I’m trying a new thing to be As Useful as Possible Without Breaking My Back; I’m going through outstanding edits with 2 yes / 0 no, of certain kinds, one being recording merges. If I would have been willing to file that edit, I vote yes and it goes down the slide.
For merges this generally means:
Identical-looking recordings within the same release group, or
Recordings that the editor says match by AcoustID, even of different releases, provided the ISRCs are a subset of what’s already on the target.
A lot of merges cite “same ISRC” as a sole basis, and here I would like to understand where they come from. An AcoustID is something that is calculated and that I can’t directly put into the system (particularly, wrongly). But I’ve seen and occasionally corrected mistyped ISRCs along the way.
Is it true that matching on ISRC should increase confidence in a merge?
An ISRC code is not required to use this service.
If you already have an ISRC code that you have obtained yourself, please enter the ISRC code when registering the song. If you do not have one, it will be issued free of charge by this service and automatically assigned to the song.
Machine-generated ISRCs are of course, by definition, meaningless.
TuneCore Japan isn’t the only company that machine-generates ISRCs without considering whether the audio matches prior releases.
At least they don’t mistakenly reuse an existing ISRC for a different version of the recording.
The more problematic cases for merges is when ISRC are reused for different versions.
Which is not very often but it did occasionally occur in my MB editing experience.
A minimum of knowledge about the artist discography is also welcome, @emmadilemma.
if the ISRC provided is “bogus” (i.e. one just randomly generated and applied)
any special information for the ISRC (i.e. [2003 Remaster])
any differences in publishing / rights holder
any differences in release date
confirmation of the duration (which, as always can be out by +/- a few seconds)
There is no “secret formula” to this work, you have to use your noggin, some educated guesswork (for example some compilation labels will likely re-use the same recording across releases) and a handful of salt (because record companies / distributors are known for applying incorrect ISRCs to recordings) to work this sort of stuff out.
I know that “un-merging” recordings later is difficult, but the occassions this occurs in my experience are very low; and these “alternate” recordings may not even trigger an obvious difference (be it duration, ISRC or AcoustID fingerprint).
Ultimately, the best method is when you have a variation of a recording is to provide an annotation that explains how it differs.
Someone may consider the few seconds difference to be conincidental, and originally try and merge it into the original recording - however that would be incorrect as I identified (with my ear holes) that this compilation version has been edited with an aggressive fade out at the end, thus generating a unique recording. This should now prevent it from being merged.
Of course, over-time we have 1000’s of recordings that have been added with no annotation/detail - so it’s a case of going back over what we can and working stuff out like that.
So in short.
ISRC’s are not a gold standard, neither are fingerprints, and neither are durations. Everything must be taken on a case-by-case basis, but due to the sheer volume of de-duplication we need to perform, some stuff will be errenously merged until someone notices - at which at that point, an “unmerge” has to happen and hopefully that new recording will have the necessary annotation/disambiguation applied to it.
This actually seems pretty back breaking to be honest
(But more voters is always great! Particularly if you stumble across some inaccurate edits)
If you haven’t seen it yet there may be some reports of interest to you here, when you need an intermission from voting: https://musicbrainz.org/reports
Welcome to hell.
Thank you record company.
Right Here - 7" Radio Edit and Right Here - 7" Radio Edit, look the same - right?
Are they the same?
Absolutely not.
One is the original mix, one is the considerably different remix.
So with this, there is absolutely no way of knowing (unless you listen to the source) which is which…
With the criteria I listed above (that didn’t include the two additional and subjective Stay in My Lane and Whose Edit Is This), I’m only looking for sure things; stuff I don’t know I leave be, and I haven’t run across any bad edits with these criteria - those don’t tend to get to the +2 / -0 filter.
The only problem edits I’ve run across are when I play Spin the Bottle, which also happens - use the same criteria, except for not caring about existing votes, and start at page 10. Sometimes I’ll discover that I’m the only other person on the site who has heard of, say, Queen of Hearts. It’s a very low-yield drill though.
I work with the first half of the 20th century, and when I prepare to merge, I first need to identify the versions created or existing for my track. imho ISRC’s are worthless by themselves. Matching accousticIDs are always welcome, especially when track timings slightly vary, or the title and/or artist name have been modified. For instance, the recording was released in 1936 by Teddy Wilson and His Orchestra, but now reads “Billie Holliday”. That’s where an accousticID is handy as a closer. You still have to do all the other research, using just one identifier is risky and won’t get my vote.
A good way to up the productivity level is to hit that ‘subscribe’ button on artists and labels you know and care about.
I’ve definitely never had to worry about looking for more things to vote on thanks to my subscriptions, and it’s all stuff I have a vested interest in which makes it much more interesting.
I have done, and the only reason I’ve been casting about lately is that there’s been stacks and stacks of bootlegs being entered on my favourite songbirds lately, and I have nothing to say about those.
It would be particularly impressive for a bootlegger to take the time to add ISRCs to live recordings. One with a true data nerd sense of humour… And the anarchist would probably deliberately add wrong ones
