SongKong not finding releases I know are in Discogs and Musicbrainz

[CBD]

SongKong seems like the perfect software for me in my attempt to tag some of my collection I know are not on MB by augmenting/adding Discogs data to my files.

I’ve just downloaded the trial version and I’m dissapointed to see that it can’t find two test releases I tasked it with. I reckon I must be doing something wrong, but not sure what. I even retried default mode and it recognised some songs from MB but they’re not from the actual release.

I have logs I can email. I’m keen to get your software working in trial so I can be confident to buy a full version but I can’t do that while I can’t get it to do what I want.

thanks

EDIT:

So, I just pasted the latest logs into ChatGPT and it said that SongKong is skipping the correct match because the MusicBrainz entry linked to the Discogs release (ID 6210459) only lists 15 tracks, whereas my copy and the actual Discogs release contain 26. Because of this mismatch, SongKong is filtering the release out automatically.

It also noted that SongKong won’t search Discogs independently unless there’s an existing link to MusicBrainz via Albunack. So even though the Discogs release is a perfect match in terms of title and track count, it gets ignored because the MusicBrainz version is incomplete and the AcoustID coverage on my files is too low to override it.

Is chatGPT telling the truth? If so, can I override this behaviour? I really thought SongKong would search MB and if it didn’t find it then search Discogs. Your software seems pretty comprehensive so I’m guessing I’m just missing something in the settings?

[paultaylor]

Hi, can you please run Admin:Create Support Files so I can see your reports then I can give you a more accurate answer. But briefly, yes SongKong can search and find Discogs matches without needing a MusicBrainz match so CgarGpt is incorrect.

[CBD]

thanks @paultaylor for the quick reply! I have sent the support files. Let me know if you need to know anything else about mysetup etc.

[paultaylor]

Hi, slight issue, the Support File sonly send the last 5 reports for any task, so I have 9-13. Its look like report 12 and 13 were for Bronco BullDog but its look like you cancelled the report creation midway so I cant see the details, and if you had already run it before repoert 9 i cannot see it.

Could you please rerun to completion without cancelling report creation and then rerun Create Support Files

[CBD]

Done. This is the release I expected it to find in Discogs

https://www.discogs.com/release/6210459-Various-Bronco-Bullfrog

Thanks for your help.

[paultaylor]

Hmm, actually you seem to have encountered a bug

java.util.IllegalFormatConversionException: d != java.lang.String
	at java.base/java.util.Formatter$FormatSpecifier.failConversion(Formatter.java:4515)
	at java.base/java.util.Formatter$FormatSpecifier.printInteger(Formatter.java:3066)
	at java.base/java.util.Formatter$FormatSpecifier.print(Formatter.java:3021)
	at java.base/java.util.Formatter.format(Formatter.java:2791)
	at java.base/java.util.Formatter.format(Formatter.java:2728)
	at java.base/java.lang.String.format(String.java:4386)
	at com.jthink.songkong.reports.fixsongsreport.CompletenessReportSummary.addSortFieldCheck(CompletenessReportSummary.java:175)
	at com.jthink.songkong.reports.fixsongsreport.CompletenessReportSummary.addCompletenessSummaryToTextReport(CompletenessReportSummary.java:82)
	at com.jthink.songkong.reports.fixsongsreport.FixSongsSummaryReportSection.getMetadataCompleteness(FixSongsSummaryReportSection.java:328)
	at com.jthink.songkong.reports.fixsongsreport.FixSongsSummaryReportSection.outputSummary(FixSongsSummaryReportSection.java:382)
	at com.jthink.songkong.reports.fixsongsreport.FixSongsReport.outputChart(FixSongsReport.java:185)
	at com.jthink.songkong.reports.AbstractReportCreator.outputSummary(AbstractReportCreator.java:79)
	at com.jthink.songkong.reports.fixsongsreport.FixSongsReportCreator.createBodyOfReport(FixSongsReportCreator.java:83)
	at com.jthink.songkong.reports.fixsongsreport.FixSongsReportCreator.createBodyOfReport(FixSongsReportCreator.java:30)
	at com.jthink.songkong.analyse.toplevelanalyzer.controller.AbstractModifySongsController.createReportBody(AbstractModifySongsController.java:578)
	at com.jthink.songkong.analyse.toplevelanalyzer.controller.AbstractModifySongsController.start(AbstractModifySongsController.java:479)
	at com.jthink.songkong.ui.swingstartcontroller.AbstractPrimaryTaskSwingWorker.doInBackground(AbstractPrimaryTaskSwingWorker.java:47)
	at com.jthink.songkong.ui.swingstartcontroller.AbstractPrimaryTaskSwingWorker.doInBackground(AbstractPrimaryTaskSwingWorker.java:16)
	at java.desktop/javax.swing.SwingWorker$1.call(SwingWorker.java:305)
	at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:317)
	at java.desktop/javax.swing.SwingWorker.run(SwingWorker.java:342)
	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144)
	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642)
	at java.base/java.lang.Thread.run(Thread.java:1583)

I will come back to this tomorrow

[paultaylor]

So that issue is preventing report displaying summary information, raised bug https://jthink.atlassian.net/browse/SONGKONG-2753 and fixed for next release due sometime next week.

