SongKong Jaikoz

SongKong and Jaikoz Music Tagger Community Forum

Contemplating re-tagging my music database (165k Files) with Songkong

#62

Fix Songs has now run through. First thing I did was to look at the non matched albums. Most are obviously niche content that is unlikely to be on Musicbrainz or Discogs. But some matches fail, even though the exact album with the same tracks exists:

Monserrat Caballe - Norma should match to
Release “Norma” by Vincenzo Bellini; Montserrat Caballé, Fiorenza Cossotto, Plácido Domingo, Ruggero Raimondi, Carlo Felice Cillario, London Philharmonic Orchestra - MusicBrainz

I tried to album match to specific Musicbrainz release, but it wouldn’t match either. I’ve had this issue with several albums. Doing an album match to the discogs release usually works, but not with Norma.
Is there any way for me to see directly why the Caballe Norma is not matching to the Musicbrainz album?

I have uploaded the support files

#63

When you tried the Match to one Album task did you enable the Force Allow Low Scoring matches option?

#64

Yes I did, but no success.
I then used Jaikoz and that worked with Norma.
With Belohvlavek / Martinu Symphonies and using Jaikoz I could not get a match using the Musicbrainz ID, but I could get a match using the barcode listed in Musicbrainz. Strange indeed.
I have uploaded the Jaikoz support files.

#65

Have not looked at your logs yet, however we released SongKong 10.9 yesterday and this includes an improvement to the classical artist checking that I noticed as an issue preventing some of your matches a while ago https://jthink.atlassian.net/browse/SONGKONG-2614

So please try out SongKong 10.9 by rerunning Fix Songs and see if that fixes some more albums. You have already resolved the Norma album so can’t check that but let’s see if it does fix some other missing ones. Rerun will be much faster than first time round since will only try to find a match for albums not already matched.

I will look at your logs later today.

#66

okay I see you did use 10.9 in FixSongsReport0032 and of 56,592 songs it had matched 43,297 songs to album and 40,483 to a MusicBrainz album compared to 41,389 songs and 37,518 songs to a MusicBrainz album when ran using 10.8 in FixSongsReport0029 so the fix seems to have had some effect.

Here are some examples of files that were updated in report 32

image
image1920×1040 68.5 KB

And when I checked a few the change was that they had been matched to a MusicBrainz release when they were not before

e.g

image
image1920×1040 76.3 KB

So I think that fix has made a difference but I will look into it more later.

#67

I thought it had not been matched at all but I can see it matched https://musicbrainz.org/release/5d644e91-56fa-48ef-a86c-81b652cff32a in Report 27

image

but the problem is you want to match to this release instead https://musicbrainz.org/release/f11a7ab6-7d1c-49fa-a6f0-b8b33004e832 ?

If so since it has already been matched running Fix Songs again will not match it to a different release it will only update with if any changes for the match (unless on the basic tab you change For Songs Already Matched option to Rematch)

But Match to One Album does do a rematch. Is my summary of the problem correct, if so I will delve further.

#68

Hi, this is because you entered 627c9f04-b033-40db-a0d1-317bb69dc6f2 and this is a MusicBrainz Release Group Id rather than a MusicBrainz Release Id

e.g https://musicbrainz.org/release-group/627c9f04-b033-40db-a0d1-317bb69dc6f2

#69

Unfortunately your matching attempt for this album was lost in recycled logs, if you could retry Match To One Album and then immediately aferwards send me support files for this album or a different album with same issue I can look into it further.

#70

Ah, maybe you are using ReleaseGroup Id instead of Release Id like you were with Jaikoz

#71

I now recopied the unmatched original version so I could do a rematch with MB ID for Belohvlavek / Martinu.
I Album matched to ead35746-f850-4d93-9836-f731827410c5 which seems the correct ID to me, but SongKong still would not match. Sent support files.
Jaikoz matches, but delivers this message:

jaikoz
jaikoz769×307 55.4 KB

#72

I recopied the unmatched original version, so I could try a rematch for Norma.
SongKong would not match to
https://musicbrainz.org/release/5d644e91-56fa-48ef-a86c-81b652cff32a
nor to
https://musicbrainz.org/release/f11a7ab6-7d1c-49fa-a6f0-b8b33004e832
SK says it matched 30 songs, but not the album
Sent support files
Jaikoz matches 30 songs also
Sent support files
Jaikoz matches

#73

Okay so with Fix Songs if the track length of a song is more than 10 seconds different to the track length of the MusicBrainz track it will not allow match because such a difference indicates they are not the same track or at least not the same version of the track, it would not be safe to automatically allow that match when you could be runing Fix Songs against thousands of albums

But with Match To One Album it should allow if you enabled Force Allow Low Scoring Matches option

However, I have only receved one new Jaikoz support file and no new SongKong files, are you sure you sent one ?

