satprint -- Satellite Terrain to 3D-Printable Models

Pick a place on a map. Get a watertight STL of its terrain and buildings, a multi-color 3MF ready for a multi-material printer, and a GLB with satellite imagery draped over it.

satprint turns real elevation data into a solid relief model scaled to your printer: the terrain on top, four walls and a flat base. Cities can carry their OpenStreetMap buildings, and the multi-color 3MF splits the model into land, water, buildings and a border frame, one filament each.

Documentation: suchanek.github.io/satprint

satellite elevation  ->  heightmap (meters)       ->  STL solid
AWS Terrain Tiles        clamp sea, smooth,           terrain + walls + base
(SRTM, ASTER, GMTED)     scale to mm, exaggerate      + OSM buildings
                                                      + land/water/border 3MF

Mount Fuji from real elevation tiles, 120 mm wide, 1.2x exaggeration:

Features

• Real elevation, no API key. Elevation comes from the public AWS Terrain Tiles set (SRTM, ASTER, GMTED and others), cached under ~/.cache/satprint/tiles. You can also upload a 16-bit PNG, TIFF or GeoTIFF heightmap, or use a procedural demo terrain that needs no network.
• Print-aware output. True plan scale, vertical exaggeration or a fixed relief height, base thickness, sea-level flattening, Gaussian smoothing and grids up to 1024 columns. Each build reports model size, scale, triangle count, volume and an estimated PLA weight.
• Watertight by construction. Every edge is shared by exactly two outward-wound triangles, so slicers (Bambu Studio, PrusaSlicer, Cura, Chitubox) load the STL without repair.
• Buildings from OpenStreetMap, as closed solids standing on the terrain. Landmark towers keep their setbacks where OSM maps them as building:part shapes, domes, onion domes, cones and pyramids get their roof shapes, and overlapping footprints are merged so no two solids pass through each other. A few landmarks, such as the Sphere in Las Vegas, get exact shapes. Heights are in true proportion by default, with a multiplier. OpenFreeMap, Overpass or Overture Maps.
• Multi-color 3MF. Land, water, buildings and an optional border frame are separate parts of one object, already on filaments 1 to 4 in Bambu Studio. Water is the OSM sea, rivers and lakes, plus the flattened sea.
• Border frame. A rectangular rim around the model, 1 mm above the base by default, in the STL, GLB and 3MF.
• Textured GLB. Satellite imagery draped over the terrain and the roofs, for viewing in Blender, macOS Quick Look or a web viewer. The web preview shows it.
• Search and presets. Search any place by name, or pick one of 67 presets: 37 cities and landmarks, which turn buildings on, and 30 mountains and landscapes.
• Web app, CLI and REST API. Builds run as background jobs, and the page shows each stage with tile counts. Interactive API docs are at /docs.

Install

satprint needs Python 3.12 or 3.13.

git clone https://github.com/suchanek/satprint.git
cd satprint
poetry install                   # add --extras geotiff for GeoTIFF georeferencing

Without Poetry, pip install -e . in a virtual environment installs the app. See Installation for Docker.

Run the web app

satprint serve                   # http://127.0.0.1:7417
satprint serve --host 0.0.0.0 --port 7417

1. Choose an area. Search for a place, pick a preset, click Draw rectangle and drag on the map, or type the bounds. The hint line shows the real size of the area and the model it makes.
2. Set the print. Model width, base thickness, vertical exaggeration (1.5x to 3x reads well for most landscapes; 1x is true scale) or a fixed relief height. Resolution sets the grid columns: 256 is fine for FDM, 512 to 1024 for resin. Check Add buildings for a city, and Multi-color 3MF plus a Border frame width for a multi-material print.
3. Generate. The textured model appears in the 3D preview. Download the STL to print in one color, the 3MF to print in several, or the GLB to view.

After you change the code, restart satprint serve to pick up the change.

Print in several colors

The multi-color 3MF holds up to four parts, in this filament order:

Filament	Part	Display color
1	land	green
2	water	blue
3	buildings	white
4	border	dark gray

To print it on a Bambu printer with an AMS:

1. In Bambu Studio, set up four filaments in the left sidebar, or sync them from the AMS.
2. Open the 3MF. Bambu Studio reports that it loads "geometry only", as it does for any 3MF it did not write; the part names and filaments still come through.
3. To check or change a part's filament, open the object list: in the left sidebar, under Process, click Objects and expand the model.
4. Slice. The preview shows each part in its filament's color.

