It's crowd sourced. They have software that returns timestamps when you click a button and you use it when watching a movie. It's definitely error prone and the

It's crowd sourced. They have software that returns timestamps when you click a button and you use it when watching a movie. It's definitely error prone and the data should be treated as if you got it from Mechanical Turk. There's so much data in movies that is really hard to get automatically. We should really be digitizing and cataloging old films if we're able. It's hard to find films in the public domain but this is what we should be doing with them: 4k scanning and digitizing. Perform a rough automatic restoration on all frames and audio then put it all online, every frame in 4k, and accept pull requests. During this process all shots and scenes should be accounted for. They are timed with actors and locations marked. Subtitles should have crucial metadata such as the actor speaking the lines. Music and musicians should be identified.