Decimation that respects the silhouette
Zlim collapses redundant edges and quantizes vertex attributes within a tolerance you control, so a high-poly mesh becomes a lighter one without a visible change in shape. Typical results are 70–80% fewer faces at similar quality.
Because it is deterministic, the same model and tolerance always yield the same face count — which makes it safe to run in automated asset pipelines.
Why poly count alone is the wrong KPI
A model can have a low triangle count and still run badly if it has hundreds of draw calls or hundreds of megabytes of uncompressed textures. Zlim optimizes the whole budget: faces (decimation + quantization), draw calls (mesh and material merging), and texture memory (KTX2). Each is measured and reported per job.
zlim optimize model.glb -o out.glb -p maximum
# prints a JSON report of input vs. output: faces, vertices, bytesFAQ
How much can I reduce without it looking worse?
Most assets tolerate a 70–80% face reduction with no perceptible change because decimation is bounded by a geometric tolerance rather than a blind ratio.
Can I control the aggressiveness?
Yes — presets (fast, balanced, maximum) set the tolerance, and the result is deterministic at each setting.
Reduce your model's poly count
Start free — 25 optimizations a month. No card required.