AcoustID and instrumental mixes

Tags: #<Tag:0x00007fe84d734460>

Here’s a special edition release that includes a bonus CD of instrumental mixes (of the primary CD), https://musicbrainz.org/release/035c98c5-29dd-4416-94ce-a42886bbc91b. There are six AcoustIDs that are both linked to the standard/non-instrumental recording as well as the instrumental recording.

I look at them feeling like I need to do something or clean them up somehow. Is it a problem that this is the case? Is it just less than ideal? Are there any actions to take? I realize from the guide that AcoustID “has trouble” with instrumental mixes, but what are the implications? Just this?

AcoustID guide: https://musicbrainz.org/doc/Guides/AcoustID

AcoustIDs in question:
https://acoustid.org/track/0f1629ed-298c-436a-94fa-b2df667e7201
https://acoustid.org/track/f3edf76a-a38e-407b-baa6-292baf51c21d
https://acoustid.org/track/3cc65ded-93bb-4538-9857-1b5083681471
https://acoustid.org/track/da4bf3d1-4ccc-4e3f-a578-1897ea5e63bd
https://acoustid.org/track/aef0a412-f5d8-4f56-9216-eb97d4aa0860
https://acoustid.org/track/b7bed84c-baee-4c86-900b-e03ba113255e

2 Likes

AcoustID is only a guide, and it is no surprise when it gets fooled like this when a track is so similar.

Do you have the CDs? You could load it up in Picard and hit SCAN. That would let you do a manual check of these numbers. I have seen some cases where the wrong details have been submitted, swapping CDs round. If you have the tracks yourself this is an easy check to do.

2 Likes

I do have the CDs and have loaded them into Picard (used it to tag the files). Perhaps my question is more regarding what to do next? The scan button is currently greyed out, do I need to put the disc back in the drive?

I have Picard set up and configured per some other MB guide, to use AcoustIDs. So it looks like it automatically uploaded the AcoustID when I tagged it. How might I see the AcoustIDs I’ve submitted? I don’t see a way to view them in Picard.

AcoustID only scans the first x seconds of a song (not sure how much sorry). So if the lyrics start later than x in the non-instrumental version, the AcoustID will be the same (correctly, though inconveniently).

If this doesn’t apply and the AcoustID’s are different between the two versions, the next step is to find which ones are linked to the wrong recording/s and then unlink them. You do this from the recording page > fingerprints (e.g. this page).

Hope that helps! It can be tricky but fun to untangle this stuff. With something like AcoustID’s an early incorrect match can certainly compound itself as people scan + tag + add more data with the wrong links over the years :slight_smile:

1 Like

Sorry, I may have confused things by saying “CD”. You need the ripped tracks. I drag the tracks into a Cluster on the left hand side. If already tagged you may have to drag them back from the right to the left.

Now hit SCAN on the cluster and the AcoustIDs get added back into the tracks. From here you can now visually compare the generated AcoustIDs numbers with the tracks in the database. Mainly to see how much of a difference you get between the Vocal and Instrumental versions. These will also be the same numbers as found on the recordings attached to the release.

2 Likes

The guide linked says “2 minutes (or 1 minute in a few cases, due to the older fingerprints).” I just checked the first AcoustID I listed (A Rite of Passage, https://acoustid.org/track/0f1629ed-298c-436a-94fa-b2df667e7201) and my files (from discs) in Picard are showing:

standard: f610aeb2-ef9a-4e2d-a549-a26122fac642
instrumental: 0f1629ed-298c-436a-94fa-b2df667e7201

The standard one looks good, only linked to 148 sources of the standard track. The instrumental however, has 162 sources for the instrumental and 347 sources for the standard track. The two most popular fingerprints (Fingerprint #11927135, Fingerprint #11927000) both look quite similar. Is this a case where both recordings legitimately share the same AcoustID? Even though there are indeed vocals between 0:30 and 2:00 in the standard song? It seems odd to ‘disable’ a recording with 347 sources.

1 Like

An AcoustID_ID is a group of fingerprints. It’s because AcoustID glosses over certain differences that the system works, so there are many cases where different fingerprints have been decided by the AcoustID server to be similar enough to share their AcoustID_ID.

Here’s some extreme examples where an AcoustID was assigned to different recordings (as defined by MusicBrainz Style):

  1. https://acoustid.org/fingerprint/47041132/compare/47041137 from https://acoustid.org/track/a145f243-7276-46a6-a000-b1696dd1cbbf

  2. https://acoustid.org/fingerprint/12635077/compare/14351739 from https://acoustid.org/track/af660683-4ca9-4f87-83bb-7c023e9419f5

I don’t know what causes AcoustID and/or Picard to scan a different amount of time. Your example appears to be a case of this:
https://acoustid.org/fingerprint/11927135/compare/43309863

4 Likes

As it gave you two distinct AcoustIDs I’d unlink the standard version from the second AcoustID.

I think overall this seems to be a corner case, where individual fingerprints are similar enough for the matching on AcoustID. E.g. the two most commmon fingerprints on both AcoustIDs are quite similar:

https://acoustid.org/fingerprint/43309863/compare/11927135

Other fingerprints on the instrumental AcoustID show bigger differences. But I’d really unlink the standard version from the instrumental in this case anyway, because it could likely have been just wrong matching. And wrong matchings like this can get repeated if people get the wrong match, but then just notice the correct title and submit the same fingerprint / MusicBrainz recording ID combination again.

4 Likes

I’ve used my AcoustIDs from the disc to appropriately unlink inaccurate recordings. Thanks everyone!

The vocals in track 4 ‘The Shattered Fortress’ do not begin until 2:20, which I understand will result in the same AcoustID. These are still so:
https://acoustid.org/track/da4bf3d1-4ccc-4e3f-a578-1897ea5e63bd
https://acoustid.org/track/aef0a412-f5d8-4f56-9216-eb97d4aa0860

3 Likes