Review: Matter and Form Three 3D Scanner

As a chronic tinkerer, I’ve always wished there was a simple, efficient, and effective way to get a full-resolution 3D model of a part without spending hundreds of hours learning 3D modeling or thousands of dollars on a high-resolution 3D scanner. This is the problem Matter and Form wanted to solve with the Three, a 3D scanner that claims to be simple, intuitive, and inexpensive.

It’s important to emphasize that 3D scanners are not the end-all, be-all of creating 3D models. They are a tool to be put into the arsenal of a 3D modeler, but you shouldn’t expect perfect results from a scanner with zero touch-ups. The Matter and Form Three is small and affordable, mostly self-contained with onboard processors and internet connectivity. It requires a laptop with a browser to connect to the scanner over a local internet connection, and costs $2,399 MSRP but is often on sale for $1,499, which includes the scanner, a motorized turntable, calibration equipment, a tripod, and a carrying case.

Easy Setup

Photograph: Henri Robbins

The Three is incredibly simple to set up and has a robust series of tutorials that explain nearly every aspect of its operation. Besides more advanced settings hidden in the menus, every part of the calibration and operation is detailed within the interface and can be adjusted quickly.

The most impressive part of the scanning is the lack of prep needed. While other scanners might require tracking points, the Three can be operational within minutes after you’ve started it up, waited for the system to initialize, and connected to the computer. This can be done by connecting the scanner to your device wirelessly with the included antenna, connecting it to your router via an Ethernet cable, or connecting directly to your device via Ethernet. The first two methods will allow wireless connection to the scanner with any device, while the third will only allow connection to the device the scanner is plugged into.

The application for the scanner runs entirely in-browser, accessed through an IPv4 address that’s displayed through the scanner’s projector. Once you’ve connected, the website looks simple and sleek, with prominent tutorials and an intuitive interface.

Henri Robbins

The Three uses a projector to illuminate the subject being scanned. This, along with adjustments for shutter speed and gain, can be used to fine-tune the exposure of the captured object. The exposure settings are well-explained, and the software clearly states whether any part of the object is over-exposed by highlighting these sections in red.

One of the major downsides of a scanner like this is the reliance on visual lines of sight and lighting. Some details can’t be captured easily due to either distinct shadows being cast by other parts of the object or overlapping components blocking a full scan. Any objects with black and white sections will be difficult to scan in one go and will instead need to be scanned twice with different exposure levels. However, with dedication and patience, you can scan nearly everything. For details that really refuse to scan, even basic 3D modeling skills can get you a passable result.

Another downside is the reliance on a secondary device. While using a PC to visualize and manage scans is convenient, it also means the Three can’t be used on its own, and is ultimately tethered to a laptop or desktop computer. The scanner can still operate wirelessly without needing an internet connection, but the ability to operate it without a connected computer (or even mobile support) would be nice.

Scan Quality

Photograph: Henri Robbins

When properly set up, the Three can create unbelievably detailed and high-resolution scans, capturing even the smallest changes in shape and texture. That's thanks to two 13-megapixel Sony sensors to capture details down to 0.033 mm. It utilizes these cameras with its “ChromaSpec” system, which allows the scanner to capture color alongside shape. The scanner also uses a Texas Instruments-designed projector to apply patterns onto the scanned object. These projections function similarly to tracking points and allow the cameras to easily determine the depth of specific points on the object.

It can create incredibly high-resolution scans of close-up items and passably detailed scans of items at greater distances. While the 0.033 mm of accuracy is rated for items that are 220 mm away from the scanner, an item at 700 mm will only have 0.4 mm of accuracy. However, moving the scanner back will drastically increase the scan area, from around 155 x 115 to 480 x 356 mm. Essentially, when scanning larger items, the scanner can be used close-up with multiple scans, or it can work at a distance to capture the entire item with one or two passes. This also means large and small scans can be combined if you only need small details in some parts of a scan.

One of the scans I took, a singular dye-sublimated PBT keycap, had enough detail that I could make out small indents on the underside of the keycap left by a wire keycap puller, and the divot left by the injection-molding process. With a double-shot keycap, I could even make out the difference in depth between the main plastic of the key and the second plastic of the text. Textures on surfaces were captured accurately, and I was impressed by just how much the scanner was able to pick up.

