Loading

← AI Library

Axonometric map from open data

Turn the Seongsu map yourself, then follow how open map data becomes a neighborhood axonometric.

Format
Interactive
Tools
Claude Code · OpenStreetMap · Overture Maps
Made
15 Nov 2026

How it’s madePrompt

How it’s made

Open map data becomes a 3D model, drawn with the line hierarchy of an architectural drawing. The picture is not an image: the browser draws the model every time.

  1. 01

    Get the map data

    Take building footprints, streets and rail from OpenStreetMap and Overture Maps. Turn the map to the main street so the buildings line up with the drawing’s axes.

    Plan of the Seongsu Station area with every building footprint in black
  2. 02

    Raise the heights

    Use the floor count where the map has one. Where it does not, estimate it from the footprint and the shadows in aerial photos. Only 72 of the 294 buildings here had a height on the map.

    White massing model with heights only
  3. 03

    Draw in lines

    Lines instead of grey shading. Find them in the depth and normal buffers, then thin them in order: outline, edge, joint, window. Walls that touch a neighbor get no windows.

    The same model drawn with the line hierarchy alone
  4. 04

    Add materials and light

    Red brick, green waterproofed roofs and ginkgo street trees, in three flat tones. The shadows follow the real sun over Seongsu on 15 November.

    The finished drawing with materials, shadows, street trees and people

    Change the hour above ↑

Prompt to reproduce

The opening request to send to Claude Code. Change the neighborhood and the extent. It will not give the same result word for word.

Request to Claude Code

Make an axonometric map of [neighborhood], drawn the way an architect would. Center it on [landmark] and cover about 500 m by 350 m.

1. Data: from Overture Maps (which includes OpenStreetMap), take the building footprints, floor counts and heights, streets and rail in this area. Estimate the floors of buildings without them from their footprint and their neighbors, and list the estimated ones separately.
2. Model: turn the map to the main street and raise every building to its height. Put parapets, stair cores, water tanks and AC units on the flat roofs, and sidewalks, lane lines and crosswalks on the streets.
3. Drawing: WebGL2 with an orthographic isometric camera. Find the lines in the depth and normal buffers and thin them in order: outline, edge, joint, window. Fill faces in three flat tones and cast shadows from the real sun position of the day.
4. Screen: one HTML file. Drag to move, Shift + drag to turn and tilt, Ctrl + wheel to zoom, a time-of-day slider (windows light up at night), a plan view with only the buildings in black, and an exploded view. When the page opens, play one short tour: start from the plan, tilt into the axonometric, turn a full circle, then explode and settle back. Stop the moment someone touches the map.

Made by Planby with Claude Code. Map data © OpenStreetMap contributors, Overture Maps Foundation. The 3D model data on this page is available under ODbL 1.0. Heights of 222 of the 294 buildings are estimated, and materials and windows are chosen for the drawing, so they may differ from the real buildings. No AI-generated images.

Let's talk about AI for your team

We'll suggest the right products and how to adopt them.

We collect your name, company, work email, phone, department and role, plus visit data (page URL, referrer, UTM and time of visit) to see how you found us. We use it to respond and send materials, keep it for one year, then delete it.