A scanned Persona stands in for a face a headset hides

What the record documents

Apple announced that Vision Pro would go on sale in the United States on 2 February 2024 in a newsroom statement published 8 January 2024, which describes Persona as an authentic spatial representation of an Apple Vision Pro user that enables others on a call to see their facial expressions and hand movements, created in just minutes using Vision Pro with machine learning, and notes the feature was available in beta at launch. Apple's support documentation, retrieved 16 September 2026 as a living guide, sets out the capture step in more detail: the wearer is instructed to take off the headset, hold it facing themselves at eye level, and follow spoken prompts through a set of expressions and head turns, with a hands-free option to prop the device up instead.

What creates the effect, and why a headset call needed one

The mechanism follows directly from that capture step: a Vision Pro cannot photograph its wearer's face while being worn, because the wearer's face is covered by the device. Apple's answer is to scan the face once, beforehand, with the headset held outward, then animate that scan in real time during a call using the headset's internal sensors while it is being worn, standing in for a live camera image that is physically impossible to capture at that moment. A companion guide describes spatial Persona, which places up to five participants' Personas around the wearer so they can move around in space and be repositioned by looking and pinching, rather than appearing in a flat video tile.

Why this became an avatar call, not a video call

This is the constraint worth naming plainly: a Persona is a driven likeness built from a one-time scan, not a live optical image of the wearer's current face, in the same sense that a Codec Avatar or a Holoportation capture is not a live volumetric feed of that instant. Apple's own materials do not claim otherwise; the newsroom statement calls it a spatial representation, not a video stream. Anyone comparing a spatial Persona call to a matched-room telepresence system or a light-field booth is comparing a driven-avatar mechanism to a live-capture one, and the two rest on different evidence: one on how faithfully a pre-scan tracks live expression, the other on how faithfully a camera rig reconstructs a body at that instant.

Questions to bring to a demo

• How old is the underlying face scan, and does it update as someone's appearance changes.
• What happens to the Persona's accuracy for an expression the original scan did not capture.
• Does the call fall back to a flat video tile for participants without a compatible headset.

The record here is Apple's own: a scan taken once, held outward, and a live drive of that scan during the call itself. That is a coherent and disclosed mechanism, and it is worth stating precisely because headset call and video call are not the same claim, however similar they may look in a demo.