[mb-style] policy on merging of recordings

Frederic Da Vitoria davitofrg at gmail.com
Tue Jul 12 08:06:35 UTC 2011


2011/7/11, Paul C. Bryan <pbryan at anode.ca>:
> Three examples:
>
> 1. Remasters take original source material, transfer those to digital
> medium, with levels typically adjusted¹ (often to the detriment of the
> recording's dynamic range), then use this new master recording as the
> basis for releases. In most cases, I would expect such a recording have
> the same fingerprint over its multiple releases. Same recording as the
> original release?

Not for me. I'd definitely want to be able to separate good transfers
from bad ones.

> 2. I have demonstrable examples where a known recording was transferred
> from master to digital medium more than once. This is not a remaster per
> se—it's the exact same master recording—but levels were appreciably
> different because the audio data was sampled under different
> circumstances, and track time is occasionally different (e.g. slightly
> different fade-outs). Depending on what audio fingerprinting algorithm
> you use, each could have a different fingerprint. Are these the same
> recording?

I believe these should be merged, except if the audio is definitely different.

> 3. The same recording is released with multiple encodings, each of which
> exhibits its own its own (arguably audible) artifacts: DSD, 96 kHz 24
> bit PCM, 44.1 kHz 16-bit PCM, MP3, AAC, Vorbis. Are these all the same
> recording?

In a way yes. The fingerprints could be different, but from my point
of view, we should separate recordings where there is a practical
advantage to do so (as in your first case). In this last case,
separating the encodings as Recordings would actually be a kind of
redundancy. The only problem I see here is: those encoded recordings
would be the same as what? With a track with significantly different
digital transfers (case 1), how do we decide which of those transfers
was used for the final encoding, at least in the case of lossy
encodings. Sometimes there are indications, but when there are none?

-- 
Frederic Da Vitoria
(davitof)

Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org



More information about the MusicBrainz-style mailing list