Can I use a phone 3D scan for a part that has to fit something?

Usually not, and the reason is resolution rather than effort. A phone depth sensor is built for room-scale AR and is specified in centimetres; photogrammetry from phone photos does better on shape but recovers no absolute size at all. Either way the point spacing is coarser than the features that decide a fit — a 0.5mm chamfer, a bore diameter, a thread. The scan can be an accurate record of the overall shape and still be useless for a mating surface. Use it as a reference to model against, measure the fitting features with calipers, and build those in CAD.

Does the LiDAR sensor in a phone or tablet make it a real 3D scanner?

It makes it a real depth camera, which is not the same instrument. Phone time-of-flight sensors were specified for AR occlusion, room mapping and autofocus assistance — tasks where centimetre-class depth over several metres is the right trade. Manufacturers do not publish part-metrology accuracy figures for them because that is not the design target. A dedicated scanner solves a different optical problem: a calibrated camera and projector on a rigid known baseline, working over a short fixed distance. That is where the 0.05–0.1mm figures on consumer handhelds come from, and it is not something firmware can add to a phone.

Why does my phone scan come out the wrong size?

Because photogrammetry reconstructs shape from photographs only up to an unknown scale factor — the geometry is recoverable, the size is not. Apps that use the depth sensor do carry a real scale, but with the sensor's own error on it, so a 120mm part can land several millimetres out with nothing warning you. The fix is the same for both: measure one real feature across its longest span with calipers, scale the mesh to that number, then check a second independent feature. If the second one is off, the error is not uniform and scaling cannot fix it.

Will scanning spray and markers improve a phone scan?

Yes, and they are the cheapest improvement available. Photogrammetry needs visible surface detail to match points between photos, so a plain matte face is nearly as hostile as a shiny one — a light matte coat plus scattered marker dots gives the solver features to lock onto. Depth sensors benefit too, because glossy surfaces reflect the emitted light away and matte black absorbs it. Keep the coat light and even: a heavy wet layer pools in corners and is captured as real geometry, which swaps a gap for a wrong surface.

Is photogrammetry or the depth sensor better for small parts?

Photogrammetry, with caveats. On an object small enough to fill the frame, photographs carry far more lateral detail than a phone depth map does, so shape and surface markings come out better — but the result has no inherent size and needs scaling. The depth sensor wins on large, low-detail, texture-poor subjects like rooms and furniture, where its scale is useful and its coarseness does not matter. Below roughly a hand-sized object with millimetre features, neither is the right tool and the ceiling is the phone, not the technique.

When is a scanning service cheaper than buying a scanner?

When the job is occasional, large, metrology-critical, or due to somebody this week. A professional capture is one fee against a machine that only pays back across repeated work, and the shop absorbs the surfaces, the fixturing and the cleanup that cost you evenings. Buying makes sense when scanning is a recurring step in how you work rather than a one-off problem to clear. The directory at /providers lists providers by location and process, and one quote usually costs less than the third failed weekend of rescanning.

Why Phone 3D Scanning Apps Fail on Real Parts

You walked around the bracket with your phone, the app churned for a minute, and a complete-looking mesh dropped into your files. Then you imported it, measured the bolt spacing, and it was four millimetres out. Phone 3D scanning apps do not fail randomly or because you moved too fast — they fail at a boundary set by what the sensor physically measures, and no amount of rescanning moves it. This is where that boundary sits, what sits comfortably inside it, and what the next rung up actually buys you before you reach the point where having the part scanned by a service is the cheap option.

This post contains affiliate links. If you purchase through these links, 3D Prototyping Hub may earn a small commission at no extra cost to you.

This post also contains Amazon affiliate links. As an Amazon Associate we earn from qualifying purchases.

First, Work Out What Your App Actually Measured

"Phone 3D scanning" covers three different techniques that fail in three different ways. Most apps use one or blend two, and the loading screen rarely tells you which. Match yours before you blame your technique.

Technique	What it does	Where it breaks
Photogrammetry	Reconstructs shape by matching points across dozens of photos	No absolute size at all; needs surface detail to match on
Depth sensor (LiDAR / time-of-flight)	Measures distance directly, point by point	Centimetre-class depth, coarse point spacing, built for rooms
AI reconstruction	Fits a plausible surface to sparse or missing data	Invents geometry confidently and tells you nothing

The third one is the dangerous one. A gap-filling model produces a smooth, closed, attractive mesh over data that was never collected, and the result looks more trustworthy than the honest noisy capture it replaced.

The Size Is Not Real Until You Give It One

Photogrammetry recovers geometry from photographs only up to an unknown scale factor. That is not a limitation of the app — it is a property of the maths. Photographs of a real bracket and photographs of a perfect half-size model are the same set of images, so the solver cannot tell them apart and picks something.

Depth-sensor captures do carry a true scale, but with the sensor's own error riding on it. A part that measures 120.0mm can come back at 123mm with nothing flagged, because from the sensor's point of view that is a perfectly ordinary result.

Both are fixed the same way, and it takes a minute:

Measure one real feature across its longest available span with digital calipers — the longest span available, because a scale factor derived from a 10mm feature multiplies its own error across the whole part.
Scale the mesh to that number in your mesh tool.
Check a second, independent feature at the other end of the part.

