Questions and answers

Aftermath Premium turns a raw 3D scan into a clean, watertight, CAD-ready model, ready to manufacture. You upload a mesh or point cloud, add a few photos of the part if you have them, and get back an STL. It runs in the browser and as a desktop app for Mac and Windows.

If your scanner software can export STL, OBJ or PLY, or a point cloud in ASC, XYZ, PTS, PCD or PTX, yes. That covers almost every desktop scanner, handheld scanner and phone scanning app. It doesn't need to know which scanner made the file.

Meshes in STL, OBJ or PLY, and raw point clouds in ASC, XYZ, PTS, PCD or PTX. The output is a watertight STL for any slicer, CAD or CAM software.

Not yet. The output is a watertight STL. Most CAD programs can import it as a mesh body, and you can use it as the reference for a parametric rebuild.

No, but they help. Photos show what the real part looks like where the scanner missed or smoothed things over, like sharp edges, gear teeth or the side the part sat on. You can add up to 10 per scan. Plain phone photos in good light work.

It depends on the scan. A clean, complete scan comes back close to the real part. Where the scanner missed surfaces, the engine completes them the way an engineer would, but it can't recover detail that was never captured. Reference photos help most in those spots. Try the free sample scans, or your one free scan, to see what to expect from your own scanner.

Yes. Phone scanning apps are great for a part's overall shape and size. Small details like threads, ribs, thin clips and small holes are usually too fine for a phone to capture, so where your scan doesn't have them, the model keeps them simple. A dedicated 3D scanner, or its raw point cloud export, captures far more. We're building a way to rebuild those details from your reference photos (coming soon).

Your scan and photos are uploaded to our servers to be processed. To check the model, your photos and pictures of your scan are analysed by a service provider we use for image analysis (listed in our Terms). Your scan, photos and finished model are kept for 2 hours so you can download the model and tell us if something's wrong with it, then deleted. If you choose to help improve Aftermath, we keep your scans to learn from. That's off unless you turn it on, and you can turn it off any time (what this means). We never sell or share your scans.

Each file you process counts as one scan once it finishes, whatever its size. A run that fails or that you stop doesn't count, and sample scans never count. Processing the same file again counts again.

Click "Something's wrong with this model" under the result and tell us what's off. We keep that scan so we can look at exactly what happened, and we'll email you. If we can't make it right, you get the scan back or your money back. Report it within 2 hours of the run, while we still have the files.

Create a free account on the demo page. The sample scans are free to run as many times as you like, and you get one scan of your own free. No credit card needed, just a phone number to confirm by text. After that, buy single scans or pick a plan.

Yes, any time on the manage plan page. Upgrades apply right away and you pay only the difference for the rest of the period. Downgrades start with your next billing period. There are no penalties or lock-in periods.

Yes. The desktop app is free to download and install on any computer you use. Your account can be signed in one place at a time, either on the website or in the app. If you sign in somewhere new, we show you where you're signed in now, and you can sign out there with one click.

Yes, through the web app in any modern browser. A Linux desktop app isn't available yet. Email info@aftermathprem.com if you'd like to hear when it is.

3D scanning measures the shape of a real object and records it as a large number of points in 3D space. Software turns those points into a surface, usually a mesh of triangles. The result is a digital copy of the outside of the object that you can measure, 3D print, or use as a reference in CAD.

Most scanners for parts use triangulation. The scanner projects light onto the object, either a laser line or a pattern, and a camera set a known distance away sees where that light lands. From that known distance and angle, the software works out the 3D position of every lit point. Long-range scanners for buildings and rooms often use time of flight instead, timing how long light takes to bounce back.

Structured light scanners project a pattern of light and capture a whole patch of surface at once. They're fast and capture fine detail on small and medium parts. Laser scanners sweep one or more laser lines. They usually cope better with dark or shiny surfaces and bright surroundings, and handheld laser scanners are common in quality inspection. Both can be very accurate. The surface and size of the part usually matter more than the technology.

