Someone mentioned that I should be generating AcoustIDs from WAV (lossless) rather than MP3s. Will this make a significant difference, or will using MP3s be 99% as good as WAVs?
I mostly now rip in mono and with lame setting v2 and I see that AcoustID is the same as with when I ripped in FLAC when I submitted them, initially.
Even in mono, I was surprised.
※ Before I am asked, I use mono in my walkman because I am deaf in left ear.
Interesting!! Can someone confirm?
I always assumed using WAV or FLAC was preferable, just because AcoustID has separate stats for fingerprints for each, why would that be? Edit: Well, I’m sure it did at some point, I don’t see them differentiated on the stats page anymore!
It’s preferable to fingerprint as many different versions/iterations of a track as possible (AFAIK). But the “pure” FLAC/WAV would be a good starting point. The AcoustID from that and any derived lossy file should be the same, even if the exact fingerprint might not be.
@lukz is the brainz behind AcoustID, so he might know more.
The AcoustID from that and any derived lossless file should be the same, even if the exact fingerprint might not be.
You meant lossy, not lossless, right? Losslessly compressed audio is the same as the original by definition, so the fingerprint calculated from it in a deterministic way will also be identical.
For lossy compression you are right; the fingerprint may be slightly different, but still similar enough that it should get clustered with the original and other lossy versions.