Fixing incorrect track matching inside a correctly matched album?

[KeebCat]

When I try to add tags, Beets often completely messes up track matching. An example is the following:

  Match (32.0%):
  Ginger Root - SHINBANGUMI
  ≠ tracks, album, artist
  MusicBrainz, Digital Media, 2024, XW, Ghostly International, GI443, None
  https://musicbrainz.org/release/f4583804-b928-4b7d-b573-3cf17a2dabc2
  ≠ Artist:  -> Ginger Root
     ≠ (#0)  Think Cool.opus (0:49) -> (#1) Welcome (0:40)
     ≠ (#0)  No Problems.opus (3:44) -> (#2) No Problems (3:45)
     ≠ (#0)  Better Than Monday.opus (2:50) -> (#3) Better Than Monday (2:50)
     ≠ (#0)  There Was A Time.opus (2:49) -> (#4) There Was a Time (2:49)
     ≠ (#0)  Take Me Back (Owakare No Jikan).opus (3:03) -> (#5) All Night (2:57)
     ≠ (#0)  Welcome.opus (0:40) -> (#6) CM (0:48)
     ≠ (#0)  Only You.opus (3:29) -> (#7) Only You (3:29)
     ≠ (#0)  Kaze.opus (2:15) -> (#8) Kaze (2:15)
     ≠ (#0)  Giddy Up.opus (3:26) -> (#9) Giddy Up (3:26)
     ≠ (#0)  CM.opus (0:48) -> (#10) Think Cool (0:49)
     ≠ (#0)  Show 10.opus (5:12) -> (#11) Show 10 (5:12)
     ≠ (#0)  All Night.opus (2:57) -> (#12) Take Me Back (Owakare No Jikan) (3:03)

Interestingly, if I import them "as Tracks," there's no issue, and they match correctly. Is there a way to manually reorder the tracks, or is there anything I can do, automatically or manually, to improve the matching? Importing them as tracks saves them as individual singles, which is not what I want.

I have already looked at other discussions here, but their issues seemed to be not having the chroma plugin installed or not having the correct track lengths, neither of which is the case here.

On a side note, I'm not sure why, but this wasn't a problem when I first set up my library six months ago. Back then, it only occasionally failed when the names weren't an exact match, and I could easily fix it by renaming the problematic songs to what Beets expected. However, when I tried adding some new albums over the last few days, only two out of six were matched properly.

[arsaboo]

Is .opus part of the tag? I will add that I've seen this behavior (although infrequently).

[KeebCat]

I don't know if I am misunderstanding your question, but it's the filename extension. It tells programs what type of file it is. I'm guessing you know that, though, so could you clarify the question, please?

Nonetheless, while I assumed Beets ignores those, I just removed them and tried reimporting the album. The exact same mistakes occur, even without the extensions.

[arsaboo]

Here's another example of incorrectly matched tracks:

  Match (99.2%):
  Various Artists - Ek Deewane Ki Deewaniyat
  ≠ tracks
  Spotify, None, 2025, None, PlayDMF, None, None
  https://open.spotify.com/album/31F4ueQAgDjPMQB2crda8D
  * Artist:
  * Album: Ek Deewane Ki Deewaniyat
  * Album: Ek Deewane Ki Deewaniyat
     * (#1) Deewaniyat (4:17)
     * (#2) Mera Hua (3:43)
     * (#3) Bol Kaffara Kya Hoga (3:27)
     * (#4) Mera Hua (Arijit Singh) (3:56)
     ≠ (#10) Hum Bas Tere Hain (4:29) -> (#5) Hum Bas Tere Hain (4:28)
     * (#6) Khoobsurat (4:17)
     * (#7) Dil Mere Naam (3:29)
     * (#8) Mera Hua (Female) (3:30)
     * (#9) Deewaniyat (Unplugged) (1:55)
     ≠ (#5) Hum Bas Tere Hain (4:28) -> (#10) Hum Bas Tere Hain (4:29)
➜ [A]pply, More candidates, Skip, Use as-is, as Tracks, Group albums,
Enter search, enter Id, aBort, eDit, edit Candidates, Print tracks,
Open files with Picard? a

I suspect the matching can be improved if we incorporate duration and artists into the matching process.

cc: @snejus any thoughts?

[KeebCat]

I suspect the matching can be improved if we incorporate duration and artists into the matching process.

I believed that to already be the case, is it not?
In a discussion I found while searching for the solution, weights for different types of data were shown, by Snejus, actually. The song duration was extremely high, which is why I'm especially confused by the mistakes in my example, as the duration is way off.

You know, as I was checking the comment to link it, I just noticed that they said that's their own config, so probably not the defaults. I'll go and add that to my own config file then and see if it helps.

Update: It doesn't even change which track mistakes happen, despite such an enormous track_length weight, so sadly it doesn't fix anything, although it does seem to have some effect as the album match went from 32% to 59%.

[arsaboo]

I played with track_artist and track_index and that seems to have helped. I will keep exploring

[KeebCat]

Good to know, although, sadly, on my end the files are completely free of metadata relevant to tagging besides duration and name. I guess if all else fails, I can manually add the track numbers to everything new to my collection, but that does sound like quite a bother.