The feature debuted during the Google I/O developer conference.

“It’s incredibly powerful for both agent and robotics applications, as well as for human exploration,” said Jack Parker-Holder, a research scientist on DeepMind’s open-endedness team, speaking to TechCrunch. “That has always been the core idea behind Genie.”

He illustrated this with a scenario involving a new robot deployed in London, where sunlight is rare. According to Parker-Holder, Genie can simulate those infrequent moments when sunlight glints off Victorian buildings, ensuring the robot isn’t startled by sudden brightness.

Loading the player…

“Alternatively, you might say, ‘I’m visiting New York City, but not during this season,’” he added. “‘It’s going to snow. I want to see what that block looks like under snow.’”

For two decades, Google has gathered Street View data using camera-equipped cars and individuals wearing “tracker backpacks.” The tech giant has amassed over 280 billion images across 110 countries and all seven continents.

“Street View gives us imagery from a vast portion of the world,” Jack explained. “Imagine the potential of combining this rich real-world data with the ability to simulate entire worlds.”

Last August, Google released Genie 3 for research preview, and in January, it granted access to Google AI Ultra subscribers in the U.S., enabling them to create interactive game worlds from text prompts or images. The aim is to leverage Genie for educational experiences, gaming, and robotics training.

Genie 3 is already powering one of Waymo’s simulators to train self-driving cars on “extremely rare events” such as tornadoes or unexpected elephant encounters. Incorporating Street View data could help Waymo prepare for launches in more global cities.

Waymo uses its own simulator to expand operations to 11 U.S. cities and test its AI driver in several others. Parker-Holder notes that while these simulations are from the car’s perspective, Street View allows for simulating a world anchored to a real location and shifting the viewpoint to other agents, such as humans or robots.

Google is rolling out Street View in Genie to select Ultra users in the U.S. today, with broader access for global Ultra users coming in the next few weeks, according to the company.

Diego Rivas, a product manager at DeepMind, stated that the researchers aim to make this capability widely available. He cautioned that Street View, and Genie in general, remains experimental, with significant room for improvement in accuracy.

In the demos shown by Google—including an underwater simulation of a neighborhood I once lived in—the results are impressive and recognizable but still resemble video game graphics rather than photorealism. The models also lack physics awareness, meaning they don’t yet grasp cause and effect. For instance, in a simulation of a woman running through snowy Joshua Tree, she passed directly through cacti and bushes.

Contrast this with Google’s image generator Nano Banana, which can now generate perfect text in infographics, or its video generator Veo, which understands that paper boats drift on water currents, smoke disperses into the air, and fabric drapes over forms.

Physics isn’t hard-coded into these models; they learn it intuitively over time through passive observation, much like a living being.

“I believe this type of model is about six to 12 months behind video in terms of accuracy and quality, so I think we will resolve that,” Parker-Holder said.

Jonathan Herbert, director of Google Maps who started on the Street View team as an intern 12 years ago, noted that Genie cannot yet create a faithful reconstruction of a street. He believes the real breakthrough lies in the AI’s spatial continuity: if you turn 360 degrees, the AI correctly remembers and simulates the environment behind you. From there, the model can build a new environment on top of that foundation.

“We’ve long considered how to build the best and richest model of the world using Street View data,” Herbert said. “It’s definitely been our goal to use Maps Data in new ways and for new kinds of AI research for quite some time.”