@lxo "playing the video is supposed to be useful to the user, whereas the proof of work [...]"
I don't really see the qualitative difference here. When I play the video or audio there are parts of the decoding that are redundant for me, for example the video may be too high resolution or contain a padding intro/outro I'm not interested in or audio frequencies I can't here.
"there has to be a better way"
Maybe. I don't run an Invidious instance so I don't know. On wikis, QueryCaptcha works.