Is there any way to delete a busted AcoustID and start over? That or at least permanently purge the garbage data like dead recordings and wrong tracks? What can be done when one gets super messed up?
Yes, you can delete the obvious busted up AcoustIDs. Just login to the AcoustID server and you’ll get a Delete button.
Top right of that AcoustID page it says “Sign In”. Just click that and login with your Musicbrainz credentials.
You can’t MOVE AcoustIDs, but you can delete. A good clue as to bad acoustIDs is if you see a low count of samples.
It is technically possible to get two tracks submitting the same AcoustID, but I think I have only seen this twice after looking at many thousands of samples.
You will also find that some editors will add a reference to an acoustID at the top of the recording page and a link to some releases it is found on. (Example) Sometimes some heavy remastering can create a new AcoustID. (Example of remastering changes)
Remember also people with vinyl copies submit their acoustIDs. These can come out slightly different
There can be an awful mess with AcoustIDs on some artists. Much of this can come from the noobs who use Picard with 10,000 files… get a messy inaccurate match up that is not checked… and then hit that Submit AcoustID button
How can this be technically possible? The fingerprints are absolutely specific to a particular sound recording. Two recordings with different length which are are based on the same recording (a full length and a shortened version) may submit identical fingerprints (if compared) but they go to different acoustIDs because of their different length.
If you have verified an acoustID for your recording, you can safely disable all other recordings linked to it.
@Echelon666 - we are talking AcoustIDs, not DiscIDs here…
Oh I totally agree… but spotted it. And added them to a thread you were in. Just totally forgotten what they were now.
I could verify them as I owned both tracks. Freaky maths.
It is rare enough that you are pretty safe 99.9999% of the time disabling other matches. Especially when they have a low sample count. But sometimes need to be wary of a badly credited bootleg with an incorrect track list.
Popular recordings often have more than one valid acoustID, especially if there are vinyls involved. Different (individual) splits can cause separate acoustIDs and sometimes very worn records can submit fingerprints that are not similar enough for the same acoustID.
You can compare fingerprints of all attached acoustIDs to your verified fingerprint. All not similar can be disabled. Sometimes after such a clean up there are more than 10 remaining that are similar enough (according to your judgment).
I don’t quite trust your memory. It is because of the way fingerprints are calculated. Only the peaks of the frequencies at a time are used - the loudest sounds. If the pattern is similar, it must have come from recordings that sound very similar. Only very quiet sounds can be different.
Also, do AcoustID that match to 0 recordings (from them being disabled) eventually delete themselves, like empty release groups, or do they sit there forever clogging up the space?
FWIW and unrelatedly to anything else in here - I seem to remember back in the day the same song performed with the same backing track by two different jpop singers got given the same acoustid. The code might have been made more discerning in the years since though.
1 and 2 will end up doing the same thing. Generate a new AcoustID and read the value in Picard. Then look that up and disable the rest.
And yes, when I said “Delete” I meant “Disable”.
Worst case, if you disable something that was correct then it will soon get re-enabled again. Especially if you are doing a heap of work on a specific artist.
You can’t disable all active acoustIDs of a recording. Yours is only one specimen and there are most often other valid acoustIDs. But if your acoustID is shared with other (different) recordings, you can of course disable them.
My workflow is as follows:
I determine the AcoustID of my version: I usually look for the specific fingerprint by re-submitting it, but all available fingerprints would be similar enough to compare it to others (you can use the acoustIDs returned from a scan).
I compare this fingerprint to the most often submitted fingerprint of other acoustIDs attached to the recording:
at those similar to mine, I disable all recordings that are not this recording
at those not similar to mine, I disable this recording (and leave others, if present)
During this process I look at other linked recordings that are possibly duplicates. I look at their fingerprints and if they are (mostly) also similar, I initiate a merge. If it’s unclear, I leave it and do not disable it from the acoustID. When I’m finished with the clean up, the recording will only be represented by similar = verified acoustIDs and no clearly different recording will remain at these acoustIDs.
