MeshLab

What MeshLab does

MeshLab is a processing environment for triangle meshes and the point clouds that usually precede them. It is not a modelling program in the sense of sketching or extruding shapes; it is the place scanned or generated geometry goes to be cleaned up, checked, simplified and converted. The filter catalogue is the heart of the application and covers the operations a scanning pipeline needs: removing duplicate and unreferenced vertices, reorienting face normals coherently, closing holes, filling small gaps, smoothing noise, decimating millions of triangles down to a printable count without destroying the silhouette, and reconstructing a surface from an oriented point cloud using Poisson or ball-pivoting methods.

How a typical pipeline runs

Work begins with an import, and the supported format list is long enough that almost any scanner or photogrammetry output opens without conversion: OBJ, PLY, STL, OFF, 3DS, X3D, COLLADA, glTF and many more. The first pass is diagnostic. You look at statistics, check for non-manifold edges and boundary loops, and confirm that normals point outward, because every later filter depends on that being true. Cleaning follows: duplicate vertices are merged, isolated components that belong to background clutter are deleted, and the mesh is cropped to the region of interest.

Repair comes next. Hole filling closes missing patches left by occlusion or reflective surfaces, and the quality settings decide whether the patch is a simple fan or a smooth reconstruction. Decimation then reduces the triangle budget, either to a target count or a target percentage, with quality preservation that keeps curvature where it matters. If the model came from photogrammetry, texture handling is a separate stage: the mesh can be parametrised and retextured, or existing textures can be projected from registered rasters. Measurement and inspection use the distance tools, which compute per-vertex deviation between two meshes and report statistics including the maximum and mean error, the standard technique for comparing a scan against a reference model. Finally the result is exported in the format the next tool expects.

Settings, performance and limits

The interface exposes a large number of filters with many parameters each, and most dialogs have live preview, which is essential because the correct setting depends entirely on the data. Undo is available but not unlimited, and heavy filters on huge meshes can take minutes; saving intermediate states as separate files is the practical habit. The application is memory bound and a 64-bit build is essential for anything serious. It has no parametric history, no constraints, no assemblies and no toleranced geometry, so it cannot produce a manufacturable drawing. Automatic reconstruction also produces dense triangle soup rather than clean quad topology, which matters if the model must later be rigged, subdivided or edited by hand.

Who should choose something else

If you need to design a part rather than clean one, use a parametric CAD program or a polygon modeller with proper editing tools. If your data is predominantly registered point clouds and the questions are about alignment, sections and volumes, a dedicated cloud processing package is more direct. And if your whole workflow is 3D printing, a slicer with repair functions may fix the small problems without a separate tool. MeshLab is the right choice when geometry arrives from a scanner and needs to become usable.