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
here's how it works Karissa Bell / Engadget : Facebook's 3D photos no longer require portrait mode Brittany A. Roston / SlashGear : Facebook taps AI to turn almost any 2D photo into a 3D image Stephen Shankland / CNET : Facebook's 3D photos expand to millions more phones thanks to AI Tweets: Robert Scoble / @scobleizer : Cool tricks with cameras. And my Tesla has eight cameras. Hmm. https://twitter.com/...
Context & Ripple Effects
When Facebook launched 3D Photos in late 2018, it was a hardware-gated feature: only iPhone portrait-mode shots with their dual-lens depth data qualified, as covered in the original portrait-mode rollout. The pipeline built on earlier work — Facebook had already added glTF 2.0 object rendering to News Feed posts and, separately, researchers had shown a neural net could infer 3D faces from flat front-facing photos in the face-to-3D research two years prior.
First-order effects
- The feature's addressable base jumps from dual-camera portrait shots to essentially any photo in a user's library — CNET's framing is 'millions more phones' — since the neural net synthesizes depth that the hardware never captured.
- Users on single-lens Android devices, previously excluded entirely, can now post 3D Photos without buying new hardware.
Second-order effects
- Dual-lens portrait mode loses one of its differentiating payoffs: if software can fake the depth map, the hardware advantage that justified premium camera arrays narrows for phone makers competing against Facebook's free conversion.
- News Feed fills with retroactively-3D'd legacy photos, deepening Facebook's investment in depth-capable media formats alongside its existing 360-photo and glTF object pipelines.
Third-order effects
- If depth becomes a software inference problem rather than a capture-time requirement, camera hardware specs stop gating immersive media — the same pattern Facebook established by accepting 360 uploads from any camera rather than requiring specific rigs.
- Platform-owned neural nets converting users' back catalogs position Facebook as the layer where flat media gains dimensionality, independent of what device took the picture.
The trend: Camera features are migrating from lens hardware to neural-network inference, letting platforms retrofit capabilities onto years of existing photos.