gotiler
gotiler is a fast, lightweight and portable converter to transform massive point clouds or Gaussian splat datasets into OGC 3D Tiles.
Jastrzębia Góra, Poland, 104M Gaussian splats3D scanning data created and provided by Andrii Shramko, Teleportour · teleportour.com
Features
Turn massive scans into streaming 3D Tiles in seconds.
Built for huge datasets and simple workflows: point gotiler at a point cloud, a folder of them or a Gaussian splat scene, and get a tileset that CesiumJS and other 3D Tiles viewers can stream.
Fast and out-of-core
Uses every CPU core and fast NVMe drives, and tiles billion-point clouds and scenes of hundreds of millions of splats without running out of RAM.
Self-contained
One executable and its bundled data folder, for Linux, macOS and Windows. Nothing else to install to run it.
Point cloud tiling
LAS, LAZ, E57 and PLY point clouds into compressed 3D Tiles 1.0 or 1.1, reprojected, colored and with their attributes.
Gaussian splat tiling
Trained 3D Gaussian splat scenes into 3D Tiles 1.1 with levels of detail, for CesiumJS to stream.
Point cloud tiling
Fast, feature-rich tiler for billion-point clouds.
Reproject, colorize and compress your LAS, LAZ, E57 and PLY point clouds, all in one self-contained tool.
LAS, LAZ, E57 and PLY
Reads LAS, LAZ, E57 and ASCII or binary PLY inputs.
Compressed by default
Compresses 3D Tiles 1.1 point clouds with meshopt compression and quantization, cutting tileset size by up to 70% or more.
Automatic reprojection
Reads the CRS from LAS GeoTIFF or WKT metadata and transforms coordinates with the embedded PROJ library.
3D Tiles 1.0 and 1.1
3D Tiles 1.0 and 1.1, optionally packed into a single .3tz archive.
Per-point attributes
Export intensity, classification, GPS time, LAS extra bytes or any other attribute of the input.
Colorization
Colorize point clouds from their attributes with 36 embedded ramps, or directly from an RGB GeoTIFF orthophoto. No extra tooling required.
Gaussian splat tiling
3D Gaussian splat tiler for high-fidelity visualization.
Transform your massive PLY Gaussian splat datasets into georeferenced, compressed and streamable 3D Tiles with hierarchical levels of detail.
Cutting edge algorithms
Builds every level of detail without retraining: splats merge in pairs as in NanoGS, chosen by Runnalls' closed-form bound for Gaussian mixture reduction.
PLY input format
Reads ASCII or binary PLY files of 3D Gaussian splats, as trainers and editors export them, with spherical harmonics up to degree 3.
SPZ compression
Writes the splats of every tile as SPZ-compressed glTF, with the KHR_gaussian_splatting extensions that CesiumJS reads.
Configurable SH and opacity threshold
Choose the spherical-harmonic degree to keep and the opacity below which splats are dropped, to stream lighter tilesets.
Benchmarks
313 million points in under 2 minutes.
Up to 9.1× faster than other open-source tilers, with a fraction of their memory.
The Helsinki point cloud, 313M points in a 13 GB LAS file, tiled on the same AWS EC2 i4i.2xlarge instance running Ubuntu 24.04. Reproduce the runs with scripts/benchmark.sh.
259 million Gaussian splats in under 6 minutes.
- Splats
- 258,951,032
- Input PLY
- 16.3 GB
- Execution time
- 5m 50s
The Lublin, Poland scene, tiled by gotiler splats with its default options on a desktop with an Intel Core i5-13600K (14 cores, 20 threads), 32 GB of RAM and a Samsung 980 PRO NVMe SSD, running Windows 11. 3D scanning data created and provided by Andrii Shramko, Teleportour · teleportour.com.
Quick start
One line to install, one to tile.
No dependencies and no administrator rights: the install script fetches the release for your platform, checks it and puts gotiler on your PATH.
-
Install gotiler with the command for your platform, or download an archive from the latest release.
Terminal window curl -fsSL https://github.com/mfbonfigli/gotiler/releases/latest/download/install.sh | shTerminal window irm https://github.com/mfbonfigli/gotiler/releases/latest/download/install.ps1 | iex -
Tile a file or a whole folder. Pass
--previewto see the output right away in a local CesiumJS viewer and fine-tune its position.Terminal window # Tile a LAS, LAZ, E57 or PLY file into ./outgotiler pcloud -o ./out ./input.las# Tile every file of a folder into its own tilesetgotiler pcloud -o ./out ./las_folder# Merge a folder into one tileset, packed as a .3tz archivegotiler pcloud -o ./out --join --3tz ./las_folderTerminal window # Tile a Gaussian splat scene, then place it on the globegotiler splats -o ./out ./scene.ply# A scene trained with OpenCV axes, placed in Romegotiler splats -o ./out --input-up-axis -y --longitude 12.4924 --latitude 41.8902 ./scene.plyTerminal window # Set the input CRS, here with a vertical datumgotiler pcloud -o ./out -c EPSG:32633+3855 ./input.lasTerminal window # Color points by elevationgotiler pcloud -o ./out --colorize z:viridis ./input.las# Color by intensity on a reversed, 8-band turbo rampgotiler pcloud -o ./out --colorize intensity:turbo:reverse:steps=8 ./input.lasTerminal window # Tile, then check the result on a globegotiler pcloud -o ./out --preview ./input.las# Tile a Y-up model without a CRS, then move it into placegotiler pcloud -o ./out --crs local --input-up-axis y --preview ./model.ply# Open an existing tilesetgotiler view ./out
gotiler view
A 3D globe viewer, built in.
Check any tileset on a local CesiumJS globe, and drag ungeoreferenced models into place.
- Search an address, or click Pick position and then the globe, to move the model there.
- Drag the gizmo’s arrows and rings, or type offsets and rotations.
- Save position stores the new position in
tileset.json, without rewriting the tiles. - The viewer only listens on
127.0.0.1, behind a random access token.
gotiler view ./out
Point cloud: City of Melbourne, CC BY 4.0 · Imagery: VersaTiles & OpenStreetMap contributors
Go library
Embed the tiling engine.
The tiler, readers, encoders and plugins are importable packages of the
github.com/mfbonfigli/gotiler/v3 module, with plugins for E57 and PLY input, meshopt
compression, subsampling, orthophoto colorization, 3TZ archives and S3 uploads.
func main() { t, err := tiler.NewGoTiler() if err != nil { log.Fatal(err) } opts := tiler.NewDefaultTilerOptions() ctx := context.Background() err = t.ProcessFiles([]string{"scan.las"}, "./tiles", "", opts, ctx) if err != nil { log.Fatal(err) }}Open source
gotiler is distributed under the GNU AGPLv3. Release archives ship the third-party licenses for their platform.
Read the licenseAlternative licensing
If the AGPLv3 does not fit, for example to embed the engine in proprietary software or services, contact the maintainer for alternative licenses.
Contact the maintainerSupport the project
gotiler is maintained with love. If it saves you time or resources, a donation or a star keeps it going.
Make a donation


