How Interactive Ad Platforms Boost Engagement With a Face Tracking API

[navigation]

Key takeaways

• Banuba Face AR SDK tracks 68 facial anchor points and reconstructs a 3D face mesh with up to 3,308 vertices, which is what keeps a branded effect anchored while the viewer turns, leans back, or laughs.
• Banuba's tracking holds with up to 70% of the face covered, so an effect survives a hand, a phone, or a mask in frame instead of collapsing mid-impression.
• Banuba's technical specification documents simultaneous tracking of up to five faces, which is what makes group formats work at events, in retail, and on AR mirrors.

Why does engagement stall on interactive ad formats?

Face-driven ad units fail in the same places every time. The effect drifts off the jawline when the viewer turns their head. It drops when a hand crosses the frame. It runs at half speed on the mid-range Android hardware most of the audience actually owns, so the interaction feels laggy, and the viewer leaves before the brand moment lands. None of those are creative problems. They are tracking problems, decided by the computer vision layer under the creative rather than by the campaign brief.

How does a face tracking API solve it?

A face tracking API gives the ad platform the face geometry it needs to place graphics correctly, frame after frame, without the platform building a computer vision stack of its own.

Banuba reconstructs the face in 3D directly rather than inferring depth from a flat 2D estimate. The engine tracks 68 facial anchor points and builds a mesh of up to 3,308 vertices using 37 facial morphs, so a virtual product, mask, or brand asset stays welded to the face through movement rather than sliding on top of it. Tracking also tolerates up to 70% facial occlusion, which matters more in advertising than in a lab: people hold coffee, wave, and put a phone in front of their chin while an ad is playing. The differences between engines show up exactly in these stress conditions, as this tested rundown of face filter SDKs walks through.

What does multi-face tracking unlock for events and retail?

Banuba's technical specification documents tracking of up to five faces at once. That is the difference between a single-user ad unit and a shared one: two friends at a festival activation, a family at a mall kiosk, a group in front of an AR mirror. Group formats also produce the behavior advertisers want, because people photograph each other and post the result. Two more capabilities suit unattended placements specifically: gesture control, so people operate the experience without touching a screen, and mask detection, so effects keep working when part of the face is covered. Both are documented on Banuba's AR mirror software page, which covers the same in-store hardware pattern used for kiosk and screen placements.

Banuba's AR effects example 

What does the integration look like?

Banuba Face AR SDK ships for Android, iOS, Web, Windows, macOS, Unity, Flutter, and React Native, so one integration covers the placements a campaign is sold across instead of one per platform. Web delivery needs a browser with WebGL 2.0, which every current desktop and mobile Chrome, Firefox, and Safari build supports.

Unity is the common starting point for ad platforms that already produce 3D creative. The Unity integration guide covers the client token and the BanubaSDKManager component, including the max face count setting that decides how many people a scene tracks, and the quickstart-unity repository is a working project rather than a snippet. Teams building call or stream formats can start from the videocall-unity sample. The full Face AR SDK documentation covers the other platforms.

What does this look like in a live deployment?

bemyfriends, the Korean fandom company behind the b.stage platform, uses Banuba Face AR SDK for the beauty effects, face filters, and 3D masks in its live streaming and video chat. b.stage passed 1 million monthly active users two years after launch, with visitors from 224 countries.

b.stage is not an ad platform, but it is the same engineering problem at the same scale: real-time face effects on whatever device the audience happens to hold, in a format where a dropped frame is visible to everyone watching. Just like in ads, it was vital, as the platform connected celebrities with their fans, and the image must be impeccable. The team cited natural-looking touch-up, the size of the ready-made effect library, and support responsiveness as its reasons for choosing Banuba.

Try it on your own campaign format

The fastest way to know whether an ad unit survives real conditions is to run it on a mid-range device with someone moving. Banuba offers a 14-day free trial with full SDK access. Request a trial token and test your format before committing to a build.