If the second feature lands, your scale is good. If it does not, the error is not uniform and scaling cannot save it — the capture drifted, and that is a capture problem rather than a units problem. Our guide to why a scan refuses to print covers what else goes wrong between the mesh and a usable file.

Resolution Decides Whether Your Feature Exists at All

This is the part people argue with, and it is not an argument about skill. A scan cannot contain a feature smaller than the spacing between the points it collected. Phone depth maps are a small fraction of the camera's resolution and were specified for room-scale work at metres of range; a dedicated handheld is specified in the 0.05–0.1mm band over a short, fixed working distance.

So before scanning, ask what the smallest thing you care about is:

A 20mm boss or a panel outline? A phone will capture it.
A 2mm fillet, a 0.5mm chamfer, a bearing seat, a thread? It was never in the data. Not blurred — absent.

The honest workflow at that point is a hybrid, and it is faster than a rescan: use the scan for the overall shape and the feature positions, measure the critical dimensions by hand, and build those features properly in CAD. That is the reverse-engineering workflow rather than the copy workflow, and it is what the professionals do with good scans too.

The Mesh Looks Crisp Because the Photo Is Crisp

Phone apps drape a high-resolution photographic texture over a much coarser mesh. The lettering, the grain, the fastener heads you can see in the viewport are in the image, not in the geometry — and the geometry is all the printer or the CAD kernel gets.

Switch to a flat-shaded, untextured view before you judge any capture. What you see there is what you actually have. It is usually a shock the first time, and it ends a lot of pointless rescanning.

Phones Fail on Difficult Surfaces First, Not Last

Every optical method needs light to come back usefully, and a phone has the weakest tools for making that happen. Photogrammetry additionally needs visible detail to match points between frames, which is why a plain matte face can be nearly as hostile as a shiny one.

Three surfaces defeat a phone early: glossy or machined metal reflects the light away, matte black absorbs it, and clear or translucent plastic lets it through and returns data that is present but wrong. The counters are the same ones professionals use, and they work on a phone:

A self-vanishing matte scanning spray, applied in light passes from 20–30cm. A heavy wet coat pools in corners and is measured as geometry.
Positioning marker dots on and around the part, which give the solver fixed references on featureless faces instead of letting error accumulate frame to frame.
A motorised turntable for anything bench-sized, so the working distance stays constant and your arm stops being part of the measurement.
Even, diffuse indoor light that does not change mid-capture. Direct sun is the reliable killer.

If your captures stay bad after all four, the hardware is the ceiling. Our walkthrough of noisy and hole-filled scans separates the surface problems from the tracking ones in more detail.

What a Phone Is Genuinely Good At

Phone scanning is not a toy, and the honest boundary cuts both ways. It earns its place when the tolerance you need is looser than a millimetre and the features that matter are bigger than about five:

Rooms, furniture, vehicle interiors and installation envelopes — "will the enclosure clear the bulkhead?"
Bodies, faces and soft goods for cosplay and props, where the shape is the deliverable and nothing has to mate.
Rough fit envelopes: the footprint of a part you are designing around rather than the part itself.
Visual reference for modelling, and before-and-after documentation.

For any of those, a phone and twenty minutes beat a purchase order.

What the Next Rung Up Actually Buys

If scanning is a recurring step rather than a one-off problem, here is the ladder, stated from published specifications and price rather than from a shootout we did not run:

Consumer handheld scanners are generally specified around 0.05–0.1mm accuracy, with blue-light models rated tighter on small objects — enough for most mechanical reverse engineering. Revopoint's handheld scanner range sits in this band; read each model's accuracy and scan-volume figures against your actual parts rather than against the headline number.
Standalone structured-light systems fix the camera, the projector and the working distance in a calibrated rig, which is most of what you are paying for: repeatability you can put in a report, and better behaviour on difficult surfaces. SHINING 3D's EinScan standalone range starts well into four figures and is bought for exactly that reason.

Before you spend, read why scanner accuracy specs do not hold in practice — the headline figure is a best case under conditions you will not reproduce on your bench, and knowing that is what stops you buying the wrong rung. If you are choosing rather than diagnosing, the full comparison is in our 3D scanner buyer guide.

A Decision Table You Can Use Tonight

What you are making	Is a phone enough?	If not
Room or vehicle fit check	Yes	—
Cosplay, props, body shapes	Yes	—
Replacement cover or shroud, loose fit	Borderline — scale it and verify	Handheld scanner
Bracket, housing or part that mates	No	Handheld + CAD rebuild
Sealing face, bore, thread, tight tolerance	No	Standalone system or a service
Inspection against nominal CAD	No	Metrology-grade service

When to Stop and Send It Out

Three signals that you are past the point of diminishing returns: the part is large enough that drift dominates the result, the tolerance you need is the same order as the equipment's own accuracy, or the scan is blocking somebody else's work. Any one of those, and a professional capture is cheaper than another weekend.

That is doubly true when the job is inspection rather than modelling — comparing a manufactured part against nominal CAD is a measurement task with a report attached, and it needs calibrated equipment to mean anything at all.

For everything else, work in order: find out which technique your app uses, give the capture a real scale with calipers, check the untextured mesh rather than the pretty one, and decide honestly whether the feature you need was ever inside the sensor's reach. When it was not, the providers in our directory scan for a living and will send back a mesh you can build on.

Hero photo by Creative Tools via Wikimedia Commons, licensed CC BY 2.0.