An iPhone LiDAR scanner brought depth to everyday scans

↗

Depth scanning moves from tablet to pocket

Apple first brought a LiDAR Scanner to a mainstream device with the iPad Pro, announced on 18 March 2020; the press release states the sensor "measures the distance to surrounding objects up to 5 meters away" and "operates at the photon level at nano-second speeds," working indoors and outdoors. Seven months later, on 13 October 2020, Apple carried the same sensor class into a phone: the iPhone 12 Pro announcement describes a LiDAR Scanner that measures "light distance" using "pixel depth information of a scene," and credits it with a sixfold improvement to autofocus speed in low light alongside a new Night mode portrait feature. Both releases frame the sensor primarily around augmented reality and photography, not around 3D scanning as an end in itself.

What the sensor feeds into, mechanically

Apple's ARKit documentation shows what that raw depth signal becomes for a developer. The ARMeshGeometry reference describes a live scene mesh built from vertices, faces formed from every three connected vertices, and outward-facing normals, with each face additionally carrying a surface classification such as wall, floor or table where ARKit can determine one. This is a fundamentally different kind of output from the multi-photograph photogrammetry pipelines used elsewhere in this production chain: LiDAR-based mesh reconstruction updates continuously, in real time, directly from the sensor's own measurements, rather than being computed afterwards from a batch of still photographs. That makes it well suited to what the iPad and iPhone releases actually emphasise, placing virtual objects convincingly and letting real objects occlude them correctly as a person moves the device, and poorly suited, on the documentation's own terms, to producing the kind of fine surface detail a full photogrammetry detail tier is built for.

A documentation gap worth noting

One detail is worth flagging precisely because it is easy to assume rather than read: the ARMeshGeometry reference itself states only iOS 13.4 and iPadOS 13.4 as minimum requirements and does not, in the retrieved text, explicitly state that LiDAR hardware is required to obtain mesh data. In practice, dense real-time scene meshing of this kind depends on the sensor Apple introduced in these same two 2020 announcements, and the class's practical availability tracks LiDAR-equipped hardware; but that link is an inference from the surrounding product history, not a sentence the API reference itself asserts, and it is the kind of gap a specification review should catch rather than assume.

Questions to bring to a demo

• Is the displayed scan raw LiDAR mesh data, or has it been refined afterwards with photogrammetry or manual retouching?
• Does the device and OS version shown actually carry the LiDAR Scanner, or is the mesh being simulated in software?
• What is the scan being used for, real-time placement and occlusion, or a finished asset, since the sensor suits the first far better than the second?

The LiDAR Scanner's documented contribution is a live, coarse, classified mesh of a room, not a finished 3D asset; what a display or AR session does with that stream is a separate, and separately documented, step.