Digital releases

I am still digesting all of the comment here, all good stuff I think. But I wanted to comment now only on the audio formats. I agree there will be some difficulties there, especially since there is a wide range of knowledge levels, and that the normal user needs to be assumed to be mostly ignorant, which is not meant to be insulting at all, just that many people just dont care, as others have pointed out.

While this is generally correct, it is not always correct. Lossless means it is a lossless file in comparison to its source. Typically, lossless means it to be comparable to a 16 bit 44.1 kHz source, like a CD. In this description, it is correct, a lossless file of 16 bit and 44.1 can basically be converted between formats without any issues. But if you have a 24 bit FLAC, and convert that to a 16 bit FLAC or any other 16 bit lossless, you cannot go back to the 24 bit. This also applies if you take a 48kHz or 96kHz to a 44.1kHz, once you have created loss, you cannot undo that. 16 bit and 24 bit are just higher levels of the norms for lossy like 256, 128 and 320 Kbps, 16 bit and higher just come in higher at 700+ Kbps.

Currently, MB does this by default. CDs have a standard of 16 bit 44.1kHz. This is actually tracked in MB in the medium format, for example CD vs HDCD, etc. Vinyl and cassette also have their standards making the quality differences all ingenant. The same could be done for digital, there are just different medium formats we could use vs just calling everything “digital media”. So if I see CD as a format, I know the quality, and if I see vinyl or cassette, I also know the quality and know that the CD falls in between them in quality.

Now this all assumes that the vendor is trustworthy, as others have pointed out. This is a common issue, and many do not even know if they are being lied to or not. Even CDs have been known to have been made and released from MP3s, which is nowhere near CD quality audio. There are many ways to tell this, but I do not know that going into that is beneficial now as just talking about quality and containers (file types) is causing enough damage.

On what to take into consideration for a release, I do not believe that if I have a CD and RIP it to MP3 files that it is a release. The CD is, my RIP is not. No different than if I copy a CD, that is not a release, just a copy of the release. The thing is that converting a PCM file (lossless types) to a compressed audio file like MP3 or M4A is similar to a mastering process. In mastering, you cannot master a master. Meaning that once mastering steps have been taken, mastering it again creates more loss, generation loss. So the source is a major factor in all of this and relied greatly on the encoder, encoder settings and encoding process. For example, if I take a 24 bit FLAC and use my nifty GUI sound converter to make an MP3, that could result in a total mess. What is happening to those top 8 bits for example? The encoder most times does all of this with default settings so the user does not need to have knowledge of anything, and that works great because most will use it and if it plays, they are a satisfied customer. But for our purposes, what does my MP3 have anything to do with a release? It is no different than me trying to add the disc ID of my mixtape, that is not a release. My MP3 will be different than your MP3, different if I change encoders, and even different than a M4A conversion I make from the same sources.

Lastly, on tagging digital files, as pointed out very well here, there are issues. This is one main reason I do not tag my files with Picard anymore. The original meta data tells me more about the digital release than MB can. If I tag in MB, my actual release date is overwritten by the date in MB for example. Although some find release date not important, it can be real important, especially on releases that are initially released, then re-released (different date) with the same title and track list NAMES, but on close look, one of the tracks is actually different but titled the same. Yes, this does happen and is a real thing. On the meta data side of this, the only way I can identify this on original metadata, using iTunes as an example, is via the release name + release date, or by the iTunes IDs. The tracks that differ will be clear from the different ISRC assigned in the meta data. So in this case, not using ar elease data just makes it more likely to mix up the releases. So me, I want that release date as is, not changed to what MB accepts as correct via tagging, but I want it to be correctly representing the release I have.

I read through all again and had a few more comments on some very good points.

This is absolutely an issue. even if we look here:

One of those buttons advertises a “MP3 16-bit VBR”. That is either wrong or improperly worded. What it could mean is that the MP3 files were derived from a 16-bit source. What it cannot mean is that they are 16-bit MP3 files. That is just not a possibility. In this case, with English being my native language, I read that as saying that they offer 16-bit MP3 files, saying the MP3 files are 16-bit. So if one reads it and interprets it in that way, they are mislead because it is wrong. In the case it does not go without saying, you also cannot have a 24-bit MP3 file.

