Facebook says its 3D Photos feature, which retroactively converts 2D photos to 3D using neural nets, now works on photos taken with single-lens cameras
Facebook just expanded 3D photo posting to phones that don't actually capture depth data. Using the magic of machine learning …
Context & Ripple Effects
When Facebook launched 3D Photos in late 2018, the feature was gated on hardware: it only worked with iPhone portrait-mode shots because those carried real depth data from a second lens. The new update severs that dependency — neural nets now infer depth from any ordinary single-lens photo, building on research like the AI tool that reconstructed 3D faces from flat front-facing images two years earlier.
The move completes Facebook's longer push toward richer feed media, which began with panorama uploads from 360 cameras and native 360-degree photo support, then extended to glTF 2.0 3D objects dropped directly into News Feed. What required special cameras in 2016 and dual-lens phones in 2018 now runs on the installed base of every camera people already own.
First-order effects
- Users with single-lens phones — previously excluded — can now post 3D photos, expanding Facebook's addressable audience for the feature from dual-camera owners to essentially its entire user base.
Second-order effects
- Dual-lens and depth-sensor hardware loses part of its differentiation for social sharing, since the platform effect Facebook wants can be synthesized in software; camera makers' depth features matter less where feeds are concerned.
Third-order effects
- Depth capture is shifting from a sensor problem to an inference problem: if neural nets reliably reconstruct 3D from flat images, dedicated depth hardware becomes optional infrastructure, and platforms that control the inference layer gain leverage over what devices users actually need.
The trend: Computational perception is replacing dedicated capture hardware, letting software platforms manufacture sensor data — like depth — that cameras never recorded.