MusicBrainz Case Study : The reality of MusicBrainz Database Consistency

Hi everyone, An example with kissing to Be Clever, Culture Club Album

The real content of the album : Kissing to Be Clever — Wikipédia

MusicBrainz Result :

The real album have 9 elements and the Musicbrainz version 13 ...? how is it possible

Finally when I listen the song with this title (online) and the song in my local computer this is not the same song

Here :slight_smile:

The album title is not good
The song title is not good
The fingerprint and other song ids are note good
The Musicbrainz analyse is not good

finally there is nothing we can trust except the human hearing

This is typically the reason why I don't trust old databases like Discogs and Musicbrainz, problem is the fact, there is no consistency approval, everybody can change precious information tags and finally Online Open Database are totally messy.

Wow, do you really believe that?

First of all: MusicBrainz has an approval process called "voting". Everybody can contribute, but changes are reviewed.

Kissing to be clever vom Culture Club has 16(!) different releases on MusicBrainz. With 9 to 10 and 13 tracks in different formats, release countries and release years.

There is no such thing as a "real album" - there are always many different releases.
If you are looking for the one with 9 tracks, just choose the correct one (Vinyl LP? Digital Media? CD? 1982, 1983, 1992, 1996, 2003?)

1 Like

There are simply different releases of the album with the same title. The version with the 13 tracks has 4 bonus tracks.

You'd need to choose the correct listing for your release. Whining about poor quality here, doesn't fix your tracks.

And I'm not interested in any other post of yours about "old databases". MusicBrainz has lots of current data (releases are added as we type here). Same with Discogs. I know you want Apple Music support, but your posts don't make this even one tiny bit more likely to happen.

1 Like

Up to you but I don't validate this answer, yes sometime it's possible to have different package with a new edition of an album (deluxe, remastered, bonus tracks, anniversary,...) but all of this new edition are obviously specified in the album title, whatever the source, with a title extension often between brackets, it's not possible for legal reason to have the same album title with different content. and it's just an exemple between tones of inconsistency. But feel free to demonstrate to show me I am wrong.

All the album with different version with the same name and different content are unofficial and often personal.

A good picture of what I am saying :

Finaly I understood what's happen :slightly_smiling_face:

1 - Musicbrainz indicates with a green light this is "Love Is Cold" and it's wrong, this is "Love twist"
2 - We have some different version (here live version) joined to an album because they have a bit same spectrogram, but they don't have to be joined on a clearly defined album version with 9 titles.

This is just wrong.

Every release of this 133 listed entries on Discogs is official and unique.
The "proof" for this fact is the unique catalog number and the label.
Do you really think a record label like "Virgin" would assign a catalog number for a "personal" or "unofficial" release?
If you describe a release as "unofficial" in the way that it was not released be the original label, then you are right as you can see on the 4 releases from different labels in Saudia Arabia or South Korea.

A release can have some additional text like "Anniversary Edition", but this is up to the artist and the release label and is for sure not a legal requirement.

This is just wrong.
Why should an artist or label have to change or adjust the release title, just because the original Vinyl LP was reissued on CD or now as digital download? Everyone can see the difference between the mediums. The tracks itself are the same, the album title is the same, just some tracks are in different order or the booklet is thinner and features less content or the artwork appears in black & white or full-color.

We say exactly the same thing, I agree with that, but as you are saying only the numbering will change, not the content/tracks, the number of tracks stay the same, the order of tracks can be changed, a change of content necessarily implies a change of the title, a change of numbering, no.

the problem in this case study, is the fact that some different version of a same title are wrongly added to an album and finally in the Musicbrainz a lot of album version are malformed ! I will see that directly with Musicbrainz

Every release of this 133 listed entries on Discogs is official and unique.
The "proof" for this fact is the unique catalog number and the label.

Totally agree, but the content is always the same..., only the numbering change

No, the content is not the same as you can see in this randomly choosed examples:
8 tracks example:
10 tracks example:
13 tracks example:

Mmm... I checked, I have maybe to admit I am wrong, if it's true this world is totally crazy, the same album title in different country with different vendor with different content, ok, for this point I am wrong.

