Tiny Online Tools
Turning a model back into points
Point-cloud software speaks points. If what you have is a mesh — a CAD export, a game asset, a scan someone already meshed — you need to sample it first. This free online tool does that in the browser: load a mesh, choose a density, download PLY or XYZ.
What it is useful for
- Test data for registration, segmentation or reconstruction code, with known ground truth
- ICP and alignment pipelines that only accept point clouds
- Simulating a scan of an object you already have as a model
- Feeding LiDAR-oriented tools with something they can open
Uniform by area, or vertices only
Uniform by area is the default and almost always the right one. Each triangle receives a share of the sample budget proportional to its area, and samples are placed inside the triangle using barycentric coordinates with a square-root correction that keeps them from bunching at one corner. The result has even spacing over the whole surface, whether the model is a dense organic sculpt or a flat wall made of two triangles.
Vertices only takes the mesh's own vertices and nothing else. It is instant and exact, but the density follows the modelling, not the geometry: a smooth curve may carry thousands of vertices while a large flat panel carries four.
Sampling is deterministic. The random numbers come from a seeded generator, so running the tool twice with the same settings gives you the same cloud — which is what makes two exports comparable at all.
Colours and normals
Where the mesh has vertex colours, they are interpolated across each triangle with the same weights as the position, so the cloud keeps the model's shading. Where it does not, the material's base colour is used. Texture maps are not sampled — bake a texture down to vertex colours first if you need it.
Surface normals are optional and are written into PLY files. An oriented point cloud is what Poisson-style reconstruction in other packages asks for, and normals sampled from a real surface beat any estimated afterwards. XYZ has no column for them, so the tool warns instead of dropping them silently.
Reading the density figures
The result panel reports surface area, point count, density in points per square unit, and mean spacing — the square root of area divided by points. Spacing is the number that matters: it is directly comparable to the resolution of a real scanner, and tells you at a glance whether the cloud will support the feature size you care about.
To thin, crop or denoise the result afterwards, use the Point Cloud Processor; the Point Cloud to Mesh tool goes back the other way.