Embed web interfaces into a video using object tracking.
This is not currently packaged for use, I was just racing to a proof-of-concept.
Once per video:
Object tracking in python -> output in some format -> corners-to-corners.js -> creates a trackArr variable, tuple of [portion of video completed, x1, y1, ..., x4, y4]
Each time your website runs:
use trackArr in setTrackPos to set CSS3 matrix3d transform onto the element you want tracked onto the video. Do this more than 60 times a second. Voila, you got something.