So, to sum up:

- I apparently need to download everything- I need the ability to parse video files and figure out whether they have any audio tracks. It's not that it's impossible, but I thought I won't need ffmpeg as a dependency. Parsing mp4 by hand isn't a great idea. I already tried several years ago, didn't like it but it was still better than JSON-LD. Gosh, anything is better than JSON-LD. But I digress.