The Corrs – What Can I Do (or: A Series of Unfortunate Merge Events)

While doing some routine editing, I stumbled upon this hot mess of a recording. From what I gather from the editing history, some catastrophic merge happened that merged the remixed version with the original, with 80+ linked tracks. The merge itself isn’t the biggest problem though, it’s that this track routinely gets credited incorrectly in tracklistings. Spotify’s tracklist for certain releases (ex. 1, ex. 2) indicates the remix, but actually plays the original version. I’m sure numerous VA releases also credit it incorrectly. Even the title of the official music video on YouTube makes it appear to be the original while it’s actually the remix. Apparently the label didn’t like that the original version performed underwhelmingly, re-released it as the remix later, and chaos and confusion ensued. In short, this recording is fudged.

I’ve split off some tracks that I can confirm to be either the original or the remix to other recordings, but I’ve left most releases that link to this recording untouched because 1) the tracklist often cannot be trusted (esp. for VA releases) and 2) it’d be super confusing to add the remix-but-not-really-the-remix Spotify tracks to the original. I’ve of course added disambiguation comments and annotations to clarify and warn too. In its current state, this recording is a bit of a “catch-all” garbage recording. I’ve removed ISRCs and AcoustIDs from it, but the whole ID is probably tainted. It might actually be better to (eventually) get all tracks relinked and just let the recording get deleted, instead of trying to clean it up somehow.

Here are the two new and (to the best of my knowledge) good recordings: original, remix.

There are multiple reasons why I’m posting this to the forums as well:

  1. Extra documentation.
  2. To hopefully find an expert on The Corrs that can double-check the annotations. If you’re in the possession of an album by The Corrs that features either the remix or the original, please verify that I’ve correctly identified the original as the original and the remix as the remix. My copy of Talk on Corners (standard edition) is badly scratched, my disc reader just spits it out again, so I’ve had to resort to…“other means” of checking the versions by ear. I’m 99.999% sure, but this track is confusing me so much with all of the incorrect titling that I’d appreciate a second set of ears.
  3. To call on anyone that is in the possession of one of the releases linked to the bad recording to help identify which version is featured on the release, and relink the track as necessary. Please check the annotation for more info.
  4. To recruit voters to add the bad and the two new good recordings to their subscriptions to prevent these things from happening in the future, because they will keep happening. I’ve subscribed, but I’m bad at checking my subscriptions, so more reviewers are appreciated.
  5. To hold a discussion about what should be done about the bad recording: Should we let it die? Should we attempt to restore it to a consistent state?
  6. To hold a discussion about what should be done with the remix-ETI-but-really-the-original-version tracks, primarily on streaming services like Spotify, but likely also on VA releases. It’ll probably be very confusing if we add the tracks with remix ETIs but that are actually the original, to the original recording (and might lead to more bad merges in the future), but adding them to the remix recording would be entirely wrong. For what it’s worth, it seems the ISRCs on Spotify are correct, but the track names aren’t. I’ve actually been wondering whether it might be a mistake in the digital releases, that it’s actually supposed to be the remixed version but that they entered the wrong ISRC and Spotify now plays the wrong track (if that’s even possible, I don’t know how Spotify handles compilations).
8 Likes

I would merge it with one of the split-out recordings once it has no linked tracks but I guess it doesn’t really matter either way.

Obviously tracks should be linked to the actual recording.

For wrong titles (e.g., track title specifically indicated as a remix but it is incorrect) I think the error correction principle applies. We shouldn’t keep ETI if it is demonstrably false.

In the case of non-specific titles (i.e., the track is not specifically indicated as one or the other version) I don’t think anything needs to be changed – except possibly if a single release contains both versions and uses the same title.

In either case, the release annotation can explain what’s going on. Hopefully anyone who comes along later will see the annotation before editing the release tracklist.

