I’ve developed a workflow that turns real-world satellite imagery into functional 3D placeholder geometry. By isolating open spaces and grassy zones, I can establish exactly where the roads and pathways will end up.
The Workflow: AI-Driven Simplification
The process began with a high-resolution satellite JPG from Google Maps.
A look at the 70x70m fuel yard layout—the heart of the game level that was captured from a location on Google Maps.
To turn a cluttered photo into 3D geometry, I needed to strip away the noise. I fed the image into ChatGPT, instructing it to simplify the map down to its most basic color blocks.
Color Palette simplified by ChatGPT
By flattening the image into distinct colors, I was able to take the result into a vector editor and trace the green "open areas" into clean paths. These paths were then imported into Blender and extruded to create 3D placeholders—the negative space where the citizens of the Throne will eventually walk.
Visualizing the Pathways
The final step was bringing these placeholders into A-Frame as a .glb file. By overlaying these extruded green zones onto the satellite ground plane, I can verify the flow of the world in a 1:1 scale VR environment.
Extruded green areas exported as .glb and imported/placed in A-Frame
"We aren't just placing buildings; we are defining the gaps between them where the story actually happens."
The Next Step: Raising the Steel
Now that the open areas and pathways are established, I have a logical blueprint for the structures. These "green blocks" serve as my boundaries—I now know exactly where my buildings must sit to create a claustrophobic, industrial atmosphere without breaking the player's ability to navigate.
As the project matures, I’ll be replacing these placeholders with high-fidelity industrial models and setting up the collision boxes for our first patrol AI tests.
The grid is set. The forge is heating up.
Technical breakdown by Scott Grocott · Tools used: ChatGPT, Blender, A-Frame