There is also a iTunes link and a Mastered for iTunes link there. This could be misleading as well, as there is only the MiT version. But in that case, they are not the retailer anyway, iTunes is the retailer. But it is worth noting that Mastered for iTunes is different from the iTunes Plus, also a AAC v256 file. Difference is mostly that the iTunes Plus file is derived from 16-bit sources and the MiT from 24-bit sources. This is a requirement from iTunes that it is that way.

To change and side with the “leave digital as it is” people… the above sounds great, but how do I know if I have iTunes Plus files or Mastered for iTunes files? How do I know if my FLACs are real or fake? How do I know if my 320 MP3 files are really 320 quality and not sourced from a lesser quality MP3 (or just as bad a different lossy format) or upsampled? I do not think we can expect normal users to do that process much less understand it.

I think one critical factor is the reference of the digital release. Just as CDs, we need to take the retailer/vendor at their word. Just as there are CDs out there that are not true CD quality, there are files that are not their true stated quality. But if I buy a release from the same place as you, we both at least think it is of the stated quality, thus making it possible for us to identify it. But as noted above, it would be great to have data along side the references, so such things can be noted, if they are to be in the same release. Additionally, for country stores (like iTunes) it would be beneficial to store the iTunes ID vs the actual link. The reason for this is iTunes can change the URL format anytime, but the IDs are still there. So with a little code, you can have the site just create a link given the stored ID and a country. That could be try no country first, or use the GEO format, or have a selection that allows the user to select a country from the list of available countries… one could even select which countries are on the ID vs links for each.

It is also important that only real metadata be used, which includes being untouched by Picard. I did some looking around online and checked with some of the larger pirate groups. It

is interesting to see that some of them will actually advertise “untouched files”. I am sure we all know that one of the major sources of bad data is pirated audio, especially in places like where you see artists titled “”. I think it is worth taking note that even on the pirate side, original metadata is important, or at least important enough to advertise it.

This is a bit different. In the hip-hop world, they still say they are making a record, even if it only ends up on CD or digital. A set of files is still called an album. A CD, vinyl, cassette or file is just a medium to deliver the album/release/record/etc… what ever you wish to term the collection.

If you RIP your CD to files, I do not think that is a release, but a copy of the CD. If you take your files and make a CD, that is still the files and the CD is not at all at the quality level compared to the CD that you would buy. All you are really doing is changing the medium for your purchased release. If you but MP3 files, M4A files or any other compressed lossy format, you will never have the same quality as if you buy the CD version of the release.

I’m comfortable saying that absolutely digital releases are a reality of the recorded music world, that MusicBrainz should deal with them, and that right now MusicBrainz handles them inadequately. However, I think making MusicBrainz handle them well is not a question of desire, it’s a question of successfully doing difficult design and implementation work while still keeping the existing MusicBrainz running.

This subject has been talked about before:

What do I mean by digital releases? The biggest gap for MusicBrainz are released recordings that are a) not a corollary of a release on a physical medium, and b) streamed, rather than fixed as static data, and c) single recordings rather than a group of related “tracks” that are part of an “album”.

What are some concrete examples? I’ll lift my examples from 2016:

So, are the these digital-not-physical releases in scope for MusicBrainz? OP is easier on the project than I am:

I totally do not understand. The MusicBrainz front page says,

MusicBrainz aims to be:

  1. The ultimate source of music information by allowing anyone to contribute and releasing the data under open licenses.
  2. The universal lingua franca for music by providing a reliable and unambiguous form of music identification, enabling both people and machines to have meaningful conversations about music.

There is no escape clause there for, “we only care about music information that fits in nicely with the physical albums-and-CDs conventions of the mid-late 20th century”. To achieve its mission, MusicBrainz must find a way to handle digital releases.

Furthermore, commercial music distribution is moving away from physical media, and towards non-physical, streamed, and incremental releases. If MusicBrainz is to remain relevant, I believe MusicBrainz has to follow this change.

I agree, a statement of intended direction would be good to have. Is this a suitable agenda item for the #MetaBrainz meeting? Or can @Rob or @Freso perhaps give a pronouncement?

Turning to what makes digital releases different,

OK, those are two scenarios. But Pomplamoose and Post Modern Jukebox above do not fit these scenarios. The music recordings were distributed first by streaming, unrelated to any physical disc, in single-song units.

