Persona (visionOS)
More actions
A Persona is the realistic digital likeness of a user that Apple's Apple Vision Pro headset generates from a face scan and animates in real time, so that other people on a video call see the wearer's face, expressions and hands instead of a headset. The user captures a Persona by holding the headset in front of the face while it records a short series of head turns and expressions; the headset then drives the likeness with its eye, face and hand tracking sensors during calls.[1][2] Apple introduced the feature with the headset in June 2023 and describes a Persona as "a digital representation of themselves created using Apple's most advanced machine learning techniques".[3]
Personas appear in FaceTime and in any other app that uses the Persona Virtual Camera, and the same capture personalizes the eyes shown on the headset's outward-facing EyeSight display.[1] Since visionOS 1.1, FaceTime calls between Vision Pro users can use spatial Personas, which place up to five participants as three-dimensional figures in the room around the user rather than inside flat video tiles.[4][5] The feature shipped as a beta in February 2024 and drew heavy criticism as uncanny; visionOS 26, released in September 2025, replaced it with a rebuilt version that Apple says uses volumetric rendering, and that version is no longer labeled beta.[6][7][8] An Apple executive told CNET in October 2025 that Personas use Gaussian splatting to create their 3D facial scans.[9]
This article covers Apple's Persona feature. For avatars in general, see Avatar; for Meta's comparable research and products, see Codec Avatars and Hologram (Meta).
How it works
Capture
A Persona can be captured during the initial setup of Apple Vision Pro or later from the Persona section of the Settings app. After refining the hand setup and watching a tutorial video (which can be skipped if it has been seen before), the user takes the headset off, holds it by the frame at eye level with the front facing the face, and follows spoken instructions to turn the head in different directions and make a series of expressions. The front display shows the same prompts visually: directional arrows for head turns, four rings for a closed-mouth smile, three rings for a big smile showing the teeth, two rings for raised eyebrows, one ring for briefly closing the eyes, and no rings as the signal to put the headset back on. An expression the user cannot perform can be skipped with a double-click of the Digital Crown.[1]
Apple's guide advises capturing in front-facing, even light against a simple, neutral background with no other people behind the user, and cancels a capture that has not started within 15 minutes. A Hands-Free Capture option lets the user place the headset on a surface at eye level or have another person hold it.[1] In its launch review, UploadVR reported that the process took "around a minute in total".[2] Jeff Norris, senior director of Apple's Vision Products Group, told CNET in 2025: "We only need a handful of images when we are enrolling your Persona. That includes a few facial expressions to help our networks understand how your face moves when you're talking and smiling."[9]
After capture the user can adjust lighting (for example Studio, or Contour for stronger shadows), color temperature and brightness, choose a pair of virtual glasses, and pick a default background of a color, an Environment or a photo for apps other than FaceTime. The Persona can be recaptured at any time, for instance after a change of hairstyle or outfit, and EyeSight is updated with it.[1] Norris said Apple was "focusing on just the one at a time right now", and that people who want a different look "can reenroll or just put on a different shirt and enroll again".[9]
Generation and animation
Announcing the feature at WWDC in June 2023, Mike Rockwell, vice president of Apple's Technology Development Group, said that after enrollment with the headset's front sensors the system uses an encoder-decoder neural network, a kind of machine learning model, to create the Persona, and that "this network was trained on a diverse group of thousands of individuals".[10] Once generated, the Persona is driven in real time by the headset's eye and face tracking sensors together with hand tracking.[2] Road to VR noted that the face data comes from the headset's downward-facing cameras, which need a line of sight to the mouth, cheeks and eyes, and that all processing happens on the device.[11]
