(I donât think itâs completely off-topic)
Speed ââchanges donât seem to be uncommon in mastering. Several of my old vinyls play at a slightly different speed compared to CDs of the same album. Itâs perfectly reproducible for a specific record. The majority plays at the same speed.
Itâs not audible (at least for my ears) ⊠well, maybe I would hear the difference in your examples
Interestingly, a different speed doesnât necessarily result in a different AcoustID. CD and vinyl submit to AcoustID 1981a5e5-a82f-4b6f-ad45-7a203d2f9e15 for The Return of the Giant Hogweed (same album):
â Compare fingerprints #23431747 and #91438189 | AcoustID
Even the AcoustID algorithm seems to consider small differences in speed as normal.
Unfortunately the full video submits to a different acoustID than the one attached to the album recording, because Nick Cave & The Bad Seeds ft. Kylie Minogue - Where The Wild Roses Grow (Official HD Video) - YouTube is more than 10 seconds longer than those from which these fingerprints were submitted. The full video starts with 30 seconds of âswampâ, then the song, at the end again 10 seconds swamp which was probably cut off.
Additional problem: the video of one of the currently linked releases, Ultimate Kylie has a length of 4:08 and Kylieâs video Kylie Minogue & Nick Cave - Where The Wild Roses Grow - YouTube is also 4:08 and has no swamp at the beginning. Thus its fingerprint will look very similar to the album recording.
The second release has no track lengths.
not really necessary, but interesting: album version vs music video, the album version starts 30 seconds later:
Another curiosity (or very interesting fact) I like to present:
Dire Straitsâ Down to the Waterline was included in the compilation album Money for Nothing, but it looks like they edited the track:
They c̶u̶t̶Âč a fraction of a second out of the middle of the track. Strictly interpreted, this would have to be a separate recording (âedit ofâ).
EDIT: Âč) Quite the opposite - the inserted a piece or stretched it.
Hard to tell. All recordings are merged and I can only say this for versions that I own. But in this case, it was not that random versions were thrown together.
Money for Nothing was released in 1988 - shortly after the successful CD launch of Brothers in Arms - and the tracks from their first album were probably remastered for this compilation. The initial digital masters on the first CD release of this album are much more different (and distinguishable poor): Compare fingerprints #11604631 and #10660834 | AcoustID Money for Nothing vs early CD Compare fingerprints #48422356 and #10660834 | AcoustID 1996 remaster vs early CD (the vinyl version submits a fingerprint to the same acoustID as the remaster)
But there were probably only new remasters for the two tracks from this album. And for the 1996 album they have all been remade.
I tried to find out more but according to additional metadata of the acoustIDs everything could be everything. I canât say if and when this version was reused.
All I know to add in to this is Brothers in Arms on CD and LP used different track edits. Trimmed for length to squish onto a single disc. Maybe it is from this? (Some notes here about it in âReleaseâ section on Wikipedia)
Brothers in Armas has longer versions on CD than on LP. This was probably one of the reasons why it was the first popular album to sell more CDs than vinyl copies.
But here 0.2 seconds of near silence were appended to the solo guitar intro after 3 seconds of a fading bass tone. The only reason I can think ofÂč is that they wanted to improve the piece. Maybe the beat wasnât kept exactly and the other instruments joined in too early. I would never have noticed a difference if I hadnât compared the fingerprints.
What can cause something like this?
Thereâs a perfect match on the left of the comparison column (until itâs probably muted) and itâs not the only track where such fingerprints can be found. It varies when and for how long the disturbance occurs. Some of them actually start with a perfectly matching part:
There arenât many reasons I can think of for this phenomenon. A problem with the calculation of the fingerprint? An analog tape recorder is running amok?
I left these acoust IDs linked. They certainly have something to do with these recordings.
Are these IDs from your own copies, or samples from others? How many copies of these are there?
If these number only 1 or 2 submissions. Some bad samples can appear from skipping and damaged CDs. Iâve had bad quality rips add unique IDs, and I doubt I am alone with that. Especially when working with an artist whose music is likely to be in many collections.
Or a cheap compilation chopping something up for length to fit more on a CD.
The âconfirmed fingerprintâ is from my own copy. The damaged fingerprint was found amongst the associated AcoustIDs.
That would be a very, very bad rip. Itâs a pity that itâs not possible to reconstruct the sound from fingerprints.
From a CD? With every track it starts normally (at least some of the tracks), then noise, then silence. It probably doesnât come from the reading process, but rather happens when encoding or writing the file.
