As I noted in an earlier reply. Currently all versions of “What’s Happening to Me” are under the same AcoustID “d7a7b660-f768-4314-a7ad-76b4b78aaa5f” which makes it seem like they are either the same recording or the AcoustID service is giving errors to people who don’t know fully how it works.
If the fingerprint ID was also displayed then, for example, the recording page fingerprint tab of “What’s Happening to Me” would display something like “d7a7b660-f768-4314-a7ad-76b4b78aaa5f - #14271616” and “What’s Happening to Me (instrumental)” could display “d7a7b660-f768-4314-a7ad-76b4b78aaa5f - #14892227”, thereby removing some ambiguity since right now you can only tell which ID a recording belongs too is by submitting it and seeing which source number increases.
Oh, that’s really bad.
Are the first two really distinguishable different versions? But it doesn’t matter, I would not have thought that the “underscore version” would add to the same acoustID. Its fingerprint is distinguishable different…
However, we will probably have to live with the fact that these versions share an acoustID.
I tried to make a test myself with an explicit and a clean version of a single, but unexpected problems got in my way… I just started to open a new thread.
Well, this is remarkable:
I submitted fingerprints of the clean and of the explicit version of a song. The versions only defer by parts of a word muted (twice within the first two minutes)→ f…n – would be f…k in English
It’s not only that the submissions go to the same acoustID, both versions submit the same fingerprint (!!!):
I didn’t believe at first and submitted again, making sure the files were correctly tagged. Thus each version was submitted twice - the second fingerprint is from the vinyl: Compare fingerprints #95328129 and #95324777 | AcoustID
It’s probably less then a second but it looks like it makes no difference at all!
I don’t know why this is a surprise. This is changing a few syllables of one word. I would have been amazed if any algorithm could catch that difference.
Instrumentals can often be close enough to put out the same acoustID.
I would have expected at least 2 different fingerprints with a tiny disturbance, possibly in the same acoustID, but I must admit, I had before recordings from often played vinyls (approx. VG) that also added to the same fingerprint as the digital version (despite of occasional clicks).
In a study I was reading, they noted that Chromaprint struggles with comparing segments less than 15 seconds compared to other algorithms like Panako and the shorter it is, the worse it gets. This was from 2017 though. https://www.ijcset.com/docs/IJCSET17-08-05-021.pdf