Photogrammetry builds a 3D model from many overlapping photos. Software finds the same features in different photos and works out where each camera was and where each feature sits in space. It works well on textured objects and on very large ones like buildings. It struggles with plain, shiny or clear surfaces, and the model has no true size until you give it a known measurement, such as a scale bar.

Yes. Photogrammetry apps work on any phone with a decent camera. iPhone Pro and iPad Pro models also have a LiDAR sensor, which is good for rooms and furniture but too coarse for small mechanical detail. For a part you plan to manufacture, a phone scan is a reasonable starting point. Check the key dimensions with calipers before you rely on it.

Scanners need light to scatter off the surface and back to the camera. Shiny surfaces reflect it away in one direction, black surfaces absorb most of it, and clear surfaces let it pass through. The usual fix is a thin matte coating. Scanning sprays made for this disappear on their own within hours. Developer spray and dry shampoo are cheaper but need cleaning off. A light coat adds very little thickness, but a heavy one can matter on precise parts.

Markers are small dots, usually reflective, that you stick on the part or around it. The scanner uses them to keep track of its own position, which helps on large, smooth or symmetrical objects where the shape alone gives it nothing to lock on to. The surface under each marker isn't captured, so it leaves a small hole to fill later.

Phone apps are free or cheap. Entry-level desktop and handheld scanners cost from a few hundred to around two thousand dollars. Professional handheld scanners used in engineering usually cost from several thousand to tens of thousands of dollars, and inspection-grade systems cost more. For one-off jobs, many people rent a scanner or pay a scanning service.

Hobby scanners usually list accuracy around 0.05 to 0.1 mm. Professional inspection scanners list around 0.02 to 0.03 mm, often plus an allowance that grows with the size of the part. Real results can be worse than the spec sheet, because surface finish, technique, alignment between passes and temperature all add error.

Resolution is how close together the measured points are, which sets the smallest detail you can see. Accuracy is how close those points are to the true surface. A scan can be sharp and detailed and still be wrong, for example half a percent too large because of a calibration error.

A scanner can only capture surfaces that its light and camera can both see. Deep pockets, narrow slots, the inside of holes and the side the part sat on are often missed. Shiny or dark patches and spots under markers drop out too. Scanning from more angles helps, but most scans of real parts end up with some holes.

Scan it standing up, flip it, and scan it again. Then align the two scans and merge them. Most scanner software does this with markers or by matching the shape where the scans overlap, so make sure both scans see some of the same surface.

Usually somewhere between 50 and a few hundred for a single object. Each photo should overlap the next by about two thirds, taken while circling the object at two or three heights. Even light and sharp, well-exposed photos matter more than the count.

For simple parts made of flat faces, round holes and standard sizes, calipers and a micrometer are often faster and more accurate. Scanning pays off on curved, organic or complex shapes that are hard to measure by hand. Many engineers do both: scan for the overall shape and measure the critical dimensions by hand.

Industrial CT scanning uses X-rays to capture the inside of a part as well as the outside, including internal channels, voids and parts hidden inside an assembly. It costs far more than surface scanning and is usually done by specialist labs.

A point cloud is a list of points in 3D space, each with X, Y and Z coordinates and sometimes a colour, an intensity or a surface direction. Many scanners capture this first. It has no surface, so you can't print it or measure its volume until it's turned into a mesh.

A mesh connects points into triangles so they form a continuous surface. A point cloud is only the points. Turning a point cloud into a mesh is called surface reconstruction. Poisson reconstruction is one of the most common methods.

STL stores a 3D shape as a list of triangles and nothing else: no colour, no units and no design history. It's the most widely supported format for 3D printing, and comes in a compact binary version and a larger text version.

None. STL stores bare numbers, and most software assumes millimetres. If a part imports 25.4 times too big or too small, it was saved in inches and opened as millimetres, or the other way round. Scaling by 25.4 fixes it.

A binary STL uses 50 bytes per triangle, so a scan with 2 million triangles is about 100 MB. Text STL files are several times larger. Reducing the triangle count, called decimation, can shrink the file a lot with little visible change.

All four can store a triangle mesh. OBJ can also carry texture coordinates and a material file, so it's common for coloured scans. PLY can store colour and other values for every point and is common for scan data. 3MF is a newer 3D printing format that records units, colour and several parts in one file. STL is the simplest and works almost everywhere.

