I like the idea of this, but

[ANaPhOrAX]

but as it is, it's too flawed to be viable -imo-. which I'm guessing is probably not even your fault, since 2 other tools I tried had the same issues;

it just adds way too many incorrect songs. sometimes the wrong mix of a song, sometimes the wrong language version of a song, sometimes an entirely different song.

I don't know why it can't just look for an exact match, not just by name but also by length. why are songs being replaced by some random song with a very different duration? by a different artist, from a different album.. and with not even the exact same song name, just a similar one. honestly, why?

another example: a song by Ukrainian act 'NikitA' - for some reason it decided to add the English version of the song...? why?? that's not the one on my playlist. I don't need it to think for me. that because my language is English that I probably want the English version of the song or something like that. none of that cutesy bs.

lastly, it really struggles with Japanese song titles. a bunch of haikyuu/JJK songs just got replaced by random songs from the same composer. bizarre.

I don't know if it's something you have any control over, of it's some limitation to the API or w/e - how 'strict' the script is in finding matches for songs. but as it is, it's not really workable. I can't manually go over 5k songs to check if they're even the right one.

[sigma67]

Well, try search for the song on YTM by it's Spotify title, you'll usually find the reason.

The match is mostly based on title, artist name and to a large percentage on the duration of the song.

If you're looking for higher match quality, your only option is a paid service which likely has some sort of mapping database.

[ANaPhOrAX]

so but, is it an algorithm you personally wrote? for the matching of the songs, I mean

[sigma67]

Yes. I originally wrote it for personal use to transfer some weekly playlists from Spotify to YTM.

I did a lot of refinement until I arrived at something that had good quality results. The song duration match gave the greatest improvement in quality.

The algorithm is here: https://github.com/sigma67/spotify_to_ytmusic/blob/main/spotify_to_ytmusic/utils/match.py

For sure this can be made configurable if someone's willing to invest the effort.

I just never intended for this to become as big as it has. It was just meant for personal use and when I started getting lots of stars I put it on pypi.