[navigation]
Key takeaways
- Banuba Face AR SDK reconstructs a 3D face mesh with up to 3,308 vertices through direct 3D reconstruction, rather than inferring 2D points and lifting them into 3D.
- Banuba face tracking holds under up to 70% facial occlusion and across the full -90° to +90° head-angle range, which matters when users move or hold the phone badly.
- Banuba processes every frame on the device and stores no user data, so facial geometry never leaves the handset.
Why does facial measurement matter for health-tech apps?
Products in this space sell a change the user cannot verify alone. A facial-exercise program, a whitening treatment, a filler consultation and a frame fitting all depend on the customer trusting a difference they are being told about. Showing that difference on their own face, in real time, removes the trust gap.
The measurement also drives product logic. Aesthetic simulators need to know where the nose, chin and cheekbones sit before modeling an adjustment. Eyewear needs pupillary distance in millimeters to scale a frame. Facial-wellness apps need a stable reference across sessions so a 28-day comparison means anything.
One point deserves stating plainly: this is measurement for visualization, fitting and engagement, not diagnosis. Banuba treats pupillary distance output as an estimate for fitting support, explicitly separate from medical-grade prescription measurement, and health-tech teams should present it the same way.
How does a face landmarks SDK produce usable geometry?
A traditional face landmarks SDK detects notable features such as eyes and nose and tracks them frame to frame. Banuba's engine instead builds a face mesh and tracks 37 morphs across it, which is what keeps the geometry stable when conditions get difficult.
That gives a health-tech team three things. The SDK tracks 68 facial anchor points with sub-pixel accuracy and reconstructs a 3D mesh of up to 3,308 vertices, so downstream logic works from real facial structure rather than a flat point cloud. It holds tracking through up to 70% facial occlusion and across a -90° to +90° angle range, covering a user in a bathroom mirror rather than a capture booth. And it supports binocular pupillary distance measurement in millimeters, which the input eyewear fitting depends on.
Teams weighing this against a do-it-yourself computer-vision stack end up comparing accuracy under stress and long-term maintenance cost. This technical breakdown of Banuba Face AR SDK against OpenCV covers where the two approaches diverge.
What does the privacy story look like?
For a health-adjacent product, this is usually the first question legal asks. Banuba's face tracking does not store, process, or transfer personal data to a server. Everything happens on the user's device, which keeps facial geometry inside the handset and makes GDPR compliance a matter of configuration rather than a data-processing agreement. On-device processing with no cloud dependency is the most repeated request in Banuba's inbound technical inquiries.
What does the integration take to ship?
The SDK runs on iOS 13+ and Android 8.0+ with a 1280x720 camera recommended, and supports HTML5, Windows, macOS, Unity, Flutter, and React Native from the same engine. Landmark coordinates are read per frame through a frame-data listener and transformed from camera space into screen space, documented in the face landmarks guide. Working integrations ship as full sample projects for iOS and Android, so an engineer can run the tracking loop first.
Banuba's Face AR SDK for plastic surgery simulation example
Which products already run on this?
FaceYoga by Mental Growth, a Cyprus-based facial-wellness app, built its before-and-after feature on Banuba Face AR SDK to show users the expected result of 7 and 28 days of facial exercises before asking them to subscribe. It reached 20,000+ downloads and a 4.2 out of 5 App Store rating, with the team citing documentation quality, GitHub source samples, and no device overheating.
The same engine scales beyond wellness. A SaaS fandom platform that reached 1 million monthly active users in two years runs its live-video face-tracking layer on Banuba Face AR, serving users across 224 countries. Aesthetic and dental products sit in between, driving surgical-result simulation and whitening previews from the same landmarks.
Get started
Building facial measurement into a health-tech product? Start with the Face Landmarks SDK free trial and run the sample projects against your own target devices.
-
No. Banuba's facial measurement is built for visualization, fitting, and engagement, and Banuba explicitly separates its pupillary distance output from medical-grade prescription measurement. Present it to users as an estimate, not a clinical figure. In eyewear, that measurement feeds frame scaling in Banuba's glasses virtual try-on.
-
No. Banuba's face tracking does not store, process, or transfer personal data, and everything runs locally on the user's device, which is what makes it GDPR-compliant by default. The same on-device model applies across Banuba's Face AR SDK feature set.
-
Banuba tracks 68 facial anchor points, reconstructs a mesh of up to 3,308 vertices, and holds tracking through up to 70% facial occlusion and across a -90° to +90° head-angle range. Banuba publishes per-device figures in its technical specification, which are measured under lab conditions and worth re-testing on your actual target hardware.
-
Yes. Banuba offers a free trial of the Face AR SDK, and the documentation plus the public sample apps let an engineer validate tracking quality on your own devices before budget is committed.