And there are more differences.

Yes! The economics and technology limitations of physical LP and CD product distribution meant that a release had many mostly-identical copies. But distribution by streaming is different: no “copy” remains as static data under normal use, so there is no comparing. And distribution by static music file still allows the distributor to put a customer ID in the file, meaning every copy of the music file could be different.

If you stand back and squint, the MusicBrainz “Release” entity is heavily tied to the physical LP and CD conceptual model. Handling digital releases properly will require a different structure, perhaps a new entity parallel to “Release”. I expect that that design and implementation will be tricky. However, “Artist” and “Recording” and “Work” entities are just as effective for digital releases as for they are for the physical LP and CD concept. So we don’t have to rebuild all of MusicBrainz.


I am just thinking out loud… but this almost seems like it could be called a promotion. Meaning that the promotion is for the full release, whether LP, EP, etc… and the promotion is a track(s) as a preview. I wonder if that could get worked in somehow to tie promotions to releases. So maybe one could mark tracks 3 and 8 as been available as pre-releases. The reason I think this as the pre-releases are not real releases… once the album is out, there is no record of this. The files are all tagged as part of the album release as well, so you really are getting a piece (preview) of the album. No different than buying the whole album vs select recordings from the album. Each recording is not a release, you are just buying a part of the release. :thinking:

Maybe, in some cases, the pre-releases are only promotions. Maybe, once the album is out, the pre-releases disappear. But maybe not. Still many questions to answer:

  • How does MusicBrainz record the existence of a pre-release, while it is available and the album is not?
  • What does MusicBrainz record if the album never comes out?
  • What does MusicBrainz record if the album comes out, and the track which resembles the pre-release turns out to be different?
  • If the album comes out, the pre-release corresponds to a track, and the pre-release disappears from its official distribution location, what does MusicBrainz record about the historical fact that a pre-release once existed?

And, this talk of “pre-release” doesn’t address the Pomplamoose and Post Modern Jukebox examples.

Many interesting points in this discussion.

No, it isn’t true. It is for stores like Qobuz, where most digital releases are exact copies of existing CDs, or Amazon, where many (but not all) of the releases are ripped from CDs they sell.
There are more and more digital only releases, having no physical counterpart.
Youtube, Soundcloud, Bandcamp, have plenty of digital-only releases.

MusicBrainz handles digital releases.

Among fields that can be improved:

  • Youtube videos: standalone recording vs single, Youtube importer, clarifications (single/video/recording/…)
  • Soundcloud “playlists”, being used as albums
  • Bandcamp digital “pre-releases”: one or more tracks are released, before the actual album, to promote it
  • Digital releases changing over the time: released once, then modified afterwards (track titles renamed, tracks swapped/removed/added, cover art modified, labelled after the band sign with a label, …)
  • display could be improved, players could be embedded, etc…

About digital formats:

  • there is no list of all digital formats, because new formats are created each day, at best we can have a list of most common formats.
  • creating a release for each digital format is very likely to be a serious pain to maintain, as multiple vinyl color is: a typo in a track is very likely to be duplicated, and fixing it requires a lot of edits.

I’m totally against having one release per digital format, but we should have one digital release per release date/country/label/catno/cover art as we do for CDs.

Extra information could be added though: currently we have a relationship “can be downloaded for free …” (or the like), we could have “can be downloaded for free in FLAC | OGG | MP3 | ALAC at …” where formats are selected in a list of common formats.

Also, a thing i really dislike is excessive release duplication, this is happening with many recent vinyls, released in different colors (but that’s the only difference):

  • we currently create 2 or more releases for each of those, with disambiguation comment to differentiate them (red vinyl, green vinyl, black vinyl)
  • this leads to an absurd situation: an error concerning all releases is detected, but one has to fix it in each release, it takes time, and isn’t really user-friendly.

Digital formats are very likely to create the same situation.

