Capture guide

Field guide for scanning with an iPhone and getting the files onto the Windows box. The technique here is standard LiDAR practice written down for this pipeline; no real captures have been run through it yet, so treat it as the starting recipe and improve it as field experience comes in.

The capture apps

Five iPhone apps cover the today lane. All of them use the same LiDAR sensor; they differ in what they let you export and what they charge for it.

App Free Paid Best export for this pipeline
Polycam Capture and in-app viewing; export options on the free tier have narrowed over time Subscription unlocks the useful raw exports (OBJ, PLY, LAS, point cloud, DXF) OBJ or PLY if your tier allows, otherwise GLB
Scaniverse Capture and, historically, all exports free (OBJ, GLB, USDZ, STL, PLY, LAS) Nothing needed so far PLY point cloud plus an OBJ or GLB mesh
3d Scanner App (Laan Labs) Capture and mesh exports free (OBJ, GLB, USDZ, STL, PLY, point cloud) plus an “All Data” raw dump Some processing extras OBJ or PLY. The “All Data” package is not a raw layout the pipeline recognises yet
Stray Scanner Records raw depth frames, camera poses and RGB video; no meshing in the app Small one-off price historically The whole capture folder, unchanged. This is exactly the raw package the pipeline fuses natively
Record3D RGBD video capture with limited export Unlocks per-frame exports and USB streaming Only if you already use it: export PLY, or repack frames to the generic raw spec in formats.md

Honest note: pricing tiers and export menus change often and without notice. Verify in the app before a capture trip, and do not buy a subscription on the strength of this table.

Which format, in one line. A mesh (OBJ or GLB) plus a point cloud (PLY or LAS) when the app offers both; the pipeline prefers the app’s own mesh and always writes point clouds out. USDZ works only with the optional usd extra installed, so avoid it when another format is on offer. The full list the pipeline reads: OBJ, GLB, GLTF, STL, DAE, PLY, PCD, XYZ, LAS, LAZ, USDZ, and raw depth packages (Stray Scanner’s layout, or the generic spec in formats.md).

Start easy, and know what wrecks a scan

Accuracy is won at capture time. The processing can verify scale and clean noise, but it cannot invent detail a rushed capture never recorded. Four things wreck a scan, and a bare driveway manages all of them at once, so do not judge the system on a hard first target:

Easy mode for learning: a single object on a table, indoors, with other textured things around it, even soft light, and you moving slowly in full circles. Nail that first. An open sunlit driveway is genuinely hard mode.

Technique

The sensor is honest but short-sighted. It reads best between 1 and 3 metres, and the phone stitches your movement into one model as you go, so how you move is most of the result.

Per mode

Control measurements

The phone’s scale is usually within about one per cent, and the pipeline checks it rather than trusting it. Over a 3 metre span, one per cent is 30 mm, about two finger widths: enough to matter in a drawing. The check needs real tape measurements from the site.

At the site:

  1. Pick recognisable hard corners: table ends, slab corners, coping edges. Points you will be able to find again on a screen.
  2. Tape-measure two or more distances between those corners, longest span first. Long spans expose scale error; short ones hide it in noise.
  3. Write the distances down at the site: notes app, or a photo of the tape in place. Do not trust memory on the ferry home.

Back on the PC:

aura-scan pick scans\my-scan

A viewer window opens on the first raw input (name your main pass so it sorts first, for example 01-main.ply). Shift + left click the same corners in the same order you taped them, then close the window. The points land in manifest.yaml. Add the tape numbers against them:

control:
  distance_constraints:
    - {points: [0, 1], distance_m: 3.640, note: "coping to coping, longest span"}
    - {points: [2, 3], distance_m: 1.685, note: "table top, long edge"}

On the next aura-scan run, the pipeline compares scan distances against the tape, reports the error per constraint, and only rescales the geometry when the correction is bigger than the configured tolerance. The result is written into the metadata and stamped into the DXF text block either way, so a reader downstream can see exactly how much to trust the dimensions.

Getting files to the Windows box

Route Needs Notes
LocalSend (recommended) Both devices on the same wifi Free, open source, works fully offline, no account. Install from the Microsoft Store or localsend.org, share from the capture app to LocalSend on the phone.
USB cable Apple Devices app on Windows 11 (iTunes file sharing on older setups) Reliable for big raw packages. The capture app must expose its documents to file sharing.
iCloud for Windows Internet and iCloud storage space Save the export to iCloud Drive on the phone, let it sync down. Slowest and the only route that touches a cloud.

Whatever the route, the file ends its journey in scans\<name>\raw\.

Checklists

Every capture

Asset mode extras

Projection surface extras