But I think the underlying issue is that it doesnt find the match because of a combination of the following metadata issues:

• Because there is no match on Musicbrainz the Acoustid are limited use to finding a Discogs match and there are some issues with the existing metadata that are preventing it finding the match
• The Album Artist should be Audience or maybe Various Artists as it is on Discogs but you have set to Audience, plus the members of Audience Howard Werth, Keith Gemmel, Trevor Willliams, Tony Connor*
• The Number of tracks for the release is 27 but you only have 26

I looked at the other test album https://www.discogs.com/release/6709590-Beastie-Boys-BASTARD-BUSTERS-Underground-Remixes-Part3-

The issue here is that your existing songs have no metadata, and even the folder name is awkward because although you have a Bastard Busters album folder there is no Beastie Boys allbum artist folder above it.

Then although the album is on MusicBrainz it is missing track lengths so we cant use AlbunackDisc ids method,

and none of the songs have Acoustid fingerprints so we cannot use that either

So in summary you have picked two very hard to match examples, but for these ones you can use Match to One album task instead, enter the id and enable Force Low Scoring Match and it should work

In general do you realize that you dont have to match one album at a time you can point Fix Songs to whole collection ?

Probably a better test is to run over whole collection and see how it does rather than focus on specific albums.

[CBD]

@paultaylor

So that issue is preventing report displaying summary information, raised bug https://jthink.atlassian.net/browse/SONGKONG-2753 and fixed for next release due sometime next week.

Yeah, I saw that blank page on the report and wondered what was up with that. I had performed some other searches and it displayed a ‘proper’ report summary for those.

@paultaylor

Because there is no match on Musicbrainz the Acoustid are limited use to finding a Discogs match…

Yes, I’m aware that there is no match on Musicbrainz but there is a match on Discogs. I specifically chose this album to test to see how your software coped with this.

@paultaylor

…and there are some issues with the existing metadata that are preventing it finding the match

This album was originally ripped from my CD version using Windows Media Player as Flac and the metatdata was populated from whatever WMP uses as its source. All of my 3000 CDs were ripped like this 2 years ago when I discovered Plexamp and decided to add my music files to my existing Plex server on my Qnap Nas.

I also discovered that Plex prefers matches with MusicBrainz and so ran each album through Musicbranz Picard to correctly tag it for Plex. But, as I soon discovered, Musicbrainz isn’t as comprehensive as Discogs and although I started adding missing titles to Musicbrainz (that Bastard Busters Beatsie Boys Bootleg EP for example) I discovered it was a long process, even when using the Discogs to MB Tampermonkey plugin so I decided to leave the unmatched albums and see how plex handled them. I then fixed these unmatched albums manually using the Plex web interface.

This past week I have converted my FLAC files to MP3 to give me more space on my NAS to add more music. This (obviously) has caused quite a few of these unmatched albums to to not display as well as they could in Plex AND I thought I’d take this opportunity to add proper metadata to these unmatched MP3 albums to help them get categorised better in plex and to future-proof them incase I have to start a new Plex server from scratch!

@paultaylor

The Album Artist should be Audience or maybe Various Artists as it is on Discogs but you have set to Audience, plus the members of Audience Howard Werth, Keith Gemmel, Trevor Willliams, Tony Connor

Yeah, this is a weird one as it’s debatable what the album artist should be TBH. The annoying thing about Discogs is that they don’t use Various Artists (as plex prefers), they use Various. I’d probably manually tag this one as The Audience.

@paultaylor

The Number of tracks for the release is 27 but you only have 26

Track 27 is a video track.

And here’s where I was dissapointed with SongKong’s results I was hoping it would have the intelligence to match the title “Bronco Bullfrog” and see that it’s the only title on Discogs with that title and around 26/27 tracks. And my tags for Bronco Bullfrog currently have correct track names.

