On Sun, Oct 07, 2012 at 08:56:50PM +0200, Atte André Jensen wrote:
> After running "mp3gain -r *.mp3" on the same files as I tested with
I'd expect R-128 to do a better job on content that has a somewhat
wider dynamic range than typical pop music: classical, film sound,
etc. That said, if 'mp3gain -r *.mp3' produces the results you need,
then that's what you should use.
No loudness adjustment algorithm can do 'the right thing' fully
automatically in all cases and replace the sound engineer's
judgement, for the simple reason that not everything should have
the same loudness. What R-128 and similar systems can do is to
provide reliable inputs even in conditions when you can't trust
Yesterday I recorded a live performance of 'Vortex Temporum', a
marvellous piece for piano, string trio, clarinet and flute (and
the 'bass' versions of the latter two), by the French composer
Gerard Grisey. I don't have the file here ATM, but I'll run it
through ebur128 when I'll do the editing one of these days, as
I'm curious to know what that will produce.
I was watching jkmeter during the recording. The loudest parts
were around -20 dB RMS (having attended the rehearsal as well I
allowed for some margin, peak level was around -5 dB or so). The
softest parts, lasting for a minute or so, were some 45 dB lower
which is extreme even for this kind of music.
No loudness measurement algo will be able to make sense of that.
As the recording is intended for radio broadcast, I'll have to
adjust levels by something like 15 dB, which won't be easy.
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
Linux-audio-user mailing list