The customer sits down, bends the hug arm into the guide, and the frame glows when the main zones look aligned. It keeps the process simple before backend AI does the clean cushion cutout.
Tiny light-up mode: it samples face, torso, elbow, arm channel and dropped-hand zones from the live video. This is not full skeleton AI, so it stays fast and below the 3MB website limit.
Use this if camera is blocked.