A phone 3D scan of a face takes about 20 to 40 seconds to capture and just under a minute to process, while an object scan on a recent iPhone takes three capture passes and a few minutes of processing on the device. Sending the photos to a cloud service instead takes from a few minutes for quick scans to 20-30 minutes for heavy projects, plus time in the queue.
Where you land depends on the scanning method, how much you capture, where the model is processed, and whether the first attempt is good enough to keep.
What happens during a phone 3D scan, and which part takes longest?
Every phone scan has three stages, and each one has its own timer.
- Capture. You move the phone around the subject while the app records photos, depth data or video. This is the part you control directly.
- Processing. The phone or a server turns that data into a mesh or a splat. You wait, and the wait grows with the amount of data.
- Cleanup and export. You trim the floor or table, fill gaps, scale the model and export it in the format the job needs.
For depth-based face scans, processing is the longer part. In a 2022 study in the Angle Orthodontist, 40 adults were scanned with two apps using the TrueDepth camera of an iPhone Xs. Capture averaged 20.28 seconds with Bellus3D and 40.34 seconds with Capture, and processing took 55.47 and 53.57 seconds. The reference stereophotogrammetry rig captured in 1.5 milliseconds and processed in about 21 seconds. The authors conclude that the phone apps need more accuracy from the operator and more compliance from the patient because of the longer acquisition time.
That last point is the real cost of a slow capture. A person who moves or turns their head during 40 seconds of scanning can leave you with a scan to redo, and every redo repeats both the capture and the processing.
For photo-based object scans, processing is a separate wait after capture, because the software has to match the images against each other before it can build a surface.
How long does a photo-based object scan take?
Photo-based scanning (photogrammetry) builds the model from ordinary photos, so its time is set mostly by the number of images. In Apple's WWDC21 session on Object Capture, Apple said that, depending on the object, 20 to 200 close-up images should be enough for good results, and that processing time depends on the number of images and the quality level.
Capture time is simply the number of photos multiplied by how long you take between shots. A plain mug does not need the same coverage as a carved figurine, so matching the image count to the object is the first and easiest way to save time. Our guide on how many photos a photogrammetry scan needs covers this in more detail.
Coverage also costs passes. In its WWDC23 session on Object Capture for iOS, Apple recommends three scan passes: if the object can be flipped, scan it in three orientations so every side is seen; if it cannot, capture it from three different heights. Once the passes are done, reconstruction runs on the device, and in Apple's words a USDZ model is ready "in just a few minutes". Apple lists this feature for iPhone 12 Pro, iPad Pro 2021 and later models.
Does processing on the phone or in the cloud take longer?
Each route has its own delays.
- On the phone there is no upload and no queue, but the device sets a ceiling. Apple notes that an Object Capture session by default stops capturing images when the device's reconstruction limit is reached.
- In the cloud you add upload time and a queue. KIRI Engine's FAQ says processing depends on the amount of data uploaded, that more photos or longer videos take longer, and that queue time depends on how many users around the world are processing at the same moment. Its Pro users get priority queues, which shorten the wait before processing starts.
If you need the model in front of a customer within the hour, on-device processing is the predictable choice. If you need more images than the phone can reconstruct, Apple's framework lets an app keep capturing past that limit so that a Mac reconstructs the larger set, which adds a transfer and a second processing step to the job.
How do LiDAR and Gaussian splat scans affect scan time?
A depth sensor measures distance directly, so the app does not have to work out every point from overlapping photos. The TrueDepth face scans in the 2022 study are depth scans of this kind, and their capture took 20 to 40 seconds. On iPhone, Object Capture uses both kinds of data: Apple says it saves LiDAR data with the photos, which lets it reconstruct low-texture objects that photos alone struggle with. Speed is not the only difference between the two; our comparison of LiDAR and photogrammetry explains which to pick for which job.
Gaussian splats are photo-based but are trained into a cloud of soft, overlapping blobs rather than a mesh. Scaniverse can do this on the phone itself: its App Store listing describes unlimited on-device Gaussian splat capture and processing, and the notes for version 5.1.0, released on 10 June 2025, added a heads-up when splat scans get too long, over 180 seconds, because they risk processing failure. Our explainer on Gaussian splatting on iPhone shows how to make one.
That warning is a useful rule well beyond one app: a longer scan is not a better scan. Once every surface is covered, extra footage only adds processing time.