#74

ok, I’ve done Norma again and uploaded the support files.
Interestingly SongKong matches all 30 Tracks, but not the album.
The CD set is definitely the 2015 edition.

#75

In Fix Songs Release matching failed because of discrepancies in track lengths (times shown in seconds in bold below)

----G2:Foldername:Norma (Remaster) (44.1_16):isPartOfMultiDisc:false::Checking Release Track Duration Failed:Vincenzo Bellini; Montserrat Caballé, Fiorenza Cossotto, Plácido Domingo, Ruggero Raimondi, London Philharmonic Orchestra, Carlo Felice Cillario:5d644e91-56fa-48ef-a86c-81b652cff32a:Norma:Song:Norma - Act I - Meco all’altar di Venere:165:Norma: Atto I. Meco all’altar di Venere:178
05/08/2024 21.47.58:CEST:SEVERE: ----G2:Foldername:Norma (Remaster) (44.1_16):isPartOfMultiDisc:false::Checking Release Track Duration Failed:Vincenzo Bellini; Montserrat Caballé, Fiorenza Cossotto, Plácido Domingo, Ruggero Raimondi, London Philharmonic Orchestra, Carlo Felice Cillario:5d644e91-56fa-48ef-a86c-81b652cff32a:Norma:Song:Norma - Act I - Odi? I suoi riti a compiere Norma:261:Norma: Atto I. Odi? I suoi riti a compiere Norma:249
05/08/2024 21.47.58:CEST:SEVERE:

I notice you only seem to have tried Fix Songs, does it match if you try Match to One Album and allow low scoring matches.

If your question is why have you allowed matching to song if the track lengths dont match ?

It is because matching song only is less destructive since we are only modifying a limited number of fields therefore we can allow to be less strict in our checking

#76

Looking at Acoustid for first track I can see it had two sources of length 2:45 and 2:46 but it is linked to a MusicBrainz recording with length 2:58, usually there should only be a a maximum difference of a few seconds.

image
image1340×841 55.2 KB

I dont know if the length of track in MusicBrainz is wrong or if the linking between Acoustid and MusicBrainz track is wrong.

#77

I’ve retried match to album to
f11a7ab6-7d1c-49fa-a6f0-b8b33004e832 nor to
5d644e91-56fa-48ef-a86c-81b652cff32a
but SongKong said it could not find a match.
Support files uploading now.

This matching business is complex !

BTW when I paste the MB ID to the template, the allow low probability matches field cannot be selected. I neet to change to another tab and back to make it selectable.

Screenshot 2024-08-06 103703

#78

You just need to tab out of field because we only enable option if MusicBrainz Url field has value

Do you know why track lengths don’t match, do you have the original CD?

1 Like
#79

I can’t find the CD. I think I got it from qobuz
Bellini : Norma (Remastered) ((Remastered)), Carlo Felice Cillario - Qobuz
The timings of the qobuz edition match with the timings of my files and disagree with the timings on MB. Probably some cuts were placed at different places in the digital download than on the CD.

The Belohlavek / Martinu I mentioned above is strange. The CD booklet I have in front of me lists the same timings as MB, but the rip from that CD which is on my PC has different timings for two tracks, as SongKong noted in the screenshot I posted August 5th.

Stuff like this will make automatic batch matching extremely difficult. I think your approach to be a bit restrictive when matching is prudent. Maybe the criteria could be relaxed a bit more when ‘low probability matches’ is selected?

#80

Would be worth listening to those two tracks to see if they have additional space, premature cut off compared to the CD.

Yes, because they are not an accurate match

Well yes I thought it already did this but checking the code it appears not.

Okay so I have raised https://jthink.atlassian.net/browse/SONGKONG-2623 and Im going to try and fix that today and upload a new Windows version for you to test out.

#81

Okay for Windows I have uploaded new version of SongKong 10.9 with fix for this, this is what you need to do to install it:

image

Please note on Windows you must uninstall existing version in order to replace with new version with same version number.