Almost All Recordings in Musicbrainz are for Merging!

I’ve seen that this problem is partially mentioned in the forum but I decided to start a new thread because in my view there is a very big problem in the site, moreover, it is a real mess.

I hardly see recording in the site that is not for merging.

Let’s put live recordings aside and discuss only the studio ones.

I think, this must be clear that, in most of the cases, there are only One Studio Recording of a work (song) by an artist.
In addition, which rarely happens, artists are possibly to get together for another studio re-recording, later in the years. This will be a Second Studio Recording of a work.
And the above recordings are released numerous times in the years.

To be a useful encyclopedia, we should see exactly this in Musicbrainz.
Now, I see tons of recordings here, created mistakenly by the users.

For example, this Boney M. song have three studio recordings:

Recording 1
Recording 2
Recording 3

You know, I have a release with that song and I cannot add it in the database because the database is blocked with these recordings.
Now, I have to merge them and wait for a long time.

I think, this is an emergency for admins to emphasis on the fast procedure of merging all these recordings.


Probably yes, when we are talking about “normal” artists. I wish it was like that but then there are those funny artists, remixers or DJs or whatever they are called, and all of a sudden you see heaps of different recordings a la Song “In and Out of Love” - MusicBrainz . Good luck with sorting any such mess out.


You can add the release anyway, the merge does not block the database.


It is a common problem, but no automatic answer. There are just not enough experts on all the different bands out there to do the merges.

I’ve been spending the past few weeks on Ian Dury. Complicated by the fact many compilations credit Ian Dury and the Blockheads for tracks. Or vice-versa.

It gets a tricky puzzle when different edits exist for singles, albums, 12"s. Some bands are a minefield of remixes. And this scares many editors away from working on tidying things up.

Or you get some bands where VA collections get edited and trimmed very different to the original release.

Ian Dury is pretty simple as many cases there is only one recording. But look at someone like Pink Floyd and not only do you have different recordings, but albums like The Wall can get sliced up in some very different ways over the decades.

The reason for the seven days is so errors can be spotted. Even when you do a ton of research, it is amazing what else comes to light. Seven days is not a “long time” when many of these tracks have sat in the database for over a decade.

All we can hope is enough editors do their bit and wade in at times and clean up their favourite artists. Even if many of the VAs can get heaped in together it will improve things. This database is here for the long term. We should not rush into just mass merging without checking as that can lead to more errors that are then harder to untangle.


And a quick read on wikipedia: Rivers of Babylon - Wikipedia

Now the problem starts to appear. The single is a different mix to the album.

Along with “Ma Baker”, “Rivers of Babylon” helped establish what was to become a habit of Boney M. singles – namely that the original pressings featured an early version that was soon replaced by a more widely available mix.

The initial single mix of “Rivers of Babylon” is most notable for lead singer Liz Mitchell’s ad-libs (“Dark tears of Babylon, you got to sing a song, sing a song of love, yeah yeah yeah yeah yeah”) between the two verses. On subsequent single pressings, only the 'yeah’s were retained. However, the full ad-libs re-emerged in the US-only 12" version, and the original earlier fade-out point is kept in the album version.

The single mix differs from the album version by having Liz Mitchell singing all of the verse “Let the words of our mouth …” with Frank Farian; on the LP, Farian sings the first half of this as a solo part. Additionally, it edits out the instrumental passage before the last “humming” part and fades out a little later ("Oooooh, have the power… yeah yeah yeah yeah" can only be heard in full in the single mix) despite being slightly shorter overall.

So which version is used on the VAs? AcoustID is a problem here as it can’t tell these apart.


Then there are the remastered versions. Yes, It is the same song. Even the same mix of a song. But it sounds different. With different people working on it. Different Sound engineers, Different studios, different release years. Different ISRC codes.

This is the opposite issue. All different versions lumped into one recording.


Different ISRC does not make different sound, so we don’t create new recording just for that.

Remastering was chosen not to make new recording any more.


I cannot speak to genre’s outside bluegrass since that is where most of my works.

I find this issue in bluegrass with old original LP releases that are re-released on CD, some with remastering, but most without. The recordings from these same old original LP releases also appear many VA releases. Almost always they are the same recording, some with less than 5 seconds length difference, some trimmed and edited, but still the same master recording.

Labels CMH and Rural Rhythm are the most prolific at releasing the same “old catalog” recording over and over on different named VA releases. I had projects to try and complete the old Rebel Records, CMH, and Rural Rhythm catalogs. I have a spreadsheet of the CMH label that showed where most of the original recordings came from and how they appear on different VA releases. Since I cannot prove they are the same (some with and some without times) I have not merged them.

I agree it is a big problem, I am not sure how to proceed. I fell a recording is a recording regardless of its length due to editing, trimming, remastering, or added silence (gaps) added on CD releases. A remaster is not a remix, but we do have the issue of sound variations due to level changes or compression added.

Just my thoughts.


Actually, if it sounds really different, it was probably remixed too. I’ve got at least one “remastered edition” with a different mixing engineer noted.
The problem is the use of the term “remixed”. In most cases it’s only used for the creation of a really different work, not for an improved version, but in my opinion, a new mix should be a new recording as well.

… and, of course, the problem remains, how to know if it’s the same, if you don’t know all the records :wink:


This is how it’s supposed to work:

The problem, as already noted, is that it’s difficult to make an informed decision on whether two tracks from different release groups are actually the same “recording” by MusicBrainz definition.