Other slicers open the same parts with every part on filament 1; assign the filaments by hand.

Command line

# Matterhorn, 120 mm wide, 2x exaggeration
satprint build --bbox 45.93 7.58 46.02 7.72 --width 120 --exaggeration 2 -o matterhorn.stl

# Fixed 15 mm relief, smoothed, with a hillshade preview PNG
satprint build --bbox 36.02 -112.25 36.20 -111.95 --relief 15 --smoothing 1 \
               --preview canyon.png -o grand-canyon.stl

# Midtown Manhattan with buildings, plus a textured GLB
satprint build --bbox 40.7414 -73.9997 40.7684 -73.9683 --width 150 \
               --buildings --glb midtown.glb -o midtown.stl

# Venice as a four-color 3MF: land, water, buildings and a 5 mm border frame
satprint build --bbox 45.43 12.32 45.446 12.343 --buildings --frame 5 \
               --3mf venice.3mf -o venice.stl

# From your own DEM in meters, 12 km across
satprint build --file dem.tif --ground-width 12000 -o dem.stl

# Offline demo terrain
satprint build --synthetic -o demo.stl

Bounds are SOUTH WEST NORTH EAST in decimal degrees. The command prints a JSON summary: elevation range, scale, triangle count, volume and a watertight check. satprint build --help lists every option.

REST API

Method	Path	Purpose
GET	/api/presets	Named example areas, grouped
GET	/api/search?q=...	Place search; each hit has a bbox
POST	/api/upload	Multipart heightmap upload; returns an upload_id
POST	/api/model	Build a model; returns stats, a preview PNG and download URLs
POST	/api/jobs	Same body as /api/model, built in the background; returns a job_id
GET	/api/jobs/{id}	Job status, current stage with done and total, and the /api/model response as result when done
GET	/api/model/{id}/{name}.stl	Binary STL
GET	/api/model/{id}/{name}.glb	Textured GLB, when glb_url is set
GET	/api/model/{id}/{name}.3mf	Multi-color 3MF, when threemf_url is set ("multicolor": true)
GET	/api/model/{id}/heightmap.png	Hillshade preview

curl -s localhost:7417/api/model -H 'content-type: application/json' -d '{
  "source": "terrarium",
  "bbox": {"south": 35.28, "west": 138.65, "north": 35.45, "east": 138.82},
  "width_mm": 150, "exaggeration": 1.2, "resolution": 512, "name": "fuji"
}' | python -c 'import json,sys; j=json.load(sys.stdin); print(j["stl_url"], j["info"]["height_mm"])'

How the scaling works

• Plan scale is width_mm / ground_width_m. The depth follows the true aspect ratio of the area, from great-circle distances along the center lines of the bounding box.
• Vertical scale is the plan scale times the exaggeration, so exaggeration = 1 is a true-scale model. With relief_mm set, the vertical scale puts the highest point exactly that far above the base, and the effective exaggeration is reported back.
• Sea level. Samples below 0 m (bathymetry in the source data) are clamped to 0 by default, so coastlines print as a flat plane.
• Buildings use the plan scale for their heights, so building_scale = 1 keeps them in true proportion to the map.

Data sources

Data	Source	Cache
Elevation	AWS Terrain Tiles	~/.cache/satprint/tiles
Imagery (GLB texture)	Esri World Imagery	~/.cache/satprint/imagery
Buildings and water	OpenFreeMap vector tiles, zoom 14, rebuilt from OSM about weekly	~/.cache/satprint/vtiles
Buildings, fallback; roof shapes	Overpass API, 0.01° tiles	~/.cache/satprint/osm
Buildings, optional	Overture Maps GeoParquet on S3, per release	~/.cache/satprint/overture
Place search	Nominatim, one request per second	in memory

