Convert a physical object into a 3D model ready to print It's no longer just for labs or expensive equipment. Between photogrammetry, 3D scanning apps, and smartphones with ToF or LiDAR cameras, today you can scan anything from a figurine to a huge statue and send it to your 3D printer without leaving home. Use your mobile phone's ToF camera as a 3D scanner to print with this guide.
In this article you will see, in great detail, how to take advantage of the your mobile phone camera (including ToF or LiDAR) To capture objects, we'll cover what free software to use, what tricks you need to get decent results, and how to prepare the model for 3D printing or augmented reality viewing. The idea is that, by the time you finish reading, you'll know which app to try, how to take the photos or perform the scan, and what steps to follow to get the final STL or OBJ file.
From real object to digital file: photogrammetry and mobile scanning
The basis of almost everything you're going to use Using your mobile phone to create printable 3D models is photogrammetry, also known as Structure From Motion (SfM). In short, it involves taking many photos of the same object from all possible angles and letting software detect common points between images, calculate the camera's position for each shot, and reconstruct the 3D geometry from those pixels.
Although it sounds very technical, what you actually do is fire a bunch of photographs around the objectLoad them into a photogrammetry program (on your PC or in the cloud) and wait for it to generate a point cloud and, from there, a 3D mesh. This mesh can then be cleaned, holes closed, smoothed, and converted to formats such as STL o OBJideal for 3D printing or uploading to platforms like Sketchfab.
In addition to pure photo-based photogrammetry, there are increasingly more mobile phones with ToF or LiDAR sensors that directly capture depth information in real time. By combining that depth with the RGB camera, apps like Polycam or Scandy Pro are able to generate the 3D model as you move around the object, without needing to then upload a hundred photos to the computer.
Photogrammetry step by step: taking good photos makes all the difference
Photogrammetry has one huge advantage: you can do it with a a regular mobile phone camera or a DSLRWithout any extra hardware. The key is how you take those photos. The better you take the shot, the less trouble you'll have later on your computer.
In terms of quantity, it's advisable to shoot at least 20-50 photos For small objects, you can easily increase the ISO to 80 or more if you want to capture a lot of detail. Don't worry about going over the limit: the software can discard photos that don't blend well with the rest.
The most common technique is to move in circles around the object while maintaining a high overlap between shotsAround 60-80%. This means that most of the visible surface in one photo should also appear in the next. Furthermore, it's important not to move the object or change anything in its surroundings between shots, as this confuses the algorithms.
Lighting is just as important, if not more so, than the camera.The ideal lighting is soft and diffused, without harsh shadows or strong backlighting. A cloudy day outdoors is almost perfect; indoors, it's better to use several soft light sources than one very intense spotlight. And, of course, avoid having anything moving during the session (people walking by, pets, leaves rustling behind the subject, etc.).
Another basic rule is that the object occupies a important part of the framing in every photo. If it appears too small in the image, the program will have fewer usable pixels to reconstruct details, and the quality will suffer. If your phone allows it, use manual focus or an app like Open Camera to ensure everything is sharp and avoid blurry photos that the software will eventually discard.
Easy objects, troublesome objects, and capture tricks
Some elements seem made for photogrammetry: statues, rough surfaces, richly textured piecesThe software has plenty of visual markers to follow, and the reconstruction usually turns out surprisingly well, even on large models that would be impossible to fit into a desktop 3D scanner.
In contrast, shiny, very smooth, or transparent materials are the classic nightmare for these systems. A ping-pong paddle, for example, has large, smooth, and reflective faces; as you move, the light reflects differently, and the program can't find stable points that coincide from one photo to another, so It barely generates points in those areas. And that's where deformed meshes come from.
The trick here is to force an artificial texture: you can cover shiny surfaces with masking tape or with a fine powder (flour, talcum powder, matte spray, etc.). The goal is to eliminate reflections and create subtle variations in tone that the algorithm can track. The same applies to glass and other transparent materialsIf you don't give them a temporary matte coating, the result will be a "ghost" object full of gaps.
