On Tue, Nov 4, 2008 at 1:53 AM, Bob van der Poel <bob(a)mellowood.ca> wrote: I see that there is a replaygain script for Amarok, though I haven't used it. If you can get it to work, replaygain is probably the best solution, since it doesn't change the data (as normalize-audio does, I believe) but rather sets an id3 (?) tag. Be warned though that there is no perfect solution, since different types of music are compressed differently. No matter what strategy one adopts, a classical piece will have quiet parts and loud parts, and whichever of these you try to match with the volume of a pop piece, the other will sound wrong. Then of course there's the tricky issue of whether to change the volume per-song or per-album... HTH :) jmmcd

On Tue, 2008-11-04 at 11:25 +0700, Patrick Shirkey wrote: it doesn't need a CPU, it would need a time machine to travel into the future to measure the loudest sample still to come.

On Tue, 2008-11-04 at 14:08 +0700, Patrick Shirkey wrote: that doesn't meet my understanding of the term "on the fly". look, normalization is a simple process: find the loudest sample. get the ratio of its value to the maximum possible sample value. multiply every sample by this ratio. done. no shortcuts, but also no complexity. what you seem to be after is a compressor, which will keep the level of the output within some defined range. entirely different beastie.

On Tuesday 04 November 2008 02:29, Patrick Shirkey wrote: I have a script that makes highly compressed, reduced frequency and bitrate mp3s of my entire collection, and for that, plain old sox works fine. Normally, I would use a plugin for this sort of thing, but since I most often listen to mp3s in the car and no portable mp3 player I've ever owned (from a $30 MP3 CD player to my current 160GB Archos jukebox) has had the kind of compression I'm looking for, maintaining the second set of mp3s works better for me. Here are the relevant lines: (where $mp3 is the mp3 file, and $resamp contains a sox resample clause when the mp3 isn't 22.05KHz to begin with) my $maxfac = `sox "$mp3" -e stat 2>&1 | grep Volume.adjustment`; if ($maxfac =~ /Volume adjustment:\s+(\S+)/) { $maxfac = $1; } else { $maxfac = 1.0; } system("sox -v $maxfac \"$mp3\" -r 22050 /tmp/mp3carencode-$$.wav $resamp compand 0.1,0.1 -120,-20,-60,-15,-40,-15,-20,-9,0,-8 0 -6 0.1"); I'd paste the whole script, but since it's oriented around traversal of my own mp3 tree and creating a second, compressed tree, it wouldn't be of general use. The above compression values are pretty severe, so someone who just wants their collection to all be the same volume will probably want to tweak them (the comma-separated values are input and output volumes in dB.) My partner used the same compression settings without a frequency conversion for his own listening on headphones and speakers, but his ears started going long before he did. Rob

On Tue, 4 Nov 2008 17:32:36 +0100 "Emanuel Rumpf" <xbran(a)web.de> wrote: You are right. I don't want the dynamics of the tracks changed. I want the overall 'loudness' of each track to be similar so I don't have to keep twiddling the volume control. -- Will J Godfrey http://www.musically.me.uk

You might want a loudness model, which predicts the psychoacoustically experienced loudness of a sound. This is more or less related to the excitation on the auditory nerve, and should work to bring even different styles of music to similar "loudness". I don't know if something like this exists for linux (a long time ago I implemented a loudness model written in C for my master thesis, probably based on the one propagated by Brian Moore: B. C