GLB CompressorNEW
Compress GLB models for lighter and faster web previews
Checking your access…
The slowest part of putting a 3D model on a web page is rarely the GPU — it is the file size. A single 4K texture can run to several megabytes, and a model with tens of thousands of triangles plus a pile of uncompressed PNGs multiplies the cost of every download and every load. The AllFreeTool GLB Compressor exists for exactly this job: it shrinks a .glb right inside your browser and still produces a standard glTF 2.0 single file that any engine or viewer can open.
The pipeline runs in five steps. Duplicate vertices with identical coordinates are welded together, triangles are reduced with uniform-grid clustering, vertex attributes are rewritten as fixed-point integers (POSITION, NORMAL and TEXCOORD go through the KHR_mesh_quantization extension), textures are re-encoded (opaque images become JPEG, transparent ones stay PNG), and finally any material or texture that no mesh references is dropped. Animation tracks and skinning data can be removed on top of that, turning the model into a static mesh.
Why use GLB Compressor on AllFreeTool?
Weld and Decimate
Duplicate vertices are merged, then uniform-grid clustering cuts the triangle count while keeping the silhouette.
Fixed-Point Vertices
Coordinates become 16-bit or 8-bit integers, with normals and UVs quantized to match, shrinking vertex data dramatically.
Texture Re-Encoding
Opaque textures become JPEG, transparent ones stay PNG, with a maximum edge length and a quality dial.
Runs Locally
The model is processed in browser memory, never uploaded, and the compressed file is ready to download at once.
How to use GLB Compressor
Choose a GLB
Click the drop area or drag a .glb file in; the tool lists its meshes, faces, vertices and textures first.
Tune the Options
Set the mesh detail ratio, the quantization bit width, texture compression, and whether to drop animation and skinning.
Compress and Download
Press compress, compare the sizes, and download the resulting .min.glb file.
Common questions
Does compressing change how the model looks?＋
Only very slightly. Welding does not alter the shape at all. Quantization introduces floating-point-scale error — at 16 bits that is roughly 1/65535, which is invisible to the eye — while grid clustering pulls neighbouring vertices towards their cell centre, so the silhouette contracts by a fraction of a percent and shrinks more as you lower the detail ratio. If the model must look identical, set mesh detail to 100% and turn quantization off; you still keep the savings from welding and texture re-encoding.
Can rigged or animated models be compressed?＋
They load and compress normally, but to avoid breaking the skinning transform, skinned meshes are automatically skipped for vertex quantization: skinning matrices act in node space, so moving a mesh into a scaled child node would detach it from the rig. Welding and decimation still run safely, because bone weights are remapped together with their vertices. If you only need a static preview, switch on "Remove skins" to turn the model into a static mesh, which lets the whole pipeline work on it; animation tracks are dropped the same way with "Remove animations".
Will the compressed file still open in Blender, Unity or three.js?＋
Yes. The output follows the glTF 2.0 specification and the vertex quantization uses the standard KHR_mesh_quantization extension, so engines and mainstream viewers support it. The result is still a single .glb with geometry, materials and textures inside.
How much size can I expect to save?＋
It depends on what dominates the file. Texture-heavy models benefit most and often halve or better after re-encoding; geometry-only meshes save through welding and quantization, typically somewhere between 30% and 70%. The result panel reports the original size, the compressed size and the percentage saved.