That’s where some of the puzzle lies. In that image is “Audio 1” a single recording session? Or a multi-track tape from the session?

The way I read it in MB terms is if the recording is from a single session, and someone 20 years later remixes the levels, makes the vocals clearer, reduces the volume of the drums. Then this is still “Same Recording” in MB’s rules even though we can hear a difference.

If that engineer splices in a new bit of tape with an new sample of a single ding of a bell, then we have a new recording.

This is what happened in the above Boney M example when different vocals were mixed into the album releases. Something that is easy to spot if you can get to hear all the VA editions, but not something that can be spotted on lengths or AcoustIDs.


I suppose, it’s easier to understand.

So, let’s change the direction.

Some comments say - how to merge recordings without prove that they are one and the same. → This should be turned into: how to separate recordings without prove that they are different.

I think, the documentation should say: There is only one studio recording by one artist unless someone proves that there is one completely different and then it can be separated.

This way, the encyclopedia will be tidy, at least. Currently, the chaos start ruling the place.

Furthermore, don’t forget that after all, this should emphasize mostly on the artists-musicians whose names are above all in this stuff, not the engineers.
This is a Music Encyclopedia not a technical guide. People mostly want to know how many recordings executed Elton John himself not the technical staff around him. Now the site is blocked and fails in giving this info.

Answer the question: How many times Ozzy Osbourne was in the studio to record the song No More Tears?
It is only one time, right! → This is the useful information for the audience.

After all, each release can serve as clarification for some difference in the sound via comments in the release page. No need different recording to be created in the site because of some remastering or something - just write a comment in the release page.

That would create just a huge mess. It is much easier to merge recordings than to split an recording apart again after everything got lumped together.

And yes, merging recordings is often a difficult thing, and that’s the reason why many shy away from it. Your example from your original post just shows this, on the first look it might have seemed like merging these recordings was fine, but actually it isn’t. And I have seen many cases where the single version is different from the album version.

But at least merging two recordings is something that is doable after some research. If everything is already merged and lumped together into a single recording it becomes very difficult to tell apart which relationships, ISRCs and AcoustIDs should go to which recording after splitting it.

I have one artists for which I tried to merge the recordings as good as possible. My aim was to have only recordings that have a clear reason to exist. I managed to merge a fair amount, but there were still quite a few left were I was not confident enough that they could be merged (and there have been a few recordings added since then that probably should be merged as well).

In many cases I could only reliably decide to merge or not to merge because I have most of the artist’s discography available for comparison.


100% what @outsidecontext just said.

If you want that basic level of detail, go to Wikipedia to read the biographies. MusicBrainz is more about the trivial little details and differences because we can look at the audio closer. Because each track can turn up with AcoustIDs, lengths, sources, ISRCs then we have to be careful just mushing this all together. Once in a mush you have lost the details that a fan of that artist will be aware of.

Someone who is not a big fan of Boney M will be happy with any variation of the audio as they probably don’t even know about those subtle remixes. But there are the extreme geeky fans who really do want to own one of each variation of recording with those different audios. This is what MB excels at.


There is no point in worrying about the possibly duplicate recordings.
Just browse the releases and works and relax. :wink:


I’m sorry, but that’s just backwards. Due to acoustIDs, etc. It’s much, much, harder to try to prove something is NOT the same recording than it is. I think every recording should be new, until you can prove it’s the same. For example, clean vs. explicit. Same basic recording, but edited. They are now NOT the same recording. Sure, they are the same as far as when the artist recorded them. NOTHING IS BLOCKED!!! Why do you keep saying that?


Exactly. His Boney M links actually prove the argument against this. Recently, I’ve seen many singles that have the same exact time duration as the album version, turn out to be different recordings. It’s way more common than many editors understand.


Another question: How many times did Thelonious Monk record his famous composition 'Round Midnight? At least five times in the studio, and a dozen or more commercially released live recordings.

It would be a huge mistake to assume these are all the same recording unless proven otherwise.


While I wholeheartedly agree with @mrblond that the Recordings database for artists with many releases usually tend to be complete messes, I disagree that the approach of merging recordings until they’re shown to be different. Instead, I think the issue is the current definition of Recording is too restrictive and doesn’t follow the conventional meaning.

This is where I agree (for the most part) again with @mrblond… Simply: a “recording” is the result of one or more individuals getting together to put audio content onto a medium. A “recording” is (obviously) a track-level piece of information and should be treated as such. What happens afterwards (within reason) - for example, when a “recording” is edited for radio or re-mastered for a box set release - in my opinion, should really have little bearing when it comes to the “Recording” section of the database. I also agree with @dashv’s closing paragraph. To further elaborate…

…this is a non-issue that MusicBrainz is (nobly) rectifying, but convoluting the reality (of multiple versions of a “recording”) in the process. The possible variations in a recording are already distinctive at the Release (album/single/EP/compilation)-level. For databases purposes in the example provided, the “recording” is the same and there really don’t need to be three separate pages for each version. An exception would something like “Say My Name” and “Say My Name (Timbaland Remix)”, where the "recording of: " feature handles this case quite well. As of now, the most useful aspect of linking recordings across releases (grabbing the earliest date of a track’s release) cannot always be accomplished with the way the guidelines currently define a “recording.”

TLDR; the definition of “recording” being a little less restrictive than what it currently is would help to tidy up the database, as well as make one’s use of that particular piece of data much more effective.