I would also thank you I think I almost/or totally learn something today with your assistance, it's seems a fact that the majors in the marketing strategy decide to inflate the composition of albums in countries where the artist is not very prominent

If your trust level of “all” of the online sources currently used by mp3tag is so low, you still have the one tried and true method available: manually input the data as you add new entries to your library. Then you know 100% that the information meets “your” specific criteria. :roll_eyes:

Here it's a fact :slight_smile:
1- green result and it's not the good title (love is cold) , as you can see (song with unidentified purple cover ) love twist)

You are using MusicBrainz Picard. Why do you don't ask this here:

1 Like

I especially think and Discogs are a bit accurate than Musicbrainz, Musicbrainz is a great problem in audio tags fixing,

I did not find a good artwork online source except (Apple Music : high, Amazon : Middle, Discogs: Often Low)

But yes for this moment MP3Tag even if it's manual or /half-automated with scripts, is the best tool I found to work properly (here an example) :

My actual Progress Status :

off course I would find a good Apple Music API/Parser to automate (artwork,year, genre, numbering,..) fixing, it's very hard to find a good genre classification database

I doubt that Apple Music would "guess" much better which particular set of data should be used to tag this version of an audio piece.
If I interpret that graphic correctly, you only have data in ARTIST and TITLE. The other 2 important pieces of information, ALBUM and TRACK are missing. Of couse, these bits could easily be reconstructed if one still had the original media... The original media would then feature the required unique identifiers. And with those you would find the correct data set. Currently, the output of the scripts is only as good as your input.

I found a part of the answer

the acoustic_id that prior a song recognition in the Musicbrainz is wrong

a fastest way to get the information that MusicBrainz Picard return :

another example of the Musicbrainz/AcousticID consistency :

This is the reality of Musicbrainz

ohrenkino, Hi,

As you can see on the picture above the progress status graph, ALBUM, TITLE, ARTIST, ARTISTS ALBUM tags are almost done for :

80% of my MP3 collection for 49000 songs

this graph is more a representation of the working progress the album indicator is low because you have a lot of work to fix an album (artwork, genre, year, numbering, elements,...), so the album tag is good, the todo-list of the album consistency is low. yes I would find a identifier that can reveal all the correct data set, the best identification I found is the "ISRC" code, but seems that you can also change it, so th is is not always relevant and a lot of tunes they don't have it.

many thanks

An Answer on this Community

Again the process of the Musicbrainz database consistency protection is not reliable, then this database is not reliable,

many thanks

Again, your conclusion is wrong.
The facts:
a) Everyone can calculate AcoustID audio recognition patterns called "Fingerprints"
b) You can not calculate them with Mp3tag, but with MusicBrainz Picard
c) Everyone can submit such fingerprints with additional metadata like artist, album, tracknumber to AcoustID, not to MusicBrainz
d) AcoustID fingerprints can then be connected to MusicBrainz recordings. This helps people with very bad or not existing metadata to recognize their albums or songs.

The confusing part could be:
You can try to recognize your songs with the help of audio fingerprints (MB Picard -> Scan -> AcoustID).
If someone has already submitted the same fingerprint as you and connected it to a MusicBrainz recording, you get the matching Metadata back from MusicBrainz.

NOW the "not reliable" part is that nobody can check if your submitted fingerprint and metadata is 100% error free.
Some people care about data and submit only error free fingerprints and matching medata. Other's don't care that much.

That's the same accuracy like in other open source project, like Wikipedia. You will never reach 100% perfection.

If you see such an not yet perfect submission, feel free to fix it at the MusicBrainz side like here:
and unlink the wrong connected/linked/associated recordings.

And again I have to ask you:
What has this all to do with Mp3tag?

If you don't like MusicBrainz metadata, don't use it. If you don't like Discogs metadata, don't use it.
Feel free to add the missing data from other "more reliable sources" as @MotleyG already wrote.

1 Like

If you don't like MusicBrainz metadata, don't use it. If you don't like Discogs metadata, don't use it.
Feel free to add the missing data from other "more reliable sources" as MotleyG already wrote.

I would, this is not possible with MP3tag Mac OS Version... don't ask me why ...

The colored bar indicates similarity between your existing metadata and the track's metadata. Your screenshot does not show the tags, but I assume artist and alb probably already match perfectly, track number ptobably as well, and the track titles themselves are also rather similar. That's why it still scores rather high.

But that's kind of off-topic here and better discussed on the Musicbrainz forums.

Not only different countries. A typical case is an album getting released as a regular version in Jewel Case packaging, and then there is a "limited edition" released as a Digipack with some bonus tracks. The album title is the same, just the packaging is different. Then a few years later there is a re-release, adding again some bonus tracks. Often the album title stays the same. To distinguish between those versions MusicBrainz provides a disambiguation comment field, were one can put info like "limited edition" or "2021 re-issue".