K-FACE Demo — wireframe avatar

K-FACEDEMO

A wireframe head that mirrors your face in real time.

Demonstration version — nothing you do here is collected.

What it does

Reads your expression 30 times a second and copies it — blinks, winks, smiles, jaw, brows, gaze, head turn.
Takes typed commands: emote surprised, wink left, say "hello there".
Gives you 52 sliders for direct control of every individual muscle channel.
Publishes a machine-readable description of the face — the part a downstream AI would consume.

Getting started

Press Continue — the camera starts automatically.
Your browser will ask for camera permission — choose Allow.
First run downloads the tracker (~15 MB), so give it a few seconds.
Face the screen with reasonable light on you.

Prefer not to use the camera? Continue without the camera opens the same app driven by typed commands and sliders. You can switch between Camera and Manual at any time in the Source panel, top right. If your head turns the wrong way, flip the invert checkboxes in that panel.

This is a demo. No data leaves this page. Face tracking runs entirely inside your own browser — no video, no image and no measurement is uploaded, recorded or stored, by this site or by anyone else. Close the tab and nothing remains. The camera switches off the moment you set the source back to Manual.

What the full system adds Production versions of K-FACE do capture the expression stream and run AI analysis on it, and include a reliability mode that plays reference videos with known expressions and scores the tracker against them, so accuracy can be measured rather than assumed. Neither is running here. Any deployment that collects data will say so plainly, on screen, before it starts.

don't show this again

Works in Chrome, Edge, Firefox and Safari on a desktop with a webcam.

What it does

Getting started

K-FACE Script