Why optimize glTF you already have?
Being glTF does not mean being optimized. Raw exports from Blender, Maya, or three.js pipelines routinely ship uncompressed buffers, PNG textures, per-object meshes, and thousands of draw calls. The file loads — it just loads slowly and burns VRAM.
Zlim re-optimizes the asset in place: same look, far smaller footprint, measured per job.
The optimization budget: faces, draw calls, texture MB
Polygon count is one lever of three. Zlim optimizes all three: it quantizes and Draco/meshopt-compresses geometry (faces), merges compatible meshes and atlases materials (draw calls), and transcodes textures to KTX2 with mipmaps (texture MB). Redundant nodes are pruned and duplicate accessors de-duplicated.
zlim optimize scene.gltf -o scene.glb -p balanced
# prints a JSON report of input vs. output: faces, vertices, bytesFAQ
Does Zlim accept GLB as input, not just glTF?
Yes. Both .gltf (with external buffers/textures) and self-contained .glb are accepted and return an optimized .glb.
Meshopt or Draco?
Zlim can apply either geometry compression scheme. Draco maximizes size reduction; meshopt favors faster decode. The profile chooses sensible defaults.
Is the optimization lossless?
Quantization and decimation are tolerance-bounded and effectively lossless to the eye; node pruning and de-duplication are fully lossless. Every run is deterministic.
Optimize your model now
Start free — 25 optimizations a month. No card required.