Preset
Scene Cleanup
▼Prune unused
Dedupe
Weld vertices
Flatten hierarchy
Join meshes
Merges compatible primitives → fewer draw calls.
Instance repeats
Geometry
▼Simplify mesh
Keep50%
Error limit0.005
Lock borders
Quantize
Bit-depth applied by the active compressor below.
Position14b
Normal10b
Texcoord12b
Color8b
Mesh Compression
▼Levelrecommended
EXT_meshopt_compression — fast GPU-side decode, supported by three.js, model-viewer & Babylon.
Method
Encode effort5
KHR_draco_mesh_compression — smallest geometry, heavier decode.
Textures
▼Resize
Max dimension
Power-of-two
Format
Quality80
Mode
Cuts GPU VRAM, not just file size. Basis encode is slower — runs only when you press Optimise.
Advanced
▼Resample animation
Sparse accessors
Reorder (vtx cache)
Optimises index order for GPU vertex cache.
● settings changed
ORIGINAL
OPTIMISED
⇆
◈
Drop a GLB or glTF
Compare original vs optimised with a side-by-side wipe.
Everything runs locally — nothing leaves your machine.
.glb · embedded .gltf · drag & drop anywhere · demo loads a sample model
Comparison
Load a model, set options,
then press Optimise.
then press Optimise.
Inspector
▼No model loaded.
Export
Filename
⚠ KTX2 encoding didn't complete for this model — textures were kept in their prior format. Try WebP, or check the console.