Google Patents a Way to Turn Location Searches Into Interactive 3D Walkthroughs
Type in a place, get back a 3D walkthrough complete with simulated rain or rush-hour traffic. That's the vision behind a new Google patent that turns location searches into explorable video environments.
What Google's location-to-3D video system actually does
Today, searching for a place online gives you a flat map, some photos, and maybe a Street View snapshot. You have to piece together what a location actually feels like from those disconnected fragments. Google's new patent describes a system that would do that assembly for you, automatically.
The idea is to take a search query about a specific place and respond with a three-dimensional video of that location. That video wouldn't just be a photo stitched into a sphere; it would be a proper 3D model you could view from multiple angles, built from geographic map data and real photos of the spot.
The system could also layer on simulated conditions, showing you what a plaza looks like during a rainy afternoon or what a street corner looks like at peak traffic. Think of it as the difference between reading a restaurant review and actually walking past the front door.
obtaining a three-dimensional model that models a plurality of locations, wherein the three-dimensional model is generated based on geographic data, wherein the geographic data comprises vector data; …
Translation: The system starts by gathering map and vector data to build a base 3D world.
How Google builds a 3D mesh from maps data and photos
The patent describes a pipeline that starts with existing geographic data, including the kind of vector data (mathematical descriptions of roads, buildings, and boundaries) that already powers Google Maps. From that, the system generates a three-dimensional mesh of a specific location, essentially a wire-frame skeleton of the place.
Separately, the system processes real photographs of the same location to produce segmented images, where each part of the photo is labeled by what it depicts: sky, pavement, building facade, trees. Those labeled photos are then projected onto the 3D mesh, wrapping the skeleton in realistic visual detail pulled from actual imagery.
The result is a location-specific segment, a reconstructed 3D environment of the place. The system then renders that environment as a video showing multiple different viewpoints, so you're not locked into one fixed angle.
On top of all that, the patent describes adding simulated situational effects:
- Simulated weather (rain, fog, bright sun)
- Simulated traffic conditions
- Other context-specific overlays
All of this is triggered by a user submitting a location-related query, meaning the 3D video would be generated as a direct response to a search.
Generating a location-specific three-dimensional model in response to a location query can provide users with a better understanding of a location through providing better interactivity, better perspective, and better understanding of dimensionality.
Translation: Turning map searches into 3D models gives users a much clearer view of a place.
What this means for how you research places before you go
If this made it into Google Search or Maps, it would change how you evaluate a place before committing to it. Right now, judging whether a neighborhood feels walkable, whether a venue has enough parking, or whether a hotel entrance is easy to find requires hunting across photos, satellite views, and Street View separately. A single 3D video response collapses all of that into one thing you can actually rotate and inspect.
Google's steady investment in 3D mapping and spatial search suggests this fits a broader push to make search results feel less like a list of links and more like a direct answer. The simulated-conditions angle is particularly telling: showing you a location as it would actually look at the time you'd be visiting it is a genuinely practical upgrade over a sunny promotional photo taken in July.
This is the 670th Google filing we've tracked since May in our Google coverage, a corpus that spans work like facial recognition for home cameras and voice bots reading account codes.
The most concrete payoff here is simple: you'd stop making decisions about places based on photos that were taken under ideal conditions. A hotel courtyard photographed on a clear afternoon looks very different from what you'll find at 7pm in November. If Google's system can show you a location under realistic or even user-specified conditions, that's a real improvement in how you plan.
The harder question is whether the output would be accurate enough to trust. A 3D model reconstructed from map vectors and segmented street photos could easily introduce distortions or miss recent changes. If a building was renovated last year but the photo database hasn't caught up, you'd be exploring a location that no longer exists. That's a higher-stakes failure than a bad search result.
This is a technically ambitious filing, and the underlying components (mesh generation, image segmentation, projection) are all areas where Google has real depth. Whether this becomes a search feature or stays a research project depends entirely on how well the reconstruction quality holds up at scale.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
13 drawing sheets from US 2026/0268615 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →