3D Scanning for 3D Printing: An Honest Maker's Guide

What 3D scanning for 3D printing actually needs

3D scanning for 3D printing has one requirement that most scanning tutorials skip: the printer needs a mesh, and the mesh has to be closed. A slicer works out what is inside and what is outside your model. If the surface has holes, open edges or faces that pass through each other, it cannot do that reliably.

The usual file formats are STL, OBJ and 3MF. STL is the oldest: it was developed at 3D Systems and first documented in 1988, and it stores nothing but triangles, with no standard way to carry color. 3MF is the newer format built for printing. It is an open, XML-based standard maintained by the 3MF Consortium and now published as ISO/IEC 25422:2025. Its specification requires printable model meshes to be manifold and free of holes, and the file records a unit (millimeter by default).

So when you 3D scan an object to print it, the real question is not which scan looks best on screen. It is which route gives you a closed, correctly sized mesh with the least repair work. There are three practical routes, plus one popular technique that does not get you there at all.

Route 1: structured-light and laser scanners

A dedicated scanner is the most direct path to a printable mesh. A structured-light scanner projects a calibrated pattern of light, usually lines, stripes or a grid, onto the object. The pattern bends as it falls across curves and edges, a camera records frame after frame of that distorted pattern, and the software reconstructs the surface from it. Laser scanners belong to the same family of dedicated hardware; the structured-light kind is the one we checked in detail here.

The scanner's own software then removes stray data, aligns the frames and builds the 3D model. Some scanners add a camera that records the object's color as a texture.

The weak spots are well documented by scanner makers themselves. Artec 3D, for example, lists very dark surfaces, highly reflective ones such as chrome, transparent or translucent materials, and hair or fur as difficult for structured light. The common fixes are changing the scanning angle and distance, or giving the surface a temporary matte coating: a light dusting of powder, or a matte scanning spray that vanishes on its own after a while.

Choose this route when the part has to fit something: replacement brackets, enclosures, adapters, anything with a tolerance.

Route 2: phone LiDAR and scanning apps

If you own an iPhone Pro or iPad Pro with a LiDAR Scanner, you already carry a depth sensor. LiDAR times infrared light pulses to measure distance, and scanning apps turn those readings into a mesh.

Two apps we could check on their makers' pages:

• Polycam runs on iOS, Android and the web. Its Basic plan page lists OBJ, FBX, DAE, USDZ and STL export, with GLTF available on every plan. In other words, STL export is a paid feature, so check the plan before you count on it.
• Scaniverse, from Niantic Spatial, generates both Gaussian splats and meshes, processed on the device.

Polycam describes its LiDAR scans as accurate to within about half an inch on standard interior captures. That number tells you what phone LiDAR is good at: furniture, large props, room-scale shapes. It is not the tool for a small gear or a threaded cap. For large, forgiving prints such as a scaled-down statue or a cosplay base form, it can be enough.

Route 3: photogrammetry from ordinary photos

Photogrammetry builds a model from overlapping photos. The software finds the same landmarks in many pictures, works out where each camera stood, and then produces a textured 3D model. Any camera works.

Options we verified:

• Apple's Object Capture in RealityKit, available in iOS 17 and later and macOS 12 and later. Apps built on it turn a series of photos into a 3D model.
• RealityScan from Epic Games, a desktop program that creates 3D models from images and laser scans. Epic states it is free for individuals and small businesses with under US$1 million in annual gross revenue. There is also a free RealityScan Mobile app.
• Meshroom, an open-source project released under the MPLv2 license. Its prebuilt downloads are for Windows and Linux and are built with CUDA, which means an NVIDIA graphics card.

One catch for printing is scale. Apple's documentation says that if the photos contain depth data, the real-world size can be calculated, and that without it you may have to scale the object yourself. Assume the same with any photo-only workflow: include a ruler or a known dimension in the scene and scale the mesh afterwards.

How to 3D scan an object: capture tips

Most of the guidance on how to 3D scan an object is the same whichever route you take. These points come from Apple's Object Capture documentation and match what scanner makers say:

• Pick a rigid object. Anything that bends or changes shape when you move it confuses the matching.
• Avoid subjects that are very thin, highly reflective, transparent or translucent. Single-color, very smooth surfaces also give the software too little to work with.
• Use soft, even light. On-camera flash, direct sun and bare bulbs cast hard shadows. Bounce the light off a wall or ceiling, or shoot outdoors on an overcast day.
• Overlap heavily. Apple recommends 70 percent or more between neighboring shots, and warns that below 50 percent the process may fail.
• Go around the object at several heights, or put it on a turntable in front of a plain, well-lit background and keep the camera on a tripod.
• Capture the underside. Standing the object on different sides between passes avoids gaps and holes.
• Fill the frame without cutting anything off, and lock exposure and focus so settings stay the same from shot to shot.

On an iPhone or iPad you can lock both by long-pressing the camera view until AE/AF Lock appears.

Cleanup: from raw scan to printable mesh

Almost no scan prints as it comes. Expect four jobs.

Close the surface. Holes appear wherever the scanner could not see, typically the base and deep recesses. The mesh has to be manifold before slicing.

Set the scale. Measure one dimension on the real object and compare it with the model.

Reduce the triangle count. Raw scans often carry far more triangles than a printer can reproduce, and heavy files slow every later step.

Flatten the base and orient the model so it sits on the print bed.

Tools we checked:

• Blender. Its 3D Print Toolbox extension checks for bad geometry and fixes it with Make Manifold, calculates volume and surface area, scales a model to set bounds or a target volume, aligns selected faces to the print bed, and exports STL, PLY or OBJ. The Decimate modifier reduces the vertex and face count with minimal shape changes.
• MeshLab. An open-source program, under the GPL 3.0 license, for processing and editing 3D triangular meshes. It is aimed at raw data from 3D digitization devices and at preparing models for 3D printing, with downloads for Windows, macOS and Linux.
• Meshmixer. A free Autodesk tool with hole filling, a convert-to-solid function for 3D printing, hollowing with escape holes, and mesh simplification. Autodesk states it is no longer in development and will not be supported, although there are no current plans to retire it. It still works, but do not expect updates.

Why a Gaussian splat is not a printable model

Gaussian splatting comes up in almost every scanning discussion now, so it is worth being plain about it: a splat cannot be sent to a slicer.

The method was introduced in a 2023 SIGGRAPH paper by Kerbl, Kopanas, Leimkühler and Drettakis. It represents a scene with a large number of 3D Gaussians, soft colored blobs, starting from the sparse points found while working out the camera positions. Those blobs are tuned so the scene looks right from every viewpoint. There are no triangles, no closed surface, and no inside or outside. A file can look convincing on screen and still contain nothing a printer can use.

A splat is still useful around a print project:

• Documenting the original object realistically before you modify, repair or replicate it.
• Checking capture coverage. If the splat is blurry or empty on one side, your photos missed that side, and a mesh built from the same photos will have a hole there too.
• Showing the finished print or the reference object to someone else in a way a flat photo cannot.

Where SplatLoom fits, and where it does not

SplatLoom is a desktop app for Mac and Windows that turns overlapping photos or a short video into a Gaussian splatting scene on your own computer, with no upload and no account. It exports a Gaussian PLY, a PNG still, an MP4 orbit video and a self-contained offline HTML viewer.

It has no mesh export and it is not a measurement or CAD tool, so it will not produce an STL for your printer. For a printable part, use one of the three routes above and clean the result in Blender, MeshLab or your scanner's software.

Where it fits is the documentation side: a realistic, shareable view of the object you scanned, or a quick check of whether your photo set covers the whole thing. It costs US$19.90 once, with no subscription.