For the visionOS 26 version, Apple says the new Personas take "advantage of industry-leading volumetric rendering and machine learning technology" and are "still created on device in a matter of seconds".[7] In the CNET interview Norris said the technology uses Gaussian splatting to produce the 3D scan of the user, and described the machine learning side as "a concert of networks that come together", adding: "We counted them up, it's over a dozen, but we actually reduced the number when we moved to this new version of Personas." He said the capture stays on the headset: "right now, we love that it's self-contained to the device and that all the processing happens on the device. None of these images have to go anywhere in order for that to happen."[9] Apple's announcement of the Apple Vision Pro M5 in October 2025 stated that the M5's 16-core Neural Engine runs system features "like capturing a Persona" up to 50 percent faster than on the previous generation.[12]
Storage and privacy
According to Apple's user guide, a Persona is stored encrypted on the device and requires Optic ID or the device passcode to access; turning Optic ID off after it was on deletes the Persona. A Persona cannot be captured or edited while Travel Mode is on or during a FaceTime call, and it is not available to people using the headset through Guest User.[1]
History
Apple announced Personas together with Apple Vision Pro and visionOS at WWDC on 5 June 2023, presenting them as the way a Vision Pro wearer appears on a FaceTime call, with face and hand movements reflected in real time.[3] When Apple confirmed the US launch date of 2 February 2024, it said a Persona "can be created in just minutes", that Personas also work in the third-party videoconferencing apps Zoom, Cisco Webex and Microsoft Teams, and, in a footnote, that "Persona is available in beta".[6]
| Software release | Date | Persona changes |
|---|---|---|
| visionOS 1.0 | 2 February 2024 (US launch of Vision Pro) | Persona launched as a beta, shown in FaceTime and other video apps as a two-dimensional feed.[6][13] |
| visionOS 1.1 | 7 March 2024 | Release notes: "Adds an option to enroll your Persona hands-free", "Improves hair and makeup appearance" and "Improves neck and mouth representation"; also improved rendering of the eyes for EyeSight.[14][15] |
| visionOS 1.1 (feature activation) | 2 April 2024 | Spatial Personas enabled for FaceTime calls between Vision Pro users, for up to five participants.[13][4] |
| visionOS 2 | 16 September 2024 | Release notes: "Create Personas with more accurate skin tones, fluid and natural hand movements with effects, improved vibrant clothing colors, and more eyewear choices"; spatial Personas usable with SharePlay in the Photos app; developers can define custom spatial Persona templates.[16][17][18] |
| visionOS 26 | Announced 9 June 2025; released 15 September 2025 | "All-new Persona" with "enhanced hair, eyelashes, complexion and movement"; full side profile view; capture flow with new glasses styles and materials (over 1,000 glasses variations) and a spatial preview before saving; remote participants can join co-located shared experiences as Personas; no longer labeled beta.[7][19][20][21][8] |
For visionOS 26, Apple said Personas were "transformed to feel more natural and familiar", with "striking expressivity and sharpness, offering a full side profile view, and remarkably accurate hair, lashes, and complexion".[7] Road to VR's Ben Lang, reporting on the announcement, wrote that the earlier Personas were already "the most lifelike real-time avatars available on any headset today" but had "always felt somewhat blurry or ghostly".[21] Apple confirmed to Road to VR that the improvements also carry over to the Persona shown on the external EyeSight display.[11] UploadVR reported in June 2025 that Greg Joswiak, Apple's senior vice president of Worldwide Marketing, had publicly shared a clip of his own visionOS 26 Persona, the first time an Apple executive had done so, and that the feature was "no longer described as beta".[8]
UploadVR reported that Apple also uses Gaussian splatting for the spatial scenes feature of visionOS and that the technique is "behind the massive upgrade in visual realism we saw from Personas in visionOS 26". In the same June 2026 report on visionOS 27, which Apple released on 14 September 2026,[22] UploadVR wrote that RealityKit adds native support for rendering Gaussian splats, so developers no longer have to implement it themselves.[23]
Uses
FaceTime and video apps
On a FaceTime call a Vision Pro wearer appears as a Persona, while participants on a Mac, iPad or iPhone appear in tiles; Spatial Audio makes each voice come from the position of that person's tile.[6] Apps other than FaceTime receive the Persona through the Persona Virtual Camera.[1] In practice this means a Persona appears in iPad and iPhone apps that request the selfie camera;[2] in The Verge's review, Nilay Patel reported using his Persona on a Google Meet call in Safari.[24]