Photograph: Henri Robbins

The turntable worked impressively well. Calibration was easy, with only three steps in total: Focus the scanner on your object, replace the object with a calibration card, and press a button. As long as your object stays in the same place, you won’t need to recalibrate unless you move the turntable or the scanner. The color data was impressive, too, and is a great tool for creating 3D assets in games. At times, I found it got confused by complicated textures like fabrics, but even then, the end product was usable and could be retextured since the shape and texture data are captured separately.

Matter and Form says the scanner can work in nearly any lighting, and this is technically true, but I found the Three performs best in a well-lit and bright environment. This meant, for most of my scans, I used not only ambient lighting during the day, but gravitated toward additional light sources from lamps or studio lights for the best results. This isn’t to say the scan quality was poor otherwise, but I found the scanner easiest to calibrate and work with when the environment had consistent and bright lighting. Much like a standard camera, having more lighting meant there was more opportunity to adjust the different settings.

Hiccups and Bugs

Henri Robbins

The largest hiccups I ran into with the Three were all software-based, and these issues were only made apparent by how refined the software is as a whole. This is the best-case scenario for a device like this: The software can easily be updated and tweaked after release, something that happened multiple times during my time with the Three.

The largest issue was the alignment. The Three has two separate ways to align multiple scans. There is an automated alignment system and a point-based system. The automated system attempts to merge two scans on its own, while the point-based system has the user drag and drop markers in the same places on both scans. Both of these systems work well half the time. The other half of the time, they completely miss the mark—scans end up completely misaligned, very subtly incorrect, or refuse to merge altogether. This isn’t a huge issue, since you can always revert and realign the two components, but it got tedious at times, especially when the two halves of a scan appear to fit together in an obvious way, yet the automated system couldn't figure it out.

The most annoying instance of this was with my scan of a keycap. The system failed to align the scans multiple times. Instead, the scans would intersect at strange angles, merge at the bottom to create a trapezoid-like shape, or produce error codes until I got the alignment points in just the right spot, all of which were, quite frankly, really annoying. Something as simple as a keycap shouldn’t be that difficult. Full manual alignment of scans would be a really nice feature to have.

Henri Robbins

Once the scan is finished, you’ll need to combine the scanned points. This system works well, but if there’s any missing or extraneous data in the scans, it has the potential to make baffling decisions. One of the most significant was when I scanned a Bionicle mask. The scan turned out perfectly, with crisp edges and impressive accuracy. After the scans were aligned, I attempted to merge them together. The mask merged perfectly, but was surrounded by floating specks and large, curved forms that seemingly spawned out of nowhere. These can be cleaned up after the fact with software like Blender, but these extra forms prevented me from using features like Simplify and Convert, or Texturize before exporting.

After looking at the scans, I eventually determined that these shapes were being created by tiny stray specks picked up during the scanning process, small enough and far enough away from the main scan to be nearly imperceptible. The meshing system got confused by this and summoned massive blobs of polygons. However, this can be fixed, and you do learn to look out for these types of issues. The raw scans can also be exported for processing on your own, without any merging or meshes. However, the files I exported were large enough that, on my system, software like Blender struggled to render them without freezing up.

When the system works, it is incredibly convenient and practical. I would go so far as to say that, when things go right, it's practically seamless. Calibration is intuitive and quick, the alignment tool is both convenient and smart, and the meshing works wonderfully most of the time. More user controls during the processing would be an improvement over the relatively hands-off system that exists.

Henri Robbins

The scan processing can be slow when using the highest settings, but this isn’t a major issue in reality. The processing is all done within the scanner itself instead of being done on your system, which means it doesn’t take up any of your computer’s processing power beyond displaying the website. Instead of having to step away for a bit while the mesh is being processed, you can do other things on your computer and leave the scanner’s tab open in the background.

While I’ve mentioned small flaws, most of these are possible to work around with time and effort. When you get down to it, the scans this device makes are impressive on their own, and it is a great tool for creating 3D models of physical objects. It is very well-priced considering its resolution. Like any 3D scanner, you will need to operate within certain limitations and do some work on your own to make truly “good” 3D models, but this tool makes it incredibly easy to create high-resolution, detailed 3D models that would otherwise be tedious.