1 Like

I think this is an interesting case - the label may not be making a mistake, they may be consciously re-branding the remix as the ‘actual’ track. In which case it’s wrong but not a mistake and I wouldn’t change it. $&#@ it makes a mess of the DB though…

Needs to have a pretty solid disambiguation and annotation in place either way - incredible job so far @ROpdebee

1 Like

I’m not sure it’s the label’s concious decision to rebrand the remix as the original, there’s some conflicting evidence for that:

  • The official music video is for the remix, but the title doesn’t say that. The description does, though.
  • There’s a separately-titled single for the remix.
  • The special edition of Talk on Corners credits it as the remix.
  • At the same time, original isn’t on the special edition at all (I’d expect both to be on there)

If a VA comp credits something as the remix while it’s the original, I would agree that the error correction principle applies.

However, the reason why I’m not so sure about the Spotify situation is because I’ve also recently noticed other discrepancies with VA comps on Spotify. Take track 20 on this compilation. In the tracklist, it’s marked as explicit. If you load that release into d.ontun.es or a-tisket, it says the ISRC is USUM72009186, which belongs to the clean version. But copying the song link and pasting it into isrcfinder, the ISRC is USUM72009185, which is for the explicit version. I’m not sure which version Spotify is actually playing though, as I don’t know the differences between the two. I don’t know if this is a bug in one of the services or if Spotify is just returning different data depending on whether you load the track or the release. So given that the ISRC on the release is for the clean version, but on the track it’s for the explicit version, how can we be sure that Spotify isn’t just glitching out and playing the wrong version? Did the compiler of this release mean for it to contain the explicit version, or the clean one? :thinking:

1 Like

I think we may be saying the same thing? Using specific examples: what I meant to say is, if the tracklist says “What Can I Do (Tin Tin Out remix)” but the actual audio is the original recording, I think we should simply drop the “(Tin Tin Out remix)” ETI from the tracklist as we would correct any other error.

But if the tracklist says just “What Can I Do”, we should leave the title unchanged regardless of which version it is.

I suppose the only way to be sure is to know the differences and ear check them. It could be the problem is just in the APIs that atisket is using to retrieve ISRCs. Or maybe different people get different versions depending on where they live or their account status or the phase of the moon or whatever.

Digital releases are a mess :confused:

1 Like

Agreed, but maybe for a different reason. Tracklists can be bad on physical releases too, but at least if I buy a CD right now, the credits printed in the booklet aren’t going to magically change overnight. If I listen to a track on that CD right now, it’ll still sound the same two years from now. It won’t abruptly start playing a track from a different CD that I don’t even own. It won’t start playing a different song if I send it halfway across the world.

I’m willing to go through all of the digital media releases and listen to the track and check which version it is, then link it accordingly. I can trust my ears. But can I trust what the streaming service is feeding them?

Another fun one: I just noticed that the first Spotify example I linked above has all tracks listed as being 0:29 in length. But if you play them, you get the full length.

AHA! And another fun one. I’ve gone mad and went through all the What Can I Do (Tin Tin Out remix) results on Spotify. The albums by The Corrs themselves are fine, but every other track is wrong. And all of the wrong ones are (p) 202x Warner Music Group - X5 Music Group with information provided by them. However, there are also tracks that don’t have the remix ETI, actually do play the original version, and are also provided by Warner Music Group - X5 Music Group. The tracks are also mostly wrong on Apple Music for the VA comps. So at least it’s not a glitch in Spotify. But then the question is: Is X5 using the wrong ETI, or did they intend to upload the remix but used the wrong recording? Because it’s weird that they have tracks both with and without ETI, but which play the same version. I’ll try to contact them somehow to try to find out.

In any case, what probably happened is: Someone imported an X5 Spotify release, saw “Tin Tin Out remix”, reused a Tin Tin Out remix recording, added ISRCs from Spotify, the Tin Tin Out remix now had both ISRCs, based on the wrong ISRC the two got merged (probably this edit, actually), and now we’re here.

Do keep in mind that it also happens the other way around. Here’s a track on Apple Music without the ETI which is not the original version. Physical albums say it’s the Mangini remix, but it has a similar intro to the Tin Tin Out remix, so it’s a remix of the remix. :exploding_head: At least it has a different ISRC altogether.

4 Likes

It looks like this is because of Spotify (mis)matching tracks when a release isn’t available in your country. That’s a US release and I’m in the UK, so technically I shouldn’t be able to listen to it. However, only track 18 is greyed out for me. The rest will have been matched against identical tracks on other releases that are available in the UK. The sensible way to do this is based on the ISRC, but that can result in a remastered track not being matched to the original track because they have different ISRCs. So presumably Spotify must have some additional way of matching tracks, but this has a tendency to create false positives like this.