So I guess I’m dissapointed that SongKong couldn’t score it as a match - even if it scored it at 90% match, it’s the best match it’s gonna find on Musicbrainz or Discogs. The title are the same, it’s got more than 25 tracks and the track titles are the same. Is SongKong really going to say “well, the artist doesn’t match EXACTLY with Discogs so even th0ugh the title match, the tracks match and there’s almost the same number of tracks I consider it not a match in the slightest”.

That’s why I’m dissapointed and flagged it to you as a bug or fail or perhaps my user error in setup.

@paultaylor

I looked at the other test album… The issue here is that your existing songs have no metadata, and even the folder name is awkward because although you have a Bastard Busters album folder there is no Beastie Boys allbum artist folder above it.

But it’s got a title match, a track count match and the filenames have the track names in them. Is that not enough?

Plus it’s also on Discogs (https://www.discogs.com/release/6709590-Beastie-Boys-BASTARD-BUSTERS-Underground-Remixes-Part3-). Doesn’t SongKong move onto Discogs if it can’t find a match on Musixbrainz and just perform a new, fresh search there?

@paultaylor

So in summary you have picked two very hard to match examples

Like I said, I wanted to test it with albums I knew were in Musicbrainz or Discogs but I didn’t consider them difficult to match albums.

@paultaylor

but for these ones you can use Match to One album task instead, enter the id and enable Force Low Scoring Match and it should work

Thanks but I’m looking for an automated solution. I don’t mind if the album I’m trying to match isn’t on Musicbrainz or Discogs, but when it is and there’s more than enough clues in the files I’m using then I expect it to find those.

My workflow is : find it in musicbrainz if it’s there and add all the metedata from MB, otherwise find it in Discogs and add all the metadata from Discogs so at least my MP3s will have coverart and more comprehensive data than what Windows Media Player provided. But I want these two searches to be ONE action.

I really appreciate the time you took to look into this. Your software seems very powerful, but I guess that’s where my dissapointment comes from. I also tried using BEETS, but the Discogs plugin doesn’t work and hasn’t worked for years and no-one seems to be able to fix it.

Thanks again.

[paultaylor]

Hi, in summary since you have come this far please give just Fix Songs on complete collection a go and see how it gets on, I think you will find the vast majority of albums (including Discogs only ones) are identified.

I get your disappointment with the two you tested, and I can improve the algorithm but to explain again

Bronco BullDogs

No Musicbrainz Album to find

The Albunack Discogs Id match failed because comparing 26 track with 27, taking into account, and unfortunately nothing in Discogs to clearly identify that is a video track.

So Match is done against Discogs using metadata only, but the starting point is various combinations of album title, album artist and number of tracks and one of these matches does bring back two potential matches and because of this we don’t try the last resort of matching by track titles. But when we check the two potential matches they turn out to be the wrong album

So it could be improved to always do track title match as well and that may have worked, but needs consideration because every additional search takes time.

I was hoping it would have the intelligence to match the title “Bronco Bullfrog” and see that it’s the only title on Discogs with that title and around 26/27 tracks

So there are other albums with same name,

https://www.discogs.com/release/5127863-Bronco-Bullfrog-Bronco-Bullfrog

but not with 27 tracks, but we were expecting the exact same number rather than fuzzy 26/27 tracks you required which is usually not an issue

Bastard Busters

But it’s got a title match, a track count match and the filenames have the track names in them. Is that not enough?

So there is no trackname metadata, yes the filenames contain the trackname, and they are in an album folder, but because they are not in an album artist folder (which would be useful) that makes it difficult to parse the filepaths to extract the metadata correctly.

So actually we do do searches correctly based on trackname from filename and release name but unfortunely we combine with an artist name of beets rename project and this messes up the identification

And there are no Acoustids so we cant use those

And both the MusicBrainz and Disocgs releases dont have track length recorded so we cant use our Albunack Disc Ids method that create a checksum based on the tracklength of each track in order and compares it with one calculated for each MusicBrainz/Discogs release (Similar to how tools like DbPoweramp look up CD’s when ripping)

Doesn’t SongKong move onto Discogs if it can’t find a match on Musixbrainz and just perform a new, fresh search there?

Yes, but we have the same issue

And its unusual to have no metadata in files, incomplete folder name metadata, and have no acoustids, and no track lengths

@CBD

Like I said, I wanted to test it with albums I knew were in Musicbrainz or Discogs but I didn’t consider them difficult to match albums.

But actually they are difficult to match because we use four different methods to find matches on MusicBrainz, and three on Discogs. But the combination of your metadata and what is on MusicBrainz/Discogs conspire against any of these working on this occasion.

but when it is and there’s more than enough clues in the files I’m using then I expect it to find those.

The trouble is the clues are contradictory, such as folder name above album folder is beets rename project, this usually indicates artist name but not in this case.

