Turning a 3D Scan into a Printable or Editable Model

A raw 3D scan is the messiest file in 3D work: a few million triangles of noisy surface, holes where the scanner could not see, debris floating where reflections confused it, and no bottom, because the object was sitting on a table. Between that file and something you can print or open in CAD sits a pipeline of three steps, each fixing one class of problem. This guide walks the pipeline in order, with honest expectations about the last step, because scan-to-CAD is where marketing and reality diverge most.

What You Are Starting With

Scanner output, whether from a $300 handheld, photogrammetry, or a phone app, shares the same pathology. The mesh is enormously over-tessellated: scanners sample densely everywhere, so a coffee-cup-sized object arrives as two million triangles where fifty thousand would carry the same shape. It is open wherever line-of-sight failed: undersides, deep recesses, dark or shiny patches. And it contains debris: disconnected fragments of background, turntable, or noise.

Step 1: Repair — Close It and Clean It

Repair comes first because every later step assumes a closed solid: decimation preserves watertightness but does not create it, slicers need it, and mesh-to-BREP conversion refuses without it. Run the scan through the repair tool: hole filling closes the unscanned regions, and shell handling drops the floating debris (the default keeps shells above 1% of the surface area, which is tuned for exactly this scan-noise case). Expect the filled regions to be flat caps rather than the surface the scanner missed; repair closes holes, it does not imagine geometry. If the missing region matters, rescan from another angle instead.

Step 2: Simplify — Lose 90% of the Triangles

Scans are the best case for decimation that exists. Because the mesh is uniformly over-dense, quadric decimation removes 80 to 95 percent of the triangles with no visible change: flat and gently curved regions give up their redundant triangles first and detail survives. The payoff is practical: a 400 MB scan becomes a 20 MB file your slicer opens without choking, uploads finish in seconds, and any later CAD conversion has far less work to do. Start at 90% reduction on a raw scan and look at the result; you will usually be unable to find the difference.

Step 3a: Print It

After repair and decimation the scan is an ordinary printable mesh. Two scan-specific things are still worth checking before slicing. Scale: scans frequently import at the wrong size, so measure a known dimension against the real object and scale accordingly. And thin regions: scanned walls, especially on hollow objects, can be paper-thin where the scanner reconstructed both sides close together; a wall-thickness check tells you before the printer does.

Step 3b: Convert to CAD — Honest Expectations

Here is the part most guides soften: a scan of an organic object converts to a STEP file, but not to useful CAD geometry. The conversion merges flat regions into real faces, and an organic surface has none, so the output is a solid stitched from thousands of tiny triangular faces (when the conversion finishes at all; dense scans often exceed the timeout): it opens in SolidWorks, it can be measured and machined against, but it cannot be meaningfully edited. Where conversion genuinely works is scanned mechanical parts: brackets, housings, machined shapes, where the underlying geometry really is flat-dominated and the merge has something to find. For true reverse engineering of organic shapes, the real path is fitting surfaces semi-manually in dedicated software, or remodeling over the scan as a reference.