Out of curiosity—why no separate recordings for remasters?

For remix, it’s already the case, hopefully. :wink:

The problem is not that much because small group of people.
The problem is more that distinguishing several sound-alike remasters, this knowledge by few people willing to do it, only concerns very very few recordings, like Ivan with a couple of Pink Floyd albums, me for some small bands with only few re-editions, etc.

But in MB, there are 28,385,131 recordings. :woozy_face:

Never. :shushing_face:

1 Like

What @aerozol said above is the main reason why we don’t have:

  1. A master level under recording
  2. An arrangement level under work

They could be useful, but it’s very, very hard as far as we can figure out to make them work in a way where the user interface doesn’t become even more confusing for the majority of people who, let’s be honest, don’t care about any of this and just care about having a tracklist with some track titles. Even works and recordings are still fairly wild with a lot of works missing and a lot of recordings unmerged.


One of the things that I have liked and appreciated about MusicBrainz is that it allows the nitty gritty details to be documented. Those details are important for some of us even when they make user interface design difficult, but MusicBrainz lets different users enter different levels of detail and complexity as we each desire. For some of us, this level of detail is what sets MusicBrainz apart from the other databases because we value accuracy and documenting the truth.

For example, I can create a release which has a list of tracks and their names or I can make each track ‘a recording of a work’ (the more correct version). Then there’s the ability to add lots of other relationships such as who played which instrument etc. Users can add as little or as much detail as they want. Remastered is just another one of those details. Why is it being discounted and dismissed?
It should be possible to have a UI that is ‘incremental’.

Another objection is “It’s too hard to distinguish what is remastered”. There are three different instances of remastering:

  1. The remastering is clearly marked and documented on the packaging of existing mediums. Remastering is a selling point and intended to stimulate sales, often to existing owners. Examples are:
  1. The remastering is not documented, perhaps accidental and is only noticed by listeners. This is the instance where it is not obvious if recordings have been remastered.

  2. Remastering explicitly to take advantage of a new medium and to drive sales. This is also already happening for on-line streaming and downloads see
    qobuz.com, hdtracks. For example Joni Mitchell - Blue, 24bit remaster available at hdtracks

So, we have two kinds of remastering where the remastering is explicitly documented. It is not a problem to determine if those tracks have been remastered! MusicBrainz will need to provide ways to document the remastering if it wants to maintain accuracy in the future. It’s my belief that the remastering will increase driven by the opportunity to get more money from a record company’s existing catalog. This trend will repeat what happened with the movement from vinyl to CDs. Being able to document those remasters clearly will also help the merge problem because it will be possible to keep those known remasters separate.

Another objection has been “it’s too much work to go back and sort things out”.
Providing the ability to document known remasters in the future, doesn’t mean that the existing database content must be analysed and updated. It can be left as-is.

Another objection could be that it will complicate the database design. I don’t know the details of the MB database, but complicated relationships like remaster of recording of work are the motivations behind new database technologies such as GraphML.

My conclusion is that the requirement to document remasters will increase. I understand that implementing it won’t be easy.

1 Like

It would be easy to implement as all these are just MB recordings with a recording-recording remaster relationship (that is now deprecated but that’s still exists).


At the beginning of NGS, MB did separate recording for each master.
But it was decided to stop and I agree.

It was unneeded complexity, as in most cases, it’s another unverifiable info so recordings will proliferate, each new releaser will get its set of overkill recordings, without daring to merge them.

And then you I can no longer see if this or that release includes recordings that are or are not yet in your my collection:


Just taking Joni Mitchell as an example. Look at Track One on Blue:

How would you sort through that short list of recordings? We don’t have all the artwork to read. There is no way to know which remaster is used on the collections. Even harder when a track appears on a VA. Some popular artists the recordings page can run to four or five pages of Releases with the track on.

Relying on the text on the CD is also bogus. See above with my copies of Wish You Were Here from 1984 and 1994. That “remaster” isn’t. Was it an error in production, or just a sales repackaging con. I don’t know - but those CDs are identical.

AcoustID is no help to you with remasters either.

This is the kind of data that lives better on fan websites. Or on a Release annotation \ disambig. 90% of users just would not understand it enough to put the correct details in. As it is there is enough mess from people picking wrong recordings (album vs 7" vs 12" vs bad cover artist) when adding a release to the database. If you added in all the remasters it would be impossible to keep them pure.

The other problem comes from the more the more you split up recordings, the less likely you are to find Works, Performers, Locations added. It is hard enough to persuade people to fill out these details on one set of recordings, but if it was splintered more by each reissue then it would lead to not only missing data, but different versions and errors in that data.


It is complexity that exist - and is needed to accurately document.

Not true. In the case of remasters I have shown at least three examples where it is as verifiable as anything else in MB.

Why is this an issue? they are documented to be known to be different.

I don’t need to. I just want to accurately document going forward!

I’m not worried about the design of the database itself :slight_smile: We can easily add complexity to the schema.

I’m worried about how to make it so that the user can make use of that complexity in a way that is understandable. For example, if we were to add a master level under recording, where would that fit in the (already super long and confusing for many users) release editing process? :slight_smile: Would we add yet another tab for master? Would we have optional sections in the Recording tab (which again is already very confusing to many users) to add masters?

