Quady3D

GLB Compressor

Compress geometry and textures. Compare the result, then download.

Free · No account needed

Your model

100 MB max
Compression

Draco + WebP · up to 1024 px · all triangles kept

Advanced settings

Compressed formats need compatible viewers. Quantization and texture encoding are lossy. Existing KTX2 images are kept.

Files stay on your device

Sample preview
Original color/alpha, normal and packed ORM maps. KTX2 keeps their color spaces distinct.
  1. Open your GLB

    Choose a file or try a sample.

  2. Choose a preset

    Start balanced, then adjust if needed.

  3. Preview & download

    Check the actual output before saving.

Try a real compression example

Original files with ready-to-inspect outputs.

Original textured ceramic sample
Textured ceramic296 KiB → 30 KiB
3,980 triangles kept · Draco + KTX2
Source GLB ↓Compressed GLB ↓
Original color, normal & orm sample
Color, normal & ORM1.18 MiB → 285 KiB
sRGB color · linear normal and data maps
Source GLB ↓Compressed GLB ↓
Format support & compatibility
Supported behavior and limits
Default cleanupDeduplicate, weld identical vertices and prune unused data
TexturesWebP or KTX2 ETC1S / UASTC, with mipmaps; KTX2 encoding capped at 2048 px
MeshoptCompressed geometry and animation; decoder support required
DracoGeometry encoding and decoding; skins and animation clips retained
Triangle reductionNo decimation; triangle count is retained

Original Quady3D AI examples · digital renders. Files and license.

Troubleshooting

The output is larger

Some assets are already compact. Keep the original or try a smaller texture size. No fixed percentage is guaranteed.

The receiver cannot load it

Choose cleanup only and keep original images for the broadest compatibility. Draco, Meshopt, KTX2 and WebP each require the corresponding support in the destination.

The source already uses Draco

Draco input is decoded locally and can be recompressed with Draco, Meshopt or cleanup only. Existing KTX2 textures are kept rather than resized.

Questions

How do I reduce GLB file size?

Open the file, choose cleanup, Draco or Meshopt, and optional WebP or KTX2 textures, then compare the actual downloaded size. Processing happens on your device and does not use generation credits.

Is GLB compression lossless?

Default deduplication and identical-vertex welding preserve the rendered surface. WebP and KTX2 encoding are lossy. Draco and Meshopt use quantization, which may slightly change coordinates. Check your destination and visual tolerance.

Does the compressor reduce polygons?

No. It optimizes storage, images and redundant data without a decimation pass. For a different face layout, use the retopology guide. The output can be larger than the input when little redundant data exists.