• Buildings. A building with no height or building:levels tag gets 8 m. Building parts are extruded from the ground, ignoring min_height, so nothing floats. Buildings sink 0.3 mm into the terrain so they fuse with it when sliced. Areas are limited to 40 km² with buildings on, and very small footprints are dropped. A building that crosses a vector-tile edge arrives as two solids that meet at the edge.
• Roof shapes. Roofs tagged roof:shape dome, onion, cone or pyramidal get that shape, from roof:height or roof:levels, else a hemisphere-like height from the footprint's size. Other roofs are flat. The vector tiles carry no roof tags, so a small Overpass query adds them; if it fails, the roofs are flat and the build says so. A shaped roof starts no lower than the larger flat roofs around it, and smaller parts standing on it, such as a lantern on a dome, stand in a hole cut in the roof, so the solids never pass through each other.
• Landmarks. A few buildings that roof tags cannot describe are replaced by exact shapes from published dimensions: so far the Sphere in Las Vegas, a 157 m sphere cut by the ground at 112 m.
• Overture Maps. Choose it with "Building data" or --building-source overture. It merges OSM with Microsoft and Google footprints, so it finds buildings OSM lacks, and keeps the roof tags. It needs the overture extra (pip install "satprint[overture]", about 190 MB with pyarrow; the Docker image has it).
• Overpass. Choose it with "Building data" in the web app or --building-source overpass on the CLI, for the latest OSM edits. The default "auto" uses it only if OpenFreeMap fails. Its tiles download two at a time, busy answers are retried with backoff, and a server that times out is skipped for 10 minutes. When some tiles fail, the rest stay cached, so generating again fetches only the missing ones.
• Imagery. The texture is at most 2048 px on its longer side. Esri's terms of use govern the imagery. Set "texture": false to skip it.

Limits

• Source resolution is about 30 m (SRTM) over most land and about 10 m at zoom 14 where available, so areas smaller than about 2 km look blocky. Smoothing helps.
• Elevation and imagery downloads are capped at 64 tiles per request. Shrink the area or lower the resolution if you hit the cap.
• The multi-color split works one grid cell at a time, so a river narrower than a cell does not show. Raise the resolution to keep it.
• The GLB is for viewing. FDM slicers ignore textures, so print the STL or the 3MF.
• Built models live in memory, the last 32. A download link is valid until the server restarts.

Layout

satprint/
  terrain.py    elevation sources (terrain tiles, synthetic, file), imagery, scaling, hillshade
  mesh.py       heightmap to watertight solid, land/water split, frame, STL, GLB and 3MF writers
  buildings.py  OSM buildings to closed solids on the terrain, roof shapes
  landmarks.py  exact shapes for a few landmarks
  overture.py   Overture Maps building source (overture extra)
  water.py      water map and the multi-color 3MF parts
  osm.py        OpenFreeMap, Overpass and Nominatim clients
  presets.py    named example areas
  app.py        FastAPI backend, background jobs, in-memory model store
  cli.py        satprint serve / satprint build
  static/       web app: index.html, style.css, app.js (Leaflet and three.js from CDNs)
tests/          pytest suite, offline: every network client is faked

Development

poetry install --with dev
pre-commit install
pytest -q

The pre-commit hooks run the standard file checks, ruff, detect-secrets, ty and pytest. pycodekg and dockg index the repo for agents through .mcp.json; they are global tools, not dependencies. To build the docs site locally, run poetry install --with docs and mkdocs serve.

Docker

A prebuilt image for linux/amd64 and linux/arm64 is on Docker Hub:

docker run -d -p 7417:7417 -v satprint-tiles:/home/satprint/.cache/satprint egsuchanek/satprint

Then open http://localhost:7417. See Installation to build the image yourself.

License

MIT. See LICENSE.

Map data: Terrain Tiles © Mapzen and AWS Open Data (SRTM, ASTER GDEM, GMTED2010, ETOPO1, NED, EU-DEM and others). Imagery © Esri, Maxar, Earthstar Geographics. Street map, buildings, water and search © OpenStreetMap contributors, under the ODbL. Overture buildings © Overture Maps Foundation and OpenStreetMap contributors, under the ODbL.

Citation

If you use satprint in your research or project, please cite it:

Suchanek, E. G. (2026). satprint: Satellite Terrain to 3D-Printable Models (Version 0.2.1) [Software]. Flux-Frontiers. https://doi.org/10.5281/zenodo.23094151

@software{suchanek_satprint,
  author    = {Suchanek, Eric G.},
  title     = {{satprint}: Satellite Terrain to 3D-Printable Models},
  version   = {0.2.1},
  year      = {2026},
  publisher = {Flux-Frontiers},
  url       = {https://github.com/suchanek/satprint},
  doi       = {10.5281/zenodo.23094151},
}

The DOI is the Zenodo concept DOI, which always resolves to the newest archived release. The citation metadata is also in CITATION.cff. See the changelog for release history.