A spatial Persona requires at least one other participant on the call to be using Apple Vision Pro and supports up to five participants. Apple's current guide says the feature is on by default in such calls. Participants can reposition a Persona by looking at it and pinching and dragging, and Personas move with a shared app window. Turning spatial Persona off from Control Center keeps the call going but returns the other Personas to a tile in that user's view, and that user's Persona appears in a tile for the others; other participants are not affected unless only one other person was on the call.[5] Spatial Personas appear in the user's own space, in passthrough or in a virtual Environment, with a shared spatial context, so that windows and volumes appear at the same positions relative to every participant.[13] Their arrangement depends on the activity: when participants watch a film together with SharePlay, the other Personas appear beside the viewer, as in a cinema, while in a game they may appear across from the viewer, as at a table.[5] Brian Heater of TechCrunch listed work collaboration, watching films and playing games as uses at launch.[4]
Co-located and professional use
visionOS 26 added shared spatial experiences for Vision Pro users in the same room, and Apple said remote participants can be added "from across the world via FaceTime".[7] Road to VR described the result as a mix of in-person headset users and remote participants appearing as Persona avatars in the same virtual experience.[21] Norris told CNET that Apple was hearing about business uses in healthcare, such as doctors who create procedures and want to train others without travelling.[9] UploadVR also reported that a macOS 27 API, which lets Mac apps show and edit 3D objects on a connected Vision Pro running visionOS 27, supports SharePlay "for collaborative spatial editing with Personas".[23]
Developer support
Third-party apps place spatial Personas through SharePlay and the GroupActivities framework. The framework offers built-in spatial templates: side-by-side, where participants sit along a curved line facing the app (suited to video); conversational, an open circle with a gap for the app; and surround, a closed circle around a shared volume (suited to 3D modeling).[18] In the WWDC24 session on the topic, an Apple engineer explained that the app keeps its interface synchronized for all participants while "FaceTime, in turn, is responsible for maintaining the shared spatial context by placing participants around your app in a consistent way."[18]
visionOS 2 added custom spatial templates built from seats, each of which "can hold a single spatial Persona", positioned relative to the shared app. Seats can be reserved for roles such as teams in a game and can face a chosen point. Apple recommended placing seats at least one meter apart and warned developers that moving seats or switching templates "can be very disorienting". By default an immersive space opened during SharePlay is private, so participants cannot see each other's spatial Personas unless the app opts in to a group immersive space. Xcode 16 added simulated FaceTime calls in the visionOS simulator for testing.[18]
Research and measurements
Network and rendering behavior
In a measurement study presented at the 2024 ACM Internet Measurement Conference (IMC), Ruizhi Cheng, Nan Wu, Matteo Varvello, Eugene Chai, Songqing Chen and Bo Han compared FaceTime, Cisco Webex, Microsoft Teams and Zoom on Vision Pro. They found that only FaceTime offered spatial Personas, while the other apps showed two-dimensional Personas rendered from a fixed viewpoint.[25] Their main findings in two-user calls were:[25]
| Measurement | Reported result |
|---|---|
| Throughput, FaceTime spatial Persona | About 0.7 Mbps (0.67 Mbps on average) |
| Throughput, FaceTime 2D Persona | About 2 Mbps |
| Emulated stream of 74 compressed face and hand keypoints at 90 frames per second | 0.64 plus or minus 0.02 Mbps |
| Emulated stream of compressed head meshes (about 70,000 to 90,000 triangles, no texture) at 90 frames per second | 108.4 plus or minus 16.7 Mbps |
| Triangles in a spatial Persona mesh (read with Apple's RealityKit tool) | 78,030 |
| Persona in peripheral vision (foveated rendering) | 73 percent fewer rendered triangles and 39 percent less GPU time per frame |
| GPU processing time with five participants | About 9 ms per frame, against an 11.1 ms frame deadline |