For users with a bit more experience, a convenient alternative is to record a video surrounding the object Then extract individual frames to feed into the photogrammetry software. If you do this, set the shutter speed to something fast, for example 1/80 or faster, to avoid motion blur in the frames.
Free photogrammetry software: Colmap, Meshroom and others
Once you have the photos, the hard work begins: photogrammetry processingThere's a lot of powerful software available here, but many solutions are very expensive or have very limited free versions. Luckily, there are free alternatives that work more than adequately.
Among the free desktop options, the following stand out: Colmap (Windows, macOS and Linux), Meshroom and veteran tools like VisualSFM. Colmap is especially interesting if you want to learn how the process works at a low level, but it requires a CUDA-compatible Nvidia GPU to run smoothly, and can get bogged down with very high settings.
Meshroom, for its part, is an open-source 3D reconstruction software with a quite friendly interfaceYou drag and drop the photos, press "Start," and wait. Internally, it sets up a complete photogrammetry pipeline (feature detection, matching, structure, texturing, etc.), but you don't have to wrestle with each stage if you don't want to. Again, it really takes advantage of an Nvidia graphics card.
There are also cloud-based solutions like Autodesk ReCap Photo, which allow you to upload sets of images from 2-3 GB They handle the calculations on powerful servers. The downside is that, after the trial period, they usually switch to a subscription model, which isn't suitable for everyone if you're only going to do occasional scans.
For those who prefer to work directly on their mobile phone or tablet, many modern 3D scanning apps combine classic photogrammetry with the use of depth sensors, and many outsource the intensive processing to the cloud. This reduces the need for a very powerful PC, although it does mean uploading your photos to external servers.
Practical example with Colmap and Meshlab
If you opt for a traditional desktop solution, a typical workflow with Colmap + Meshlab It could be something like this: first you run the automatic reconstruction in Colmap and then you use Meshlab to clean and generate the final mesh suitable for 3D printing.
The easiest way is to organize the project by creating a working folder with a subfolder “Photos” where you'll put all the images. Before starting, quickly review and delete any clearly blurry or out-of-focus photos. Then, open Colmap, choose "Automatic reconstruction," and specify both the image folder and a "workspace" folder where it will save all the data and results.
You can leave many parameters at their defaults, but it's usually a good idea to set the quality to a medium value. If you switch to very high quality with many photos and a basic PC, there's a good chance the program will crash. It takes forever or it just freezes upProcessing time can range from a few minutes to several hours, depending on the number of images and the hardware.
When finished, Colmap will show you a view of the reconstructed scene with the estimated position of each camera. In the “denso/0/” folder within the workspace, you will find, among others, two key files: fused.ply (point cloud) and meshed.ply (already triangulated mesh). The output mesh can be printable after a good review, but it's usually best to process the point cloud in your own way to achieve a finer result.
From point cloud to printable mesh with Meshlab
mesh lab It is a free program specializing in point cloud and mesh processingwith many cleaning, remeshing, and simplification tools. The typical workflow starts with the fused.ply file generated by Colmap to create a mesh from scratch.
The first step is to import the point cloud using "File – Import Mesh" and selecting fused.ply. You'll see the object as a dense collection of tiny points. Next, it's a good idea to select and delete all the points that aren't part of the model (floor, walls, people who walked through, etc.). You can do this using the vertex selection tool and then "Delete Vertices".
Once only the points you're interested in remain, it's time to create the mesh. A commonly used method in Meshlab is the option “Screened Poisson Surface Reconstruction” within the Remeshing and Reconstruction filters. The most important parameter is “Reconstruction Depth”: values between 13 and 15 usually provide a good balance between detail and calculation time for most cases.
The result will be a closed mesh, but there will almost always be clean up odd edges, strange creases, or stains where there wasn't enough data. For that, Meshlab has triangle selection and "Delete faces" tools, as well as tools to remove unwanted parts of the generated mesh.
When you are satisfied with the overall shape, you can export the model from “File – Export Mesh”. If you want to keep color and texture For online sharing, OBJ, PLY, or 3DS formats are good options. If the goal is to 3D print it, the classic format is STLKnowing that you will lose color information but you will have a clean geometry for laminating.
