How this works

1. Your camera tracks the position of your irises using on-device machine learning (MediaPipe). No video ever leaves your browser.
2. You'll look at 9 dots, one at a time. We record where your irises point for each dot.
3. From those samples we build a personal gaze map and score its accuracy.
4. In the live demo, you move a cursor with your eyes and select phrases with a left-eye wink — the phrase is then spoken aloud.