STEP (ISO 10303) is the standard format for moving solid CAD models between programs. Instead of triangles, it describes exact surfaces like planes, cylinders and freeform patches, so a hole is truly round. Machine shops and mould makers usually ask for STEP. IGES is an older format that does a similar job less reliably.

You can, but a direct conversion only wraps each triangle as a flat face inside a STEP file. It's still faceted, it can be very heavy, and you can't change a hole diameter or a fillet. A useful STEP file comes from rebuilding the part in CAD with the mesh as a reference, or from auto-surfacing, which fits smooth patches over the mesh.

They're point cloud formats. ASC and XYZ are plain text files with one point per line. PTS is a text format from Leica that adds an intensity value. PTX, also from Leica, keeps each scan's grid and position. PCD is the format of the Point Cloud Library. Two others you'll meet are E57, a vendor-neutral standard for laser scan data, and LAS, which is mostly used for aerial and survey lidar.

A watertight, or closed, mesh fully encloses a volume with no gaps, so it has a clear inside and outside. Slicers, volume calculations and most CAD and CAM tools need this. Raw scans are almost never watertight.

In a clean, closed mesh every edge is shared by exactly two triangles. A non-manifold edge is shared by three or more, and a non-manifold vertex joins pieces that only touch at a single point. Both make it unclear what is inside the part, which leads to failed slices and failed imports.

Each triangle has a front side that should face out of the part. When some face inward, software reads that area as inside out. It shows up as dark or see-through patches in a viewer, or missing walls in a slicer.

Small holes are filled by bridging the edge with new triangles that follow the surrounding curve. Large holes are harder, because the software has to work out what was there. A flat base can be closed with a plane, but a missing boss or a rounded corner needs more information, such as photos, a matching side of the part, or measurements.

Every scanner has some measurement noise, which shows up as a fine grain. Bigger bumps and ridges usually come from passes that didn't line up perfectly, from reflections, or from the part or scanner moving during the scan.

Smoothing averages each point with its neighbours to remove noise. Too much of it rounds sharp edges, softens small features and shrinks the part slightly. For mechanical parts, the goal is to remove noise on the faces while keeping the edges sharp.

Decimation reduces the number of triangles in a mesh. Good decimation removes triangles on flat areas and keeps them where the surface curves, so the shape barely changes while the file gets much smaller.

Remeshing rebuilds the triangles of a mesh so they're more even in size and shape. It helps when a scan has long, thin triangles that cause trouble for later steps like smoothing or simulation.

When an object is scanned in several passes, each pass has its own position. Registration moves them into one shared position so they overlap correctly. It's done with markers, by picking matching points, or with ICP (iterative closest point), which nudges one scan until it best fits the other.

Two passes over the same surface were slightly misaligned, so the surface appears twice with a small gap between. It's common on thin parts and after the scanner loses tracking. Re-aligning the passes before merging fixes it. Smoothing over it doesn't.

Small loose pieces near the part are usually the table, the turntable, a clamp, dust or reflections that the scanner recorded. They're removed by keeping the main piece and deleting the rest, as long as the part itself is in one piece.

Reverse engineering a part means producing a usable digital model of it without the original design files. People do it to replace a part that's no longer made, repair or improve a worn one, make an accessory that fits an existing product, or understand how something was made.

In the United States, taking apart and measuring a product you lawfully own is generally legal, and trade secret law treats it as a fair way to learn how something works. Making and selling copies is a separate question. Patents, design patents, copyright, trademarks and licence agreements can still limit what you do with the result. A US utility patent usually lasts 20 years from filing. This is general information, not legal advice. Talk to a lawyer before selling a copy of someone else's product.

Scan the part, align and merge the passes, then clean and close the mesh. From there you either use the mesh directly, for example to 3D print it, or rebuild the part in CAD. A CAD rebuild means splitting the mesh into regions like planes and cylinders, fitting exact shapes to them, sketching on cross-sections and recreating the features. Finally, compare the CAD model with the scan to check it.

