Facial Recognition Login Block for React

No. The viewfinder is a drawing and the prompts are timers: the page never calls getUserMedia, never reads a frame and never stores anything, and the result says so. In a product you would ask for the camera only after the person presses Start, show the live video in place of the drawing, and send frames to a liveness check on your server or from a vendor, keeping nothing longer than the check needs.

Start sets the first prompt and a chain of four 800 ms timers advances it. The ring is drawn again for each prompt so it fills once, the face only changes classes (the head and features slide a little, the eyes close for the blink), and the cells show where you are. Cancel clears the timers, resets the drawing and returns focus to the start button, which then reads Try again. Leaving the page drops a pending timer.

Each prompt lasts 800 ms, too short to read aloud in turn, so the polite status line says once that the check-in started and what the four prompts are, and again when it ends. The cells and the line under the viewfinder show each prompt for people who can see them. The result's heading takes focus.

No. Face checks fail for people with different faces, for poor light and for anyone who can't or won't use a camera, and some places regulate biometric data. The quiet "Use your badge code" link is the other way in; point it at your own route, and ask for consent before a real camera opens.

The ring appears filled rather than sweeping, the drawn face changes pose without sliding, and the prompts still advance on their timers. The entrance on load is switched off.

The viewfinder is 224 px, the cells show short labels (Straight, Left, Right, Blink) and the Start button is on the first screen. Everything else is the same column.