| Uplink bandwidth constrained to 0.7 Mbps | Spatial Persona unavailable, with "poor connection" shown |
Because the spatial Persona used less bandwidth than any 2D video Persona and far less than streaming a mesh, the authors inferred that FaceTime probably sends compact semantic data (such as facial and hand keypoints) and reconstructs the Persona on the receiving headset, although they noted this makes rate adaptation difficult. They noted that the spatial Persona primarily tracks the eye and mouth areas for facial expressions, as well as hand movements. They also found that FaceTime applies viewport adaptation, foveated rendering and distance-aware simplification when drawing Personas, but no occlusion optimization, and concluded that the GPU cost of rendering more Personas "explains why FaceTime currently supports a maximum of only five spatial personas".[25]
Related avatar research
Research on photorealistic avatars for telepresence predates Personas. Stephen Lombardi, Jason Saragih, Tomas Simon and Yaser Sheikh's "Deep Appearance Models for Face Rendering" (SIGGRAPH 2018) modeled facial geometry and view-dependent texture with a deep variational autoencoder learned from a multiview capture rig, and described the result as "naturally suited to real-time interactive settings such as Virtual Reality".[26] 3D Gaussian splatting, introduced by Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler and George Drettakis in 2023, represents a scene with anisotropic 3D Gaussians optimized from photographs and renders them in real time.[27] Academic work soon applied it to faces: GaussianAvatars (Shenhan Qian and colleagues, CVPR 2024) rigs 3D Gaussian splats to a parametric morphable face model to produce photorealistic head avatars controllable in expression, pose and viewpoint.[28]
Reception
UploadVR later wrote that the original Personas were "harshly criticized, and even ridiculed" at launch.[8] In The Verge's review, Nilay Patel called it a "deeply weird and extremely uncanny 3D persona system", listed "Personas are uncanny and somewhat terrifying" among the headset's drawbacks, and wrote that "there's a long way to go before using a persona on a call isn't distracting at best and, at worst, extremely rude".[24] UploadVR's David Heaney called Personas "incredibly impressive technology given how quickly they can be generated and how well they run on a standalone headset", but wrote that "they remain deep in the uncanny valley" and that reactions from people on the other end "will range from laughter to shock and disgust". He also criticized the confinement of Personas to a rectangular window, which lacked the shared coordinate space of social VR platforms.[2]
UploadVR's Ian Hamilton was more positive about spatial Personas, which addressed that last point. After a hands-on, he described them as "the closest I've felt to VR successfully collapsing space between people", while noting that "you always know you're looking at a digital human".[29]
Coverage of the visionOS 26 version described a large improvement. After trying it at WWDC25, Road to VR's Ben Lang wrote that "skin looks much more detailed" and that facial motion looked "nuanced and realistic" even when he made asymmetrical movements, and that "the sensation of Personas looking 'ghostly' is greatly reduced", although the hands still looked ghostly by comparison.[11] UploadVR called the update "a step-change in quality".[8] CNET's Scott Stein wrote in October 2025 that "no one is doing anything in VR or AR headsets or glasses as advanced as Apple's Vision Personas", while pointing out that the experience remained limited to owners of the expensive headset.[9]
Lang also argued that Personas depend on cameras having a clear view of the lower face, which becomes harder as headsets shrink toward goggle or glasses form factors.[11] When Meta announced its own realistic avatar system for calls at Connect in September 2026, UploadVR described it as "its take on Apple Vision Pro's Personas".[30]
See also
References
- ↑ 1.0 1.1 1.2 1.3 1.4 1.5 1.6 "Capture and edit your Persona on Apple Vision Pro". Apple Vision Pro User Guide. Apple. https://support.apple.com/guide/apple-vision-pro/capture-your-persona-dev934d40a17/visionos. Retrieved 2026-09-29.
- ↑ 2.0 2.1 2.2 2.3 2.4 David Heaney (2024-02-16). "Apple Vision Pro Review: A Heavy Portable Cinema & Monitor With A Promising Spatial OS". UploadVR. https://www.uploadvr.com/apple-vision-pro-review/. Retrieved 2026-09-29.
- ↑ 3.0 3.1 "Introducing Apple Vision Pro: Apple's first spatial computer". Apple Newsroom. Apple. 2023-06-05. https://www.apple.com/newsroom/2023/06/introducing-apple-vision-pro/. Retrieved 2026-09-29.
