Point Cloud to Mesh Reconstruction
CloudProReconstruct a mesh from LAS/LAZ/PLY/PCD/XYZ point clouds with server-side Poisson (Open3D); export PLY/STL/OBJ. Sign-in and cloud quota apply.
Sign in to use this tool
This tool runs on Vecaria's processing service and requires an account. Sign in and you'll return to this page to continue.
Sign inDrag files here, or click to browse
Supports LAS, LAZ, PLY, PCD, XYZ
Choose fileSign-in required; cloud tasks draw from your cloud credits by weight (see pricing for free/Pro rules).
After submission this page polls the job every 2 seconds for up to about 11 minutes; if it times out, please resubmit (the server allows up to 10 minutes per job).
Want to inspect the cloud first? Use the LAZ Viewer, LAS Viewer or COPC Viewer (local, nothing uploaded).
What is it? Point Cloud to Mesh Reconstruction builds a triangle mesh from LAS/LAZ/PLY/PCD/XYZ point clouds using Poisson surface reconstruction on the server, exporting PLY/STL/OBJ. Unlike a format conversion, reconstruction infers a brand-new surface from the points (with oriented normals), so the output detail depends on depth and point density rather than on the input file.
TL;DR: Upload a LAS, LAZ, PLY, PCD or XYZ point cloud, set the reconstruction depth, and download a Poisson-reconstructed mesh as PLY, STL or OBJ — nothing to install, the heavy compute runs on Open3D in the Vecaria cloud. Requires sign-in and consumes cloud quota; watertightness depends on point density and normals.
How to use
- Sign in and upload the point cloud (LAS, LAZ, PLY, PCD or XYZ).
- Set the reconstruction depth (8 is a balanced default), tick "Estimate normals" if the file has none, and pick the output format (PLY, STL or OBJ).
- Wait for the job and download the mesh — large clouds are voxel-downsampled first, so very detailed settings take longer.
FAQ
Which formats can I export?
PLY, STL and OBJ are available.
When are files deleted?
Cloud files follow the configured retention window.
How do I choose the depth?
5–6 gives a quick coarse mesh, 8 is a balanced default, 10–12 is finest but slower and uses more quota — start at 8 and adjust.
Why does it need normals?
Poisson reconstruction needs oriented normals. Tick "Estimate normals" and we compute them; PLY inputs with normals are used as-is.
Why is it slower than plain conversion?
Reconstruction is server-side heavy compute (Open3D Poisson) — higher depth and point caps take longer, typically minutes rather than seconds; the page keeps polling for up to about 11 minutes before timing out (server-side processing is capped at 10 minutes).
What about very large clouds and normal orientation?
Clouds above the max-points cap are voxel-downsampled first. Normal orientation runs only on clouds of 300,000 points or fewer — larger clouds skip that step (normals keep their input or estimated directions), so Poisson output may have locally flipped faces in extreme cases. Pre-orient large files in a desktop tool or lower the point cap.
Is the reconstructed mesh watertight?
Poisson reconstruction solves for a closed implicit surface, so in good conditions the output is watertight — but in practice it depends on point density and normal quality: sparse or noisy regions, misoriented normals and stray far-from-surface points can produce holes or artifacts. Start at depth 8 and inspect the result; use the Mesh Doctor for STL tool to check and repair holes afterwards.
How is reconstruction different from a plain format conversion?
A conversion keeps existing geometry and only rewrites the file. Reconstruction infers a new triangle surface from the points themselves, so it needs oriented normals (tick "Estimate normals" and we compute them), and detail depends on the depth setting and point density. That is why it runs server-side on Open3D and takes minutes rather than seconds.
How does it compare to MeshLab or CloudCompare?
Those desktop tools are free and unlimited, but you must install them, learn the filter chain and have a machine with enough RAM for million-point clouds. This tool runs in your browser tab with nothing to install — the Poisson reconstruction happens on our servers, which is why it needs an account and cloud quota. If you already do reconstruction in MeshLab, keep it; if you want a mesh from a scan without setting up a toolchain, start here.
Is it free?
Cloud reconstruction consumes cloud credits per job and requires sign-in, because the compute runs on our servers (each plan tier includes a monthly credit allowance; see the pricing page). Viewing stays free and local — the LAS/LAZ/COPC viewers open files in your browser with no upload and no account. Exporting points or filtering a scan in the browser is a Pro feature.
Examples
- Upload a .ply/.laz point cloud → set depth → download a PLY/STL mesh
Supported formats
Limitations
- Requires an account and cloud quota.
- Open3D (MIT) Poisson reconstruction; higher depth means finer detail but slower runs.
- Suited to oriented surface scans; watertightness depends on point density and normals.
- Normal orientation runs only on point clouds of 300,000 points or fewer; larger clouds skip it silently (input/estimated directions are kept) and Poisson output may show locally inverted faces in extreme cases — pre-orient in a desktop tool or lower the point cap.
- Large clouds are voxel-downsampled to the max-points cap before reconstruction.
Privacy
This tool has no local processing path: your file is uploaded over an encrypted connection to our own VPS, used only for this job, and auto-deleted within 24 hours after the task — or delete it earlier at any time.
Related tools
Guide
Standards & comparison
| Dimension | Vecaria (this tool) | Online converters (cloud) | Desktop GIS / BIM suites |
|---|---|---|---|
| Privacy | Cloud tool: upload required, encrypted transfer, temporary storage | You upload the file to a third-party server | Runs locally |
| Waiting time | Server job queue | Upload plus cloud queue | Local application startup |
| Standards support | ASPRS LAS 1.0–1.4 | Varies by service, often older subsets | Full vendor protocol sets |
| Install footprint | 0 MB — runs in the browser | 0 MB | Several GB of storage |