Free tool by sizeless
Ship 3D models that load in milliseconds
sizeless GLB Compressor shrinks GLB meshes and PLY/LAS point clouds by up to 95% — meshopt compression, material dedup and voxel downsampling, processed on our servers. No installs, no accounts, free for professional use.
- Typical size reduction
- 70–95%
- Largest supported upload
- 10 GB
- Cost to use
- Free
Features
Everything our pipeline does, in your browser
The same compression stack we run on survey-scale digital twins — now available as a free web tool.
meshopt compression
gltfpack with four tuned levels — from fast basic quantization to max with 25% geometry simplification.
Material dedup
CAD exports often carry thousands of identical materials. We collapse them before compression, automatically.
Color baking & merge
Bake per-primitive material colors into vertex colors and merge thousands of primitives into one draw call.
PLY & LAS point clouds
Binary/ASCII PLY and LAS point clouds converted straight to GLB point primitives — with RGB colors preserved where present.
Voxel downsampling
Reduce massive scans with voxel-grid sampling, fixed decimation ratios, or a simple target point count.
Quantization control
Dial position and color precision per export — trade visual fidelity for size, bit by bit.
How it works
Three steps. No software to install.
- 01
Upload
Drop a .glb mesh or a .ply/.las point cloud up to 10 GB. One quick professional profile the first time.
- 02
Tune
Pick a compression level, enable color baking for CAD exports, or set voxel size for point clouds.
- 03
Download
Get a web-ready compressed GLB with full stats. Source and result are wiped from our servers within 3 hours.
Your data
Built for professionals, transparent by default
We ask who you are and what you're compressing for — once. That tells us which workflows to invest in. Your models are never used for anything except your own compression job.
Uploads and results are deleted from our servers within 3 hours.
Your profile is stored securely and never sold or shared.
Processing runs entirely on our infrastructure — your model never touches a third-party service.
One upload. A fraction of the size.
Try it with your heaviest model — the result usually speaks for itself.