I’m not saying it’s impossible - in fact, if we were something like AllMusic with paid staff that works on adding and correcting data and no need to make the tools understandable to anyone else, I think we would have implemented it. I’m just saying it’s hard to find a balance that doesn’t make MusicBrainz so confusing for new casual users that they just run away (in fact, sadly that’s already not that uncommon even now).


Of course there are cases where it is well documented, but the majority aren’t. Labels love to put stickers with “digitally remastered” on re-releases, but seldom do they provide extensive information. And on best-of and other compilations it might be mentioned that the tracks are remastered, but good luck finding out whether it is identical to a previous remaster or not.

With the given information it is often difficult enough to even figure out if two recordings in the database are the same or not. If one doesn’t have the artist’s full discography and the ability to compare the audio.

1 Like

I hope you can see that it is in no way being discounted or being dismissed out of hand! The question is simply - will this change make our database more or less accurate and complete.

If you can think of a solution where recordings aren’t merged incorrectly or not being merged anymore (with all the resulting issues) while taking mastering into account, that would be amazing. Just changing the guideline to “make separate recordings for different masters” sadly doesn’t work.

1 Like

With 3 tracks that sound different from each other over the span of 27 years, it’s highly likely that they are new mixes, which do get stored as separate MusicBrainz Recordings. For example, https://beta.musicbrainz.org/recording/db760486-ba4e-468f-8151-7bdd1ef83eb4 is a new mix (I would not call it a “remix” as it is currently listed, but a “new mix” or “re-mix”) that is known to differ from https://beta.musicbrainz.org/recording/1aea1980-37a8-41b2-bb78-526c94bb428d in both how it sounds and the credited people.

There is a difference in concept between a “remix” and a “re-mix”.


I usually put that information in the annotation on a recording when I know it. Like I’ll just add to the annotation “Remaster found on the following releases: Release 1, Release 2, etc.”


As mentioned before, I am focused on the cases where they are well documented. I’m happy to just worry about those. This isn’t an ‘either or’ choice. Just let me mark the recordings that are listed as remasters and not worry about the others. It’s a start.

A valid concern. But for those that want to document recordings and releases, not having the ability to mark recordings as ‘remaster’ is also confusing.

I believe it will make it more accurate and more complete.

How about:
Don’t merge recordings marked as remaster with those that aren’t

I understand the wish to merge recordings to eliminate duplicates, but a duplicate doesn’t cost of lot of storage space. Why do we need a high level of merging? I don’t understand why merging is such a priority and more so than accurately documenting what is on the credits - that it is a remaster.

I’m not sure people are fully reading my argument which is focused on just those that are marked as ‘remaster’ and my being prepared to accept that. Some will claim that that they can hear that it is not or that they can see if under analysis. I recognise that, but I’m just more prepared to accept the credits / liner notes as written as a starting point and document that.

What interests me is what versions do we have, what original recording, what remixes, what acoustic version, edited version, what live versions…

I want to know, do I have this one or not?
Have I ever listened to it (ratings)?

Usually there are not 36 studio versions by one artist of 1 work, usually just 1.
If 2 or 3, then you will have version name or comments to distinguish them.

But if you split remasters, you will see 108 studio versions of each song.

And remasters are not really interesting, artistically, unlike versions.


Remasters just shouldn’t ever be separate from the originals, IMO. The recording has relationships of when it was first released. If I have a 2022 remastered recording of a 1950 recording, it would say, that it’s a 2022 recording. I want the original date of the recording, not the remastered date. There were well thought out reasons why we got rid of remastered recordings. We don’t even separate remasters that are for things like “fake stereo” from mono, etc. So, a remaster that is just cleaning up hiss definitely shouldn’t be a new recording.

1 Like

Is this documented and summarised anywhere?

Understood, I note that Picard is able to replace the release date with the original date - $set(date,%originaldate%), it would be simple to do the same for explicit remasters.

It appears that you’re not understanding that I’m distinguishing between explicit remastering, and the other suspected or derived ones. For example, In the case of the Joni Mitchell example, I gave there is

The original masters for the vinyl LP released in the USA
Two explicit remasters:

Any recording is not marked as an explicit remaster is assumed to be from the original and can be merged.

So, in the scheme, I propose there would not be 108 remasters. Just two and if you wanted to combine them you could use something like Picard does with release dates ($set(date,%originaldate%)).

I care about this because I want to know which remaster I’m listening to and what I have. A lot of the early music moved from vinyl to CD really is badly mastered.

I see. Maybe if we don’t merge masters, we would have to create an unknown master recording then, rather than using the original master.

Most of the VA compilation tracks should go to unknown master.

Some might believe that remastering is a record company ploy to get customers to buy a new version.

The main problem which this presents is that the term is so widely used by labels that it is at best nebulous, at worst entirely meaningless, or even misleading. Just because a given release states that it is “remastered” doesn’t mean that it is actually true, for instance, a CD which is the first issue on CD cannot be a “remaster” if no CD master had been issued before.


Though it’s half a year since the last post, I have this example:

If you read the history of this release on Tell Your Friends (Remixed & Remastered) FLAC Download - Snarky Puppy Official Online Store you will realize that it got completely newly mixed and mastered from the raw recording material.
Imo this justifies (the existing) separate recordings if not an own release group.