vladmandic/face-api ? reverse-engineered prompt

Build me a face recognition and analysis tool that works in both the browser and Node, using TensorFlow. I want to load an image or use a webcam, detect every face, show boxes and landmarks, and then let me compare faces between photos. It should also predict age, gender, and emotion, and handle face rotation or pose tracking well enough to follow faces in live video.

Please include a simple demo page for static images and another one for webcam use, plus a basic Node example that can process images from disk. If it helps, look up the current TensorFlow and browser docs online so the setup is clean and works with modern browsers. I’d like the result to be easy to run locally, easy to understand, and ready to plug into another app.