So, i think MB needs a “variant” release: basically that’s a release sharing everything but few fields.
That’s not a duplicate (which is a copy), but instead a copy + overrides:

  • variant releases are based on a common source, when source is modified, all variants are modified
  • when a field belonging to the variant is modified, only this variant is modified

For digital releases, we can imagine:

  • one base: containing everything shared
  • multiple variants: OGG vs FLAC, different download URLs, different front cover, but sharing tracklist, label, release date, etc…

For vinyls in different colors, it would be the same:

  • one base: vinyl with release date / label / tracklist etc…
  • multiple variants: one for each vinyl color

It would help in another case: same release with unique cover arts (some artists do limited series of CD releases, having each an unique front cover)

Imho, having “variants” would help a lot in many fields, because it will reduce data duplication.


My intent here was to get the process started if the task of dealing with digital is to be addressed at all. The items you mentioned are a different topic it might seem. We could name off a ton of examples that are a bit strange, but that will get nothing done if we try to attack everything at once. Before we can think about capturing YouTube, Podcasts and what ever else, we need to firest address the idea that digital is more than just “digital release”.

In your example though, the iTunes release fits in well. YouTube however as a whole is something that needs attention on its own. Personally, I do not consider YouTube to be a release. I am not saying I am right or it should be that way, but that is my personal thoughts and feelings. Otherwise, you start getting into every video and person singing in the shower being a release of some sort.

That brings up this though… what is a release? Then, who can release a release? I think that needs to be defined a bit more clear. If anyone can release a release, then my playlists can be entered? Or what about my video of a bird singing, that can be a release? Those are crazy examples, but I hope the point is there… there needs to be some definition on what is cataloged and what is not. Currently, MB does not index bootlegs if they do not meet criteria, and I have actually had an add of a bootleg be voted no on and not allowed because there was not enough support to make it a “release”.

I do not mean to say your concerns are not valid, so please do not take it that way. Those topics for sure need addressing. I just wonder if taking step one first is the best road to progress.

I agree on your statement on this… not true. Sometimes it can be true and other times it is not. Additionally, even when it is true, it is technically not. Although a RIP can be done, unless it goes to a proper FLAC or similar, it is no longer a duplicate of the CD… for example.

Weird statement, as many artists only release their recordings on Youtube (with or without video).
Youtube is still the second source for on-demand streaming music (behind Spotify).


I don’t consider Spotify to hold releases either. Again though, that is just me. I am not saying they are not releases, but to me it is just like radio that is mostly all requests vs station selections. To me, a release is something that I can get or obtain. On Spotify, I am at best renting the rights to listen to music… or I am using the free service, but that just means I am not being charged to “rent” anything. I would think for cataloging that type, a separate section or method be there. Just like a car rental vs purchase or anything else.

An odd note… on Spotify I had a list of recordings… a playlist I guess, of an artist. As of today, about half of those recordings can no longer be played by me due to geographical restrictions. This happened once before as well, then they came back, now gone again.

Not really, a release is something made available to, at least, few people, officially or not.
Basically than can be anything, streaming, download, physical medium, in any format.
MusicBrainz should be able to catalog all of those (and, in general, is).


  • a promo CD-R, distributed to 10 people
  • a song made available on Soundcloud only during 3 days
  • a self-recorded video of a cover (as many can be seen on Youtube)
  • a streamed live concert recorded by a smartphone, made available via Facebook Live Video (a modern
  • a compilation of songs grouped in a playlist by the artist, or a third-party (Spotify, Youtube, etc…)

Those are actual releases, MB may keep track of those, if useful for at least one person.
There’s no issue with that, but the fact MB was initially made to replace Gr*ceNote, FreeDB, etc… when CDs were the main medium for music.
Things changed a lot since, and MB has to adapt.

This where MB is failing: it is very oriented to “albums”, when many artists don’t release any, but instead standalone tracks, eventually grouped in playlists.

After all, a CD is a bunch of standalone tracks, grouped in a playlist. A Deluxe CD is often the same playlist with few more tracks. A Spotify “album” is a playlist organized by the artist (or people having rights on the music).

Finally, an “album”, an “EP”, are just another way to say a “long playlist” vs a “shorter playlist”. I really dislike those in MB, because it is very subjective, some artists release a bunch of 8 tracks, total duration 40 minutes, and call it an “album”, some others do the same, and call it an “EP”. For me the important point here is: it is an artist-organized playlist, with individual recordings.