How do the methods compare on time?
| Method | Capture | Processing | What slows it down |
|---|---|---|---|
| Depth face scan (TrueDepth apps) | 20.28 s and 40.34 s on average in the 2022 study | 55.47 s and 53.57 s in the same study | The subject moving, an unsteady operator |
| Photo-based object scan on iPhone (Object Capture) | Three passes recommended | A few minutes on the device | More images; capture stops at the device's reconstruction limit |
| Photo-based object scan in the cloud (KIRI Engine) | Set by the number of photos or the video length | A few minutes to 20-30 minutes, plus the queue | Upload size, how many users are in the queue |
| Gaussian splat on the phone (Scaniverse) | Warned above 180 s | On the device | Long scans risk processing failure |
Read the table as a range, not a promise. Your own scan will sit somewhere inside it, depending on the object and how cleanly you capture it.
Why does a phone scan take longer than planned?
Often the lost time is not processing but the second and third attempt.
- Moving too fast. Blurry images are hard for the software to match, and the model comes out soft or broken. Apple's iPhone guidance is to move around the object slowly and smoothly with a steady camera, and its guided capture stops and asks you to slow down if you move too fast. See why 3D scans come out blurry for the fixes.
- Shiny, transparent or plain surfaces. Apple advises picking objects with enough texture detail, warns that textureless or transparent regions may lack detail, and recommends diffused lighting for reflective objects. Ignore that and you end up scanning again.
- Skipping the underside. A model with a missing base sends you back for the extra pass Apple recommends.
- Too much data. Processing time grows with the number of images, so hundreds of photos of a simple object cost time that the model may not need.
- The wrong quality level. For Object Capture, Apple's 2021 session describes a Preview level that is very low quality but created the fastest, Reduced and Medium levels optimized for web and mobile experiences such as AR, and Full and Raw levels for games and post-production. Since processing time depends on the quality level, a Full or Raw model for a web viewer is time you do not need to spend.
- A busy server. On cloud apps the queue comes before processing, and KIRI Engine notes that queues get longer during busier hours.
How do you make a phone scan faster without losing quality?
Plan the job before you press record. Decide what the scan is for, because that decides how much detail you need.
- Set up the space. Use even, diffuse light and, for a plain object, a textured background; Apple recommends both for Object Capture. Good light is the cheapest way to avoid a redo.
- Plan the passes. One orbit at mid height, then the flip or the high and low passes. Know where you will stand before you start.
- Move slowly and steadily. A slow first attempt beats a quick one you have to repeat.
- Match the image count to the object. Simple shapes need fewer images, fine detail needs more, and nothing needs footage of the same side twice.
- Pick the quality level for the destination. A product viewer or an AR preview does not need the heaviest model.
- Process where it is predictable. On the phone when you are on site; on a computer or in the cloud when the image set is larger than the phone will take.
For many logistics and e-commerce jobs the real goal is a size, not a showpiece model. In that case a scanner that is also a measuring tool, such as MagiScan on iPhone or Android, covers the scan and the measurement in one app. If you do need the model, export it straight to the format of the job: 3D printing, AR or an online store.
How much time should you set aside?
For a face, plan for under a minute of capture and about a minute of processing, plus time to check the result while the person is still there. For a small or medium object, plan for three passes and a few minutes of processing on the phone, or longer on a cloud service at busy times. For large objects, rooms and splats, keep each pass short rather than recording one very long take.
Above all, budget for one redo and work to avoid it. MagiScan turns an iPhone or Android phone into a 3D scanner and measuring tool and exports scans for 3D printing, AR and e-commerce, but with any app the clock is mostly set by the capture habits above.
Frequently Asked Questions
How long does it take to 3D scan a face with an iPhone?
In a 2022 study of 40 adults, two TrueDepth apps on an iPhone Xs needed about 20 and 40 seconds on average to capture a face and about 54 to 55 seconds to process it. The person has to stay still for the whole capture, so a calm subject and a steady operator matter as much as the app.
Why is my 3D scan taking so long to process?
Processing time grows with the number of images or the length of the video (KIRI Engine) and with the quality level you choose (Apple). On cloud apps you also wait for the upload and for a place in the queue, which depends on how many other users are processing at the same time.
Is a longer scan a better scan?
No. Once every surface is covered, extra images or footage only add processing time, and very long captures can fail. Scaniverse, for example, warns users when splat scans run over 180 seconds for that reason.
Is it faster to process a scan on the phone or in the cloud?
On the phone you skip the upload and the queue, and Apple says an Object Capture model is ready in a few minutes on supported iPhones. A cloud service adds the upload and the queue, and KIRI Engine quotes from a few minutes for quick scans to 20-30 minutes for heavy projects.