LAPLACE Persona Tracker is LAPLACE’s iOS face tracking app, which you can use in place of iFacialMocap or Facemotion3d. LAPLACE maintains the phone app, desktop app and transmission protocol
The app sends all 52 ARKit blendshapes, plus head rotation and position. Input derivation, eye processing and parameter mapping run on the desktop. See How Tracking Reaches the Model
Requirements
You need an iPhone or iPad that supports ARKit face tracking and runs iOS 17 or iPadOS 17 or later. The app checks whether the device supports face tracking. Connect the phone and computer to the same local network
Setting It Up
- Under Face Tracking in Persona, add a LAPLACE Persona Tracker source with Add Face Source…, tick the avatars it should drive under its Assign Avatars, then turn on Enable Face Tracking
- Open the app on the phone and tap Connect. It lists the computers that answered on this network — tap Use on yours. iOS asks for Local Network permission the first time
- Sit the way you stream and tap the reticle button (Calibrate) to make that pose neutral. Reset Calibration, beside it, undoes that
Once connected, the first button becomes Change, which is where you switch computers or Disconnect
If the network blocks discovery broadcasts, the Connect sheet takes a typed address — the same one the Local IP Address card lists in Persona's tracking section
The selected computer and calibration are saved between launches. Calibration uses the phone’s position as a reference; recalibrate after moving the phone
Ports
This source has no Port field: a single fixed UDP port, 49700, carries both directions. Several feeds on one machine are separated by Tracker IP instead, the same as with any other face source — give each source the address of its own phone and two people drive an avatar each
Tracking Data and Calibration
iFacialMocap stays supported and the two can run side by side. The protocol has the following features:
- Float weights. iFacialMocap quantizes every blendshape to an integer 0–100. Persona applies gain to some of those signals — 5× on the mouth-frown signal — so a 1% input step may become visible in the model’s expression. This protocol uses
Float32 - All 52 blendshapes in every frame, in a fixed order rather than a name-per-segment text parse
- Automatic computer discovery. The phone lists the computers that replied for you to pick from instead of asking you to copy an address across, and a typed address still works when broadcasts are filtered
The protocol sends only the pose and 52 raw values. Derived inputs are calculated on the desktop, so changes to those formulas or direction conventions can be delivered in a desktop update
Settings on the Phone
Open the gear menu in the top-right corner for these settings:
| Setting | What it does |
|---|---|
| Save Energy | Darkens the screen and pauses the preview a few seconds after the last touch, while tracking keeps streaming. Tap to wake |
| Normalized Face Mesh | Draws your expression on a generic face instead of the shape of your own — for when the phone itself is on camera. The transmitted tracking data is unchanged |
The preview draws a wireframe face rather than the camera feed: enough to see that tracking is live, without putting your room on screen. The camera feed never leaves the phone
Last updated on September 19, 2026