You have 26 contiguous tracks, indicates 26 track album but actually it is 27 track album according to Discogs, and so on.

[CBD]

OK. Thanks for the comprehensive reply.

I’m gonna take this opportunity to tidy my folder structure too. A lot of my music is artist > album > track.mp3 but some are just genre > album_title

So, I’ll try the SongKong trial on a big folder and see how I get on.

Thanks again.

[CBD]

@paultaylor

Hi, in summary since you have come this far please give just Fix Songs on complete collection a go and see how it gets on, I think you will find the vast majority of albums (including Discogs only ones) are identified.

Hi Paul. Thanks for the encouragement. I tried the Lite version on a folder of hip-hop that had spotty metedata/filenames etc and, you were right, it did a good job of matching them. Also being able to see the unmatched weird/bootleg/custom albums/tracks in the report was really useful for going into files and manually retagging using metadatics.

I also tried the folder restructuring using the Lite version and from what I recall I was pleased with that too (apologies, brain fuzzy due to trying to get this project underway for the last week trialing many different methods)

You’ll be pleased to hear that I went ahead and purchased a Standard Licence. The software is going to make my retagging project much easier than any other method I’ve tried. Plus, I’m really impressed with your customer service in the speed and detail that you answered my question here.

I am noticing that the speed that it’s processing files is using the Standard Licence is a great deal slower than the Lite version. Is this because the Standard Licence version is actually writing the files (cos I unchecked Preview) and not just previewing them?

EDIT: I’m installing on Qnap via Container Station. I was previously using it from mac to QNAP via network. Cool that it’s available on Qnap!

all the best.

[paultaylor]

@CBD

You’ll be pleased to hear that I went ahead and purchased a Standard Licence. The software is going to make my retagging project much easier than any other method I’ve tried. Plus, I’m really impressed with your customer service in the speed and detail that you answered my question here.

Great, thankyou for your trust and purchase.

@CBD

I am noticing that the speed that it’s processing files is using the Standard Licence is a great deal slower than the Lite version. Is this because the Standard Licence version is actually writing the files (cos I unchecked Preview) and not just previewing them?

That will make it slower, how much slower depends on the audio format and if it has to make more space for the metadata, if it does and metadata is stored near the start of the file (which is usually is) then the whole file has to be rewritten, and if it is a lossless format then there is more data to be rewritten. But if writing over the network that make a big difference

So im not clear if you noticed slower running on Mac over network, or locally on Qnap but Im assuming mac in which case the issue is writing over the network. Using it on Qnap will resolve that issue, although I expect the Qnap is underpowered compared to your Mac so song identification will be slower.

To sumarize if your files are stored on the Qnap Running in Preview Mode on Qnap probably slower than mac, but running with preview off probably quicker.

[CBD]

thanks. so I just set it up on my Qnap. First time I’ve used Container Station (I’m not really familiar with docker etc). Took a little while. But finally got access to the web interface and excitedly started a little Fix Songs task on a test folder and … “This task can only be run in Preview Mode with a Standard license, upgrade to Premium license here so Preview can be disabled”

I didn’t know that. So I need to give another £15 to use it on my Nas? Mate, this sucks. I’ve had so many roadblocks on this journey. Thought I’d finally gotten somewhere with this only to discover I need to stump up even more cash.

I appreciate that nothing’s free but I guess I’m just dissapointed. So looks like I’ll be copying all my files from the Nas to my mac so I can get started on this project.

[paultaylor]

Hi, I have just sent you a free upgrade to Premium.

It is clearly mentioned on the https://www.jthink.net/songkong/en/buy.jsp

Same license valid for all devices, but at least *Premium* license needed for devices that only use Web/Remote UI such as Synology, Qnap and Unraid

but was planning to change this anyway.

The history of this is we didnt originally have Web UI, so when added a Web UI it was an obvious candidate to be a Premium feature and at the time we only have a few key tasks Fix Songs, Delete Duplicates and Rename Files so we needed some more Standard/Premium differentiators. Now we have many more tasks so it is easier to differentiate Standard from Premium by having core tasks in Standard and less used tasks in Premium. So had already planned to make Remote UI available in Standard because it is an unfortunate consequence of using SongKong on a Nas that unlike the Mac or PC versions you have to use Web UI it is not optional.

[CBD]

@paultaylor

Hi, I have just sent you a free upgrade to Premium .

Wow. Many thanks. Wasn’t expecting that. Thank you so much.

Thanks for explaining the history of the thinking behind the pricing structure.

[paultaylor]

Raised https://jthink.atlassian.net/browse/SONGKONG-2755 to try and improve algorithm a little