How to reduce a point cloud file.
Start with the region and detail your next task needs. Keep the original scan, then create a smaller working copy.
1. Decide whether you need the whole scan
If you only need a roof, room or small site area, crop first. Cropping removes points outside an axis-aligned box while keeping the points inside it. In the sample pavilion, setting Minimum Z to 103 isolates the higher part of the scan. That number belongs to the sample coordinates; use your own bounds for a different file.
2. Choose a point budget or a spacing
Random subsampling keeps a requested number of points across the selected cloud. Reducing 1,000,000 points to 100,000 removes 90% of the points, but does not guarantee a 90% smaller file. Spatial subsampling instead keeps points separated by a minimum distance. Its output count depends on the scan distribution. Neither method reconstructs surfaces or fills holes.
3. Read spacing in the source units
The site does not infer meters or convert units. A spacing of 0.1 means 10 centimeters only when your coordinates are in meters. Check the source documentation and the displayed X/Y/Z extent before choosing a value. A spacing suitable for a building may erase a small scanned object.
4. Compare the result before downloading
Use Before and After to switch between the operation input and result from the same viewpoint. Check the total point-count change and inspect edges, thin structures and sparse areas. The preview contains at most 80,000 sampled points, so similar-looking previews do not prove that all detail survived. Download and inspect the working file in your downstream application when that detail matters.
5. Check the actual output size
The File size field reports the selected result. A PLY can be larger than the original compressed LAZ even after points are removed. XYZ writes coordinates as text and is not a compression option. Choose the format your next application accepts, compare the reported bytes, and keep the source if you need its original metadata or compression.
| Task | Start with | Check afterward |
|---|---|---|
| Only one region is needed | Crop | Bounds and retained objects |
| A fixed point-count budget | Random subsampling | Thin and sparse features |
| More even point spacing | Spatial subsampling | Source units and resulting count |
| A small region with too many points | Crop, then subsample | Both region coverage and detail |
Common questions
Does reduction overwrite the upload?
No. Each operation creates a separate result in the temporary session. Select the original again to try another setting. Download the original and any results you need before the session expires.
What if the file is too large to upload?
This site accepts up to 100 MB and 5,000,000 points per cloud. You need a local tool to reduce a file beyond those limits before uploading. A small LAZ file can still exceed the point limit after decompression.
Can I recover removed points from the result?
No. Return to the original or an earlier result. Subsampling a reduced result again cannot restore points already removed.