Another case, the “album-as-one-track”, where all individual recordings are appended as a single “track” (with silence or mixed between them) vs the usual album where the same individual recordings are splitted in different tracks.
MB is (very) bad at handling subtracks (see Release “Old School Flava !” by Lord Finesse - MusicBrainz vs


I just mean that me personally, I think differently. In your logic (not saying you are right or wrong), if you hear it on the radio, that is a release then, as it is streamed and goes to at least a few people. This is where I think this is a different type of release. They are difficult to track, they change (or at least can change) often, etc. Saying you listened to a recording on SoundCloud on July 8 but there is nothing there to reference now is no different than saying I was listening to 99.4 FM and heard _____.

I do not at all oppose the efforts of others to do such a thing, but for me, I would take a step back on that portion and let others handle it. I see things in the scope of we cannot properly identify an iTunes release from a HDTracks release, and those are real official releases just like a CD and vinyl. So I just feel that before we tie up loose ends, we should get the rope and knots to get those loose ends to tie up first. All of this Depending on the exact intention of MB, if I the user come to MB for information, odds say I am looking for releases such as my CD, iTunes file, HDTracks file, Amazon file or even trying to tag files I ripped from my CDs.

I went through this discussion, similar, in the topic of MovieBrainz. I had suggested to start a database to track movies and TV shows. Others started to add YouTube and all those other things. Once that happened, it sort of just died. That said, I do see a place for what you all mention, I just think there is a separation. I cannot say physical vs non-physical, but maybe tangible vs non-tangible, although that still is not proper terms.

Some ideas… an album/release/recording/etc has…

  1. a broadcast release
  2. a stream release
  3. digital release
  4. physical release

I personally don’t want a separate release for every digital store either, but I just want to point out that this could be solved by implementing

1 Like

Just wanted to share some more thoughts on the topic. Please note that nothing is intended to be argumentative, but to bring the debate to the surface, and that means things we may not like or disagree with, but if they are not discussed, nothing will happen.

Looking at a release, we could look at where is the presented material coming from. These are just my thoughts, I am in no waying saying this is how it should be, or even making this as a suggestion at this point, but I wanted to get the process started. What do I mean…

  1. A CD. The source of the CD is the vendor/label/etc. The CD was placed for availability/distribution/etc by that person/company/entity, who has permission and authority to do so. Ok, so this is a CD release, makes sense I think.
    1a) CDs like a Target Exclusive and similar, I think the store needs to be added in as a vendor along with the label;/company/etc. Difference being it is available only there vs a regular CD which can be at any retailer really. So a Target Exclusive CD will need to have a Target reference.
  2. We hear a song on the radio. This is a reproduction of a release, which in older days was a promo CD, record, etc from the vendor/lebal/distributor/etc. That is the release, not the radio boradcasting of it. So here, we have no actual release. Maybe a streaming entity of sorts as a release, but not a normal release.
  3. YouTube, this content is placed here by anyone and everyone who can create, steal, or anyone that really can use a computer. So, we look at the videos with qualification factors. If the video is placed there by the band/vendor/distributor/etc, that is a release. A playlist is not a release, it is a reproduction of releases, just like the radio, delivery of a release.
  4. Spotify is not a release, but like the radio, they reproduce the relases they are given permission to. So wht content they play is the release. This is similar in comcept of me making a mix CD and calling it a release. It is nothing more than me taking releases and reproducing them. When I use Spotify, I am selecting releases in whole or in part and playing a selection of releases. Not sure if that makes sense, but I am meaning that Spotify is not the release, but the “streaming service” is the release. So Spotify, Google Play Music, Apple Music, etc are all in the same release as a stream release. This stream release then has different references for the providers of a stream.
  5. Podcasts, I think those are then releases. They are original content placed for availability by the hosts/company/etc. This would have to include radio shows then too I guess as it is the same thing in the big picture. So you could have the broadcast release of a radio show and a CD version as some do that as well.
  6. A random video I might record of two birds in my back yard, I call them a band called The Birdies. I suppose this crazy example I brought up is a release. If someone really wants to catalog those, it does fit the criteria, I would say it unfortunately fits the criteria, but that is my opinion only.
  7. A concert recorded by the audience, I have an opinion here. Now we could say all recordings are a release, so if there is an official one that is a release and if 20 others record it, you have 20 bootlegs. My opinion would be that is there is an official recording of it, that one is the release and no others unless there is special circumstance where it may capture something unique or special. If there is no official release, then the bootleg release comes the main release should there be one. This is loosely worded and would be difficult to even track, so it is likely that any recording that is prepared afterwards is a release.
  8. If I RIP a CD, not a release. One way to look at this, I am mastering a master, that is not a release. I can use that digital copy as a representation of the release, but it is not a release of its own. Same goes for a file format conversion, RIP of vinyl or any other form of copying, replication, duplication, etc.
  9. My mix CD, playlist, etc… generally not a release. It could be a release if it is published or has something of significance. So if my material in question is posted on DatPiff, that is a release. If a band posts my bootleg on their site, that is a release. If I post a group of recordings out as a torrect, it is not yet a release. But if others take that torrent and index it on their blog, or talk about it on Facebook, or the news does a story on some pirate posting illegal music online, then it becomes a release as it has gained awareness. Similar in concept to how Wikipedia works, where pages are only allowed for entities of substance. Not a solid guideline though.
1 Like