This is an extra confusing case because USUM72009186 (clean) and USUM72009185 (explicit) should both be available in the UK, so I don’t know why it would match it to the wrong one. Could the “(NOW What’s Next!)” bit be throwing Spotify off and making it think it’s a third version of the track because the title doesn’t match the others? Every Noise at Once treats it that way, which I thought was just specific to that website, but now I’m thinking it’s a Spotify thing. Either way, loading the track link from there (https://open.spotify.com/track/01t4y7sNaoVyiw6Qz6DD6b) into www.isrcfinder.com gives the expected result of USUM72009186.

Having said all that, d.ontune.es is a better source for ISRCs. It might not have links to the tracks, but it bypasses Spotify’s (mis)matching by getting its results directly from the API.

3 Likes

I did not know that Spotify did that sort of matching. The theory makes a lot of sense though. It might just not be using ISRCs at all, and just using content-based matching similar to AcoustIDs. I presume the differences between the clean and explicit versions are minute so the fingerprints would likely be similar if not identical, and it’s just choosing the wrong one. And then the track link will be wrong, leading to “wrong” results in ISRCFinder. In retrospect, the cover art displayed in that track link is incorrect, I just didn’t put two and two together.

The X5 compilations are available in my region though, so it’s not that issue. At first it looked like that site might be useful to disentangle the Spotify recordings for the Corrs, but it looks like it’s matching solely based on track names, artists, and lengths (and maybe ISRCs). It’s grouping the tracks with the wrong remix ETI together with the real remixes, so it won’t be of any help, unfortunately. But cool resource nonetheless.

On another note: For a while now, I’ve been thinking of mass crawling streaming links periodically and comparing snapshots made on different dates to discover any changes to releases. Maybe it’s time to finally get started on that…

3 Likes

sorry not sorry to necrobump… but is this still an issue? i think i have a couple lying around

2 Likes

@agatzk recently fixed a bunch of tracks, but there are still a couple of tracks linked to the bad recording (Recording “What Can I Do” by The Corrs - MusicBrainz). There’s also a whole page of single-use recordings on compilations that can probably be merged into one of the two good recordings. So if you could help out with those, that’d be great!

1 Like

I’d propose changing the data quality on the linked Release(s) to Low if we’re fairly confident that the linkage between the track(s) and Recording(s) are a mess.

1 Like

I wouldn’t oppose, but it’s a bit of a blunt instrument that might not point the finger at the right place. In general, these releases are probably fine, correct track list, countries, bar codes… all except that bloody remix. It’s the recordings that are crap, not so much the releases.

The one I have right at hand is quite obviously the Tin Tin Out - can’t really miss where they’ve been. The original digital tag did say Tin Tin Out. Something ran it over as non-remix, and the most likely culprit is Picard. This is such a mess that our own tools might only be making a hot mess.

I’m no expert and have only a CD and a couple of compilations at hand - and now am questioning which CD I have. I have a friend who likely has more. It will unfortunately be a long process. But hey, at least you’re not trying to make sense of Kylie’s Better the Devil You Know. That one will keep a person up at night.

1 Like

Maybe we can make a more general thread like “Conflated popular recordings” for these purposes, but in the latest edition of this problem, “I Love Your Smile” by Shanice. I’ve already done some work merging/changing and adding annotations, but there is no shortage of unknown recordings left out there. When I eventually call my efforts mostly done on this task, I will ponder the question of what to do with all the unknown recordings… merge them all? Or leave them all separate with annotations?

3:49 radio version: Recording “I Love Your Smile” by Shanice - MusicBrainz
Very popular 3:51 Driza Bone remix: Recording “I Love Your Smile (Driza Bone remix)” by Shanice - MusicBrainz
3:49 alternate version: Recording “I Love Your Smile” by Shanice - MusicBrainz

Radio edit: https://www.youtube.com/watch?v=ysYyCElzB0A
Driza Bone remix: https://www.youtube.com/watch?v=pOrnltjPoRc