Smooth STL Files Online — Remove Surface Noise for 3D Printing
3D scans and high-resolution sculpts often have surface noise — tiny bumps and irregularities that make the mesh look rough even though the underlying shape is correct.
Last updated Mar 2026
Usually under 5 seconds — depends on mesh size and iterations.
When to Smooth STL Files
Clean Up 3D Scan Artifacts
Structured-light and photogrammetry scanners produce meshes with sensor noise. Smoothing 3-5 iterations removes the granular texture while preserving the scanned shape for accurate measurement or printing.
Prepare Sculpts for 3D Printing
ZBrush and Blender sculpts exported at high subdivision levels carry tool marks and brush artifacts. A quick smooth pass cleans the surface without losing the sculpted form.
Reduce Slicer Artifacts
Rough surfaces create tiny overhangs that confuse slicers, producing unnecessary support structures. Smoothing the mesh before slicing reduces support material and print time.
Improve Visual Quality for Display
STL models used for visual presentation (portfolio, client review) look more professional after a light smooth pass removes distracting surface noise.
How It Works
Taubin smoothing (published by Gabriel Taubin at SIGGRAPH 1995) is a two-step Laplacian filter designed specifically to avoid the shrinkage problem that makes simple smoothing unusable for real meshes.
Each iteration has two passes: first a positive λ (lambda) step that moves each vertex toward the average of its neighbors (smoothing), then a negative μ (mu) step that moves vertices slightly back outward (inflation). The net effect removes high-frequency noise while preserving the low-frequency shape of the model.
You control two parameters: Iterations (how many times to repeat the smooth/inflate cycle — more iterations = smoother) and Strength (λ factor — higher values smooth more aggressively per iteration). The inflate factor μ is computed automatically to prevent shrinkage.
For 3D printing: start with 3 iterations and 0.5 strength. If the surface is still rough, increase to 5-8 iterations. For photogrammetry scans with heavy noise, try 10-15 iterations. The entire process runs in your browser using pure JavaScript — no server, no WASM dependency for the smoothing step itself.