This is all a little bit complicated for my simple mind. I know that MB’s objective is to be an authoritative database, but it depends on volunteers to maintain it, so it also needs to be sufficiently ‘easy to edit’ to encourage the addition and improvement of data.

If anything, there should be an emphasis on reducing the number of releases in the case where releases share the same recording details (which I think is at least part of the thinking behind the previous post). Is there any mileage in having ‘variants’ of a release for certain characteristics? The tracklist/recordings would be ‘shared’ (i.e. only need to be entered/updated once) and the ‘variant’ data would only specify differences from the ‘primary’ release - e.g. cover art, media, packaging, etc. Possibly, even ‘bonus tracks’ on certain media (e.g. I have some downloaded albums which have bonus tracks not on the original CD) could be dealt with as variations rather than by creating a new release record.

I appreciate that this suggestion is a database structure issue rather than a ‘style’ matter, but as things stand, the risk is of ever-expanding ‘similar’ releases as the media possibilities widen. I also appreciate that the ramifications of this suggestion would need careful examination; in particular, how it fits with the present concept of ‘release groups’. An example: I have 2 CDS that, shortly after release, were re-released as a double CD with different cover art. The recordings and track details are identical, but MB has 2 release groups and 3 releases each of which can be independently edited. (I appreciate that this is not a ‘digital’ example, but it could easily be).


100 % agreed

Yes, that is one area where the situation is IMHO similar to the details of audio formats: It is a detail I definitely want to record, but it often feels wrong to duplicate all other data just to get this detail right. So my personal current approach is to add only one variant to the database and add the info in the annotation that there are more available (and leave it to others to actually add the variants as releases if they want to).

Yes, that would be cool and could handle many cases and simplify things greatly.


We’ll discuss about such possibility on next IRC meeting, i already asked @bitmap if it is technically possible, it happens we have something similar already existing for alternative tracklists. Whatever is decided, it will not happen before next schema change.


Anyone is welcome to put agenda items for a meeting, see MetaBrainz Meeting - MusicBrainz for details. Note that neither @Rob nor I are in a position where we can give a “pronouncement” on this. MusicBrainz is a community project, @Rob and I are both “slaves” to the community’s wants and needs. If anyone would have a “pronouncement”, it might be @reosarevok as the STYLE leader, but even then, he also acts on community input/feedback and tries to achieve consensus before acting on things.

What I can say, which you will likely also be able to read from the discussions you’ve linked yourself, is that we do want better ways to deal with digital music, but until we know what those ways are, we are unable to implement them. Discussions like the one here are how we figure out what needs to be changed in MusicBrainz to better support the 21st century music landscape.


This would be a great feature to have for recordings as well as releases. Various edits, mixes, etc. could share a common source recording (performers, recording date/place/engineer, etc.) and the variants would only carry the unique data (remix engineer, etc.)

1 Like

Wouldn’t this variant stuff just add another layer of complexity? Right now it is still relatively simple what a new release is: if it looks different (different cover art, different looking medium, etc) it is a different release. Now it would be “if it looks different, it is a different release unless the difference is vinyl colour, because then it is a variant of the same release.” Can’t we just create better tools for duplicating releases instead?

1 Like