Improve the model: smoothing, cleaning, and face reduction
The mesh resulting from a scan (whether by photogrammetry or depth sensors) is almost never perfect. In fact, it's normal for it to have roughness, giant faces, small holes and an excessive number of polygons. This is where post-processing filters come into play.
To remove fine noise, Meshlab offers various methods of surface smoothingOne commonly used technique is "Laplacian Smooth," where you can increase the number of steps (for example, between 8 and 15) to progressively smooth the model. However, if you go too far, you'll also smooth out important details, so it's best to proceed gradually and observe the effect.
Another useful tool is the selection of faces with excessively long edges, a common occurrence when the algorithm has attempted to "bridge" areas with limited data. In "Filters – Selection – Select Faces with edges longer than…" you can preview and adjust the length threshold to primarily select faces with excessively long edges. the large unwanted faces, then deleting them with “Delete selected faces”.
If the model has hundreds of thousands or millions of faces, handling it in a slicer or editing software becomes cumbersome. That's why it's common to apply a polygon reduction With filters like “Quadric Edge Collapse Decimation,” you can set the maximum number of pages you want to print. For home printing, working with fewer than 100,000 pages is usually more than enough; anything over a million is clearly excessive for most printers.
For even more advanced refinement, programs like Blender or Meshmixer allow you to close holes, adjust flat bases so that the model sits well on the print bed, add custom supports, or divide the piece into several parts for better printing.
Mobile apps and workflows: from ToF camera to 3D model
In addition to classic photogrammetry using photographs, today you have a wide range of other methods at your disposal. 3D scanning apps for mobile They utilize ToF sensors, TrueDepth cameras, or LiDAR. Many of them export directly to STL or OBJ, so you can use them as a base for 3D printing.
In the Apple environment, where devices with LiDAR are abundant, it shines especially brightly. polycamIt combines a fast LiDAR mode with a photogrammetry-based Photo mode. It can generate models of entire rooms in seconds, and in Photo mode, it uploads between 20 and 200 images to the cloud to reconstruct objects in considerable detail. You can export the results in various formats, including STL for printing.
Another example in the Apple ecosystem is the API Object Capturewhich simplifies photogrammetry for iPhone and Mac users. Apps like PhotoCatch or Polycam itself in Photo mode use Object Capture to convert image collections into 3D models in USDZ format almost automatically, ready for AR and also convertible to other more familiar formats for printing.
On Android and mixed platforms there are many different alternatives: from Scann3D, focused on pure photogrammetry from photos, to cloud-based multiplatform solutions like Polycam for Web, which accepts up to 250 images taken with any camera (Android mobile, DSLR, compact…) and generates a downloadable 3D model in OBJ, GLTF, FBX, DAE or STL.
The advantage of these apps is clear: with your ToF camera, LiDAR, or even just RGB With an internet connection, you can scan an object, clean up the result a bit in Blender or Meshmixer, and in no time, have it sliced and ready for your 3D printer. You don't need a powerful workstation, because the heavy lifting is usually done in the cloud.
Featured 3D scanning applications: from the human body to small objects
The ecosystem of scanning apps is very broad, and each one is geared towards a specific niche. Some focus on the human body, others on rooms, and others on... small objects for 3D printingIt's helpful to know the main ones in order to choose the one that best suits what you want to do.
For instance, 3DsizeMe It is designed to capture human anatomy (feet, legs, torso, etc.) using the Structure Sensor 3D on iOS devices. It is widely used in medical and orthopedic settings, and exports in STL or OBJ format for designing custom insoles, orthotics, or products using 3D printing.
For scanning entire rooms, renovations, and interior architecture, Occipital's Canvas is an interesting option. Use the LiDAR integrated into iPad Pro or the Structure sensor It's connected and allows you to walk through a room to build a 3D model of the space in real time. It then offers a service to convert those scans into CAD files ready for architecture, for a price per model.
