Help requested improving the MusicBrainz/Acoustid database

As you may know one of Roon major sources is the opensource MusicBrainz database, therefore the better the MusicBrainz database is the better the results you will get with Roon. But Roon themselves are not going to be able to improve that database on their own, but as a community we can do alot ourselves with not so much effort.

Warning I am going to ask for some help on a specific task here, that will benefit all users of MusicBrainz and tools such that use MusicBrainz.

MusicBrainz tagging tools make extensive use of Acoustid to audio fingerprint songs, if the fingerprint is already in the Acoustid database we can then get metadata for the song. Hopefully this will include a link to a MusicBrainz recording id and then we can make use of the data provided by MusicBrainz, otherwise Acoustid only provides us with basic artist, title, and album names.

Acoustid/MusicBrainz pairings are submitted to Acoustid by users via various programs, each time a particular pairing is submitted its Sources count is increased. This includes SongKong but mostly by the dedicated the fingerprinter tool tool provided by Acoustid.

Usually an Acoustid links to one MusicBrainz recording, but sometimes it can link to multiple MusicBrainz recordings. These may be different MusicBrainz Recordings that are actually the same song or they may be for a completely different song.

It is valid but not that common for one Acoustid to match completely different songs but sometimes applications submit incorrect pairings. This is fairly obvious when you look at an Acoustid page because the valid links will have similar titles and high source count, whereas the invalid ones will match to completely different title and have a source count of one, and the track length maybe completely different to the fingerprint length.

I wish Acoustid filtered these out but it does not so they have to be manually disabled, to aid in this I have created a number of reports identifying the most likely bad pairs.

The first report shows pairs whereby the bad match only has one source and the track it links to is incorrect length for the fingerprint it is matched to, so in almost all cases this match can be disabled.

I have disabled many pairs already but would love some help from others to clear this list, I regenerate the report every day to remove pairs that have been disabled

Procedure is as follows

  1. Create free account on Acoustid - easiest to use MusicBrainz account if have one.
  2. Choose an artist in the report
  3. Click on link in Acoustid column
  4. Find the bad link and select Disable if it looks bad
  5. Comment can be left blank and select Submit

An example

First item in the list has a fingerprint of 02:49, and has a good match (within 5 seconds of fingerprint is acceptable) to the song if you want me to by $wingin’ Utter$, and that pairing has been submitted 29 times. In contrast it also has a match to a completely different song hopeless vows by same artist (virtually impossible for two songs by same artist to end up with same acoustid), but it has only be submitted once, and the recording linked to is only of length 01:48, over a minute out

Click on the link and it is clearly wrong

so just select Disable and then Submit to disable it, (you can enter a comment, but not really necessary since it is not displayed anywhere anyway)

and now the row is shown as disabled

If you make a mistake you can click on the Enable button to renable it.

The procedure is quick with no typing required only a few button clicks and some basic data analysis skills, if I could get a few of you just to spend half an hour on this I think we could clear it in no time !

1 Like

Started on a couple of tracks.

1 Like

It’s a real pity that Roon is not making any use of AcoustID’s for better track identification. According to the Valence site: “Valence uses an approach which takes into account file names, directory structure, tag data, and file length to deliver high-confidence album and track identifications to Roon.”… so no AcoustID’s…

This leads to many albums remaining unidentified in my library, even if those are properly tagged in Picard and contain both MusicBrainz ID’s and AcoustID’s file tags… but Roon just ignores those and tries to perform its own identification, which fails and then manual identification has to be performed. Taking into account AcoustID’s would surely help with better track/album identification.

Hi, you are correct that Roon doesnt use acoustIds, but it does take note of MusicBrainz Ids (although I don’t know exactly how) so it should pick up tracks already identified by Picard. The approach of prematching with SongKong works well for getting more matches in Roon, but I dont know why the same approach would not work when using Picard.

One of the reasons for which Roon doesn’t identify albums even if MB ID’s are already present in the files is missing (or not closely matching) track duration on MusicBrainz. Picard doesn’t mind if duration is unknown as it will still match the track by AcoustID, however if Roon can’t match duration closely enough then manual identification and confirmation will be required…

In the example below, album Americanos by Contraluz is shown as unidentified in Roon, even if MB IDs are present in TAGs, so I had to identify it manually because all tracks were listed as not matched (unknown duration)… After saving it manually, all is fine, but if Roon would have used AcoustID’s for identification, then manual action should not be needed in such cases…

Acoustids matched to MusicBrainz songs without duration is a little risky, because 2-3% of Acoustids/MusicBrainz pairings are incorrect. In fact whilst its reasonable for Picard to match to songs that do not have a duration, if duration does exist and it doesnt match your tracks then it should discount the match

See this report for 185,0000 tracks that differ from the MusicBrainz match by over 30%, these are most probably wrong, currently working out a strategy for dealing with them.

So I can reasons for Roons caution here.

I suppose the current ideal here is for you to enter the track lengths in MusicBrainz, and then after a bit of time Roon will be able to automatch it, but that doesnt give you an immediate fix