Back

Draco Compression Settings

Tweak Draco parameters for the smallest meshes without visible artefacts.

dracocompressionmesh
Updated 2025-06-29

Draco encodes mesh geometry using predictive coding + entropy compression. Typical savings are 80-95 % on raw vertex data.

How Draco works (TL;DR)

  • Quantisation – Converts floats → ints within a bit budget (8-16 bits).
  • Connectivity encode – Reorders triangles (Edgebreaker) to exploit locality.
  • Entropy encode – Final pass with LZ77 + arithmetic coding.

Read Google's excellent overview on their Open Source Blog.

Key Parameters

ParameterRangeEffectTypical
Compression level1-10Higher spends more CPU, ~5-20 % extra savings8
Quantisation bits8-16Lower → smaller files but visible stair-stepping14

Choosing Values

  • Static props – Go aggressive: level 9, quant 10.
  • Hero assets – Moderate: level 7, quant 14.
  • Architectural CAD – Precision matters: level 6, quant 16.

Tips & Gotchas

  • Draco does not support morphTargets; keep originals if you need shape keys.
  • Game engines need a decoder. three.js DRACOLoader auto-loads WASM < 300 KB.
  • Precision loss is most visible on hard-edged mechanical parts. Visual check in wireframe mode.

Related Articles

Geometry Transform Passes

Deep dive into geometry transforms that prepare your model for compression.

Texture Compression Techniques

Get huge savings by converting, resizing and re-encoding textures.