If you want to capture faces, busts, or medium-sized objects, there are apps like Capture, Heges or Scandy Pro These tools take advantage of the iPhone/iPad's TrueDepth camera or LiDAR. They allow you to export models in PLY, OBJ, or STL formats, adjust lighting parameters, apply noise filters, and generally produce quite decent results for printing busts, helmets, custom figures, etc.
For small objects, Qlone proposes a curious approach: it prints a special "mat" and places the object on top; the app guides the scanning and then merges different captures to generate a mesh. Then it lets you modify size, texture and export in formats such as OBJ, STL, USDZ, GLB, or FBX. It's ideal for those who want a very guided process without too much hassle.
Photogrammetry with affordable apps: Scann3D, Trnio and others
If you don't have a mobile phone with LiDAR or ToF but want to pursue a more economical route, you can use apps purely based on photogrammetry from photosThis is where options like Scann3D for Android or Trnio for iOS come into play.
Scann3D Turn your Android smartphone into an image capture tool for reconstructing 3D models by generating point clouds. From multiple photos of the object, create a model that can be exported as OBJ, STL, or PLY files. It's compatible with augmented reality applications and platforms like Sketchfab, and is a great way to get started without needing a powerful PC.
On iOS, trio It offers three modes of use: importing existing photos, ARKit mode (which helps you move correctly around the object), and a specific object mode that seeks the best balance between texture and color. The generated models can be exported to other 3D editors for cleaning and retouching before printing or uploading to the web.
Also worth mentioning itSeez3Dwhich uses sensors like the Structure Sensor or Intel RealSense cameras along with iOS or Windows devices. The typical workflow is very simple: you scan, send the data to the cloud, wait for processing, and download a model with higher quality than what you see in the raw preview.
In the realm of Sony smartphones, the 3D Creator app (included in several Xperia models) allows scanning in different modes (selfie, full face, head, food, freeform) and export the results for 3D printing or augmented reality applications. This is an example of how some manufacturers have tried to integrate 3D scanning directly into the mobile phone system.
From mobile phone to 3D printer: cleaning, scaling and repositories
Once you have your 3D model (whether from Colmap, Meshroom, Polycam, Scann3D or any other app), there are still a few steps left to... come out of the printer properlyNormally, you should first go through an editor like Blender or Meshmixer for fine cleaning.
In Blender, for example, you can enter edit mode, select with the lasso tool all the mesh remnants around the main object (pieces of floor, wall, floating noise) and delete them. Then it's advisable close holes, check that the mesh is “watertight” (without openings) and, if necessary, create a flat base by cutting the model with a plane or adding a small platform.
Another critical point is the model scalePhotogrammetric reconstructions don't use real-world measurements unless you provide references (scale markers, known distances, etc.). This means the model can be of an arbitrary size, and you'll have to adjust it in the slicer or the 3D software itself to fit the dimensions you need, especially if it's a functional part that has to fit your car, house, or any other object.
If your goal is to share the model digitally, platforms like Sketchfab They have become established as a benchmark for textured 3D scans. To upload a textured OBJ file, you need to package the OBJ itself, the MTL file, and the textures (usually PNG or JPG) into a zip file. However, it's advisable to optimize the size of these images, as they can be 30-40 MB each and will quickly exceed your storage or upload limits.
Other repositories like MyMiniFactory, with its Scan the World initiative, have become veritable libraries of 3D scanned models ready for printingespecially of sculptures and cultural heritage. Many of these models have been generated precisely with photogrammetry, so they also serve as a reference to see how far one can go with a good capture and careful post-processing.
The combination of a mobile phone with a good camera (preferably with ToF or LiDAR), a little patience when taking photos or scanning, and free tools like Colmap, Meshroom, Meshlab, Blender, or the aforementioned mobile apps, means that today anyone can digitize real-world objects and bring them to the 3D printer with a quality that, a few years ago, was reserved for professional equipment. Although the initial mesh usually needs some tweaking, if you dedicate some time to cleaning, scaling, and preparing the file, the result can compete head-to-head with many low-end 3D scanners and open up a huge range of personal and professional projects.