I think, they were submitted only once (acoustid.org seems to be down at the moment). It was probably also noticed that the files werenât really usable.
Anything I see as an âonly onceâ or just two or three times on something this popular I just treat as bogus data and ignore. Just think what happens when you take an AcoustID from a CD (or LP) that is skipping. It would cause something like this.
Donât tempt me to spend part of today digging out my badly scratched CDs to check the AcoustID data on them⊠I have some CDs which literally have holes eaten in them⊠that causes âinterestingâ AcoustID data.
I just manually typed in the links to the Portobello example you posted. And that second one is clearly what I look at as potentially bogus data. Two samples, no other âAdditional user-submitted metadataâ connected. Those two samples likely came from the same person.
This is compared to something with 1451 samples. I know which data Iâd trust.
And can this numbers geek point out that your two âoddâ samples you link above are consecutive numbers? 45259973 and 45259974. Therefore submitted by the same person. Likely while his hardware was in the middle of a meltdown of some form. These are the samples you get when the Magic Smoke escapes from your electronicsâŠ
In my experience, AcoustIDs with a small number of submissions and only one fingerprint (with or without additional metadata) are either the correct recording and have an incorrect length (thatâs why a new AcoustID was created), or, if they are of a similar length, they are actually wrong (otherwise, more fingerprints would have been submitted to this AcoustID).
These fingerprints are mostly unique, because no one rips a file with 15 additional seconds again (at least not so similar that it adds itself to the same AcoustID) or if it is the result of mass editing, no one will have chosen the exact same wrong recording for their file. (except for versions of the same song)
And no one will ever create damaged files just like these in the example above.
Currently I leave such unique AcoustIDs (damaged or wrong length), but Iâm wondering if they should be unlinked. That would make the list much shorter and hardly anyone would miss them.
I often rip on a different PC to my final tagging. It is quite easy to submit a bad tag two or three times due to the quirks of Picard. And also many other programs submit AcoustIDs.
I leave them alone if the length is close. If length is over 20% out maybe notâŠ
When cleaning up some albums I have looked through attached AcoustIDs and it gets quite obvious that a specific rip of an album has been submitted with their own unique set of AcoustIDs. My head likes patterns so if I see every recording has an AcoustID with one fingerprint and three samples on every track⊠then I know that is one personâs work. Will usually leave it alone as extra AcoustIDs are fine. The problem comes when they are cross posted from other albums.
That makes the list longer. If they are not similar enough to be the same recording, I remove them anyway. I also have no way of knowing what track theyâre actually from and they might lure a version of that unknown track at some point to this recording.
When it comes to fingerprints with breaks, one usually thinks of a technical problem on the part of the user. But thatâs not always the causeâŠ
âHeavy Fuelâ, remastered 1996 is 14 seconds longer than the original version, released on the original album, singles and compilations (until 1996). But the additional music wasnât found on either end. It was inserted in two portions somewhere in the middle. The first part was inserted at ~1:21 and therefore shows up in the fingerprint:
Of course, these recordings must not be merged under any circumstances. And Iâm still thinking about a suitable disambiguation for the âremastered versionââŠ
Anytime I come across a âremasteredâ thatâs actually not a straight remaster, I use âversionâ in the disambiguation and put a comment in the edit notes as to why itâs not a straight remaster. Maybe something like 1996 version in the disambiguation. Or check ISRC and see if they have ETI listed there.
Yes, but what version? Itâs not about single or album, it appears on the remastered edition and many later releases. I thought about âextended mixâ, but thatâs not known to anyone and would cause confusion.
I use phrases like (extended version), (12" single mix), (1996 version) or a similar description. That disambig just needs to be descriptive, not necessarily the exact phrase fans are using.
I think of the person who may be in the editing page of a Release trying to select a recording. Something that in a few words can help select the correct version.
Good to see wordy annotations. Too many editors loose quality knowledge in an edit note that will never be seen again.
Maybe this example was a track that was originally trimmed for time to fit on an album and now someone found that earlier version for release.
No, I donât think so. Total length is about an hour. It was originally released on CD (1991) and there was no need to shorten a track by less than 15 seconds.
They made it longer for the remaster, for what ever reason (they didnât tell and it was hardly ever noticed - I didnât)
Me too, thatâs why I wanted to avoid âextendedâ. Nobody knows about it. Maybe (long version)?
(I donât think itâs completely off-topic)
I have some CDs which literally have holes eaten in them⊠that causes âinterestingâ AcoustID data.