A mesh is a surface made of triangles. A solid CAD model is made of exact faces like planes and cylinders, and a parametric model also stores the features and dimensions that built it. In a parametric model you change a hole diameter by editing one number. In a mesh you'd have to move thousands of triangles.

Design intent is the geometry the designer meant, as opposed to what the part measures today. A hole that scans at 9.96 mm was probably designed at 10 mm. Faces 89.7° apart were probably meant to be square. Good reverse engineering rebuilds the intended shape, not the worn or warped one.

Auto-surfacing covers a mesh with smooth NURBS patches and outputs a CAD solid. It's quick and suits organic shapes like sculptures, body scans and grips. The result follows the scan exactly, flaws included, and has no features or dimensions to edit, so it's a poor fit for mechanical parts.

Dedicated tools include Geomagic Design X and QuickSurface. General CAD programs like SolidWorks, Fusion and Siemens NX have tools for working with meshes, such as cutting sections and fitting surfaces. MeshLab and Blender are free and good for viewing and editing meshes. Aftermath Premium handles the step before CAD: it turns the raw scan into a clean, closed model.

Scan what's left and use the undamaged areas as the reference. Symmetrical parts can be mirrored from the good side. Worn surfaces are rebuilt to their likely original size, which a mating part, an unworn sample or a standard size can confirm. Missing pieces are modelled from photos, drawings or the part they fit.

No. A scan records the shape of one part as it is today. It can't tell you the designed tolerances, the material, heat treatment, surface finish requirements or which faces matter for the fit. Those come from drawings, the manufacturer, testing or engineering judgement.

Scanners rarely capture threads cleanly because the grooves are small and deep. The usual approach is to measure the outside diameter with calipers and the pitch with a thread gauge, then model a standard thread in CAD, such as M8 × 1.25.

No. Reverse engineering exists for when drawings are lost or were never available. If you have them, even partial or old ones, use them to confirm dimensions and tolerances.

It depends on the part and what you need back. A cleaned-up scan is the cheapest. A full parametric CAD model from a freelancer or engineering firm can start around a hundred dollars for a simple bracket and run into thousands for complex housings or assemblies. Turnaround ranges from a day to several weeks.

A scanning service gives you the scan, usually as a mesh. A reverse engineering service goes further and gives you a model you can edit and manufacture from, often a STEP file and sometimes drawings. Many companies offer both.

Yes, once the mesh is watertight and free of errors. Export it as STL or 3MF, open it in your slicer, check the size and orientation, and print. Raw scans usually need cleaning up first, or the slicer will produce gaps, missing walls or failed prints.

It depends on the whole chain, because scan error, cleanup and printer accuracy add up. FDM printers are typically within a few tenths of a millimetre, and holes tend to print slightly small. For parts that must fit others, measure the critical features with calipers, adjust the model, and print a test piece first.

Yes, but CAM software works best with a solid CAD model. Some CAM programs can cut 3D surfacing toolpaths straight from a mesh, which suits organic shapes and moulds. For flat faces, pockets and drilled holes, rebuild the part in CAD first so the machine cuts true planes and round holes.

Not directly. Mould makers need a CAD model, usually STEP, with draft angles, shrinkage allowance, parting lines and wall thickness worked out. The scan is the starting reference for that model.

After aligning a scan with a CAD model, software measures the gap between them at every point and colours the surface by it, for example green within tolerance and red or blue outside it. It's used to inspect manufactured parts against their design and to check a reverse engineered model against the scan.

Geometric dimensioning and tolerancing is the system engineering drawings use to define allowed variation in form, orientation and position, such as flatness, perpendicularity and true position. In the US it's defined by ASME Y14.5. A scan can check many GD&T requirements, but it can't tell you what they should be.

A legacy part is still in use but no longer made or supported, often with no drawings or CAD available. Reverse engineering it from the physical part is often the only way to get a replacement.

Automotive restoration and aftermarket parts, aerospace and defence maintenance, factory maintenance, prosthetics and orthotics, product design, museums and heritage work, and 3D printing. Any business that keeps old equipment running ends up needing parts nobody makes anymore.