- ↑ 4.0 4.1 4.2 Brian Heater (2024-04-02). "Apple Vision Pro's Persona feature gets collaborative". TechCrunch. https://techcrunch.com/2024/04/02/apple-vision-pros-persona-feature-gets-collaborative. Retrieved 2026-09-29.
- ↑ 5.0 5.1 5.2 "Use spatial Persona on Apple Vision Pro". Apple Vision Pro User Guide. Apple. https://support.apple.com/guide/apple-vision-pro/use-spatial-persona-tana1ea03f18/visionos. Retrieved 2026-09-29.
- ↑ 6.0 6.1 6.2 6.3 "Apple Vision Pro available in the U.S. on February 2". Apple Newsroom. Apple. 2024-01-08. https://www.apple.com/newsroom/2024/01/apple-vision-pro-available-in-the-us-on-february-2/. Retrieved 2026-09-29.
- ↑ 7.0 7.1 7.2 7.3 7.4 "visionOS 26 introduces powerful new spatial experiences for Apple Vision Pro". Apple Newsroom. Apple. 2025-06-09. https://www.apple.com/newsroom/2025/06/visionos-26-introduces-powerful-new-spatial-experiences-for-apple-vision-pro/. Retrieved 2026-09-29.
- ↑ 8.0 8.1 8.2 8.3 8.4 David Heaney (2025-06-15). "Apple Executive Publicly Shows visionOS Persona For The First Time". UploadVR. https://www.uploadvr.com/apple-executive-publicly-shows-visionos-26-persona/. Retrieved 2026-09-29.
- ↑ 9.0 9.1 9.2 9.3 9.4 9.5 Scott Stein (2025-10-28). "Apple Vision Pro's Persona Avatars Are Evolving Fast". CNET. https://www.cnet.com/tech/computing/apple-talks-to-me-about-vision-pro-personas-where-is-our-virtual-presence-headed/. Retrieved 2026-09-29.
- ↑ Andrew Tarantola (2023-06-07). "Where was all the AI at WWDC?". Engadget. https://www.engadget.com/where-was-all-the-ai-at-wwdc-163048564.html. Retrieved 2026-09-29.
- ↑ 11.0 11.1 11.2 11.3 Ben Lang (2025-06-18). "Vision Pro's New Personas Are a Big Improvement, But How Will They Scale to Smaller Headsets?". Road to VR. https://roadtovr.com/apple-vision-pro-persona-visionos-26-face-tracking-smaller-future-headsets/. Retrieved 2026-09-29.
- ↑ "Apple Vision Pro upgraded with the M5 chip and Dual Knit Band". Apple Newsroom. Apple. 2025-10-15. https://www.apple.com/newsroom/2025/10/apple-vision-pro-upgraded-with-the-m5-chip-and-dual-knit-band/. Retrieved 2026-09-29.
- ↑ 13.0 13.1 13.2 David Heaney (2024-04-02). "Apple Vision Pro Now Has 3D 'Spatial Personas'". UploadVR. https://www.uploadvr.com/apple-vision-pro-spatial-personas-launch/. Retrieved 2026-09-29.
- ↑ Juli Clover (2024-03-07). "Apple Releases visionOS 1.1 With Improvements to Persona, EyeSight, Virtual Keyboard and More". MacRumors. https://www.macrumors.com/2024/03/07/apple-releases-visionos-1-1/. Retrieved 2026-09-29.
- ↑ "About visionOS 1 Updates". Apple Support. Apple. https://support.apple.com/HT214071. Retrieved 2026-09-29.
- ↑ Juli Clover (2024-09-16). "Apple Releases visionOS 2 With 2D Spatial Photos Feature, Persona Updates, Improved Guest Mode and More". MacRumors. https://www.macrumors.com/2024/09/16/apple-releases-visionos-2/. Retrieved 2026-09-29.
- ↑ "visionOS 2 brings new spatial computing experiences to Apple Vision Pro". Apple Newsroom. Apple. 2024-06-10. https://www.apple.com/newsroom/2024/06/visionos-2-brings-new-spatial-computing-experiences-to-apple-vision-pro/. Retrieved 2026-09-29.
- ↑ 18.0 18.1 18.2 18.3 "Customize spatial Persona templates in SharePlay (WWDC24 session 10201)". Apple Developer. Apple. 2024-06. https://developer.apple.com/videos/play/wwdc2024/10201/. Retrieved 2026-09-29.
- ↑ "About visionOS 26 Updates". Apple Support. Apple. https://support.apple.com/en-us/123024. Retrieved 2026-09-29.
- ↑ Juli Clover (2025-09-15). "Apple Releases visionOS 26 With Persistent Widgets, Lifelike Personas, and Spatial Scenes for Vision Pro". MacRumors. https://www.macrumors.com/2025/09/15/apple-releases-visionos-26/. Retrieved 2026-09-29.
- ↑ 21.0 21.1 21.2 Ben Lang (2025-06-09). "Vision Pro is Getting a Major Visual Upgrade to Its 'Persona' Avatars". Road to VR. https://roadtovr.com/vision-pro-persona-avatar-upgrade-visionos-26/. Retrieved 2026-09-29.
- ↑ Ryan Christoffel (2026-09-14). "visionOS 27 is here for Apple Vision Pro with these new features". 9to5Mac. https://9to5mac.com/2026/09/14/visionos-27-is-here-for-apple-vision-pro-with-these-new-features/. Retrieved 2026-09-29.
- ↑ 23.0 23.1 David Heaney (2026-06-14). "visionOS 27 Is A Much Bigger Update Than Apple's Keynote Suggested". UploadVR. https://www.uploadvr.com/visionos-27-announced-apple-vision-pro-wwdc-26/. Retrieved 2026-09-29.
- ↑ 24.0 24.1 Nilay Patel (2024-01-30). "Apple Vision Pro review: magic, until it's not". The Verge. https://www.theverge.com/24054862/apple-vision-pro-review-vr-ar-headset-features-price. Retrieved 2026-09-29.
- ↑ 25.0 25.1 25.2 Ruizhi Cheng, Nan Wu, Matteo Varvello, Eugene Chai, Songqing Chen, Bo Han (2024). "A First Look at Immersive Telepresence on Apple Vision Pro". Proceedings of the 2024 ACM Internet Measurement Conference (IMC '24). doi:10.1145/3646547.3689006. https://arxiv.org/abs/2405.10422. Retrieved 2026-09-29.
- ↑ Stephen Lombardi, Jason Saragih, Tomas Simon, Yaser Sheikh (2018). "Deep Appearance Models for Face Rendering". ACM Transactions on Graphics, vol. 37, no. 4 (SIGGRAPH 2018). doi:10.1145/3197517.3201401. https://arxiv.org/abs/1808.00362. Retrieved 2026-09-29.
- ↑ Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis (2023). "3D Gaussian Splatting for Real-Time Radiance Field Rendering". ACM Transactions on Graphics, vol. 42, no. 4 (SIGGRAPH 2023). doi:10.1145/3592433. https://arxiv.org/abs/2308.04079. Retrieved 2026-09-29.
- ↑ Shenhan Qian, Tobias Kirschstein, Liam Schoneveld, Davide Davoli, Simon Giebenhain, Matthias Nießner (2024). "GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians". IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2024). https://arxiv.org/abs/2312.02069. Retrieved 2026-09-29.
- ↑ Ian Hamilton (2024-04-03). "Apple Vision Pro Spatial Personas Hands-On: A Step-Change For Telepresence". UploadVR. https://www.uploadvr.com/apple-vision-pro-spatial-personas-hands-on/. Retrieved 2026-09-29.
- ↑ David Heaney (2026-09-24). "Hologram: Meta Is Finally Set To Ship Realistic Avatars, And We Tried Them". UploadVR. https://www.uploadvr.com/meta-hologram-realistic-avatars-meta-vr-glasses-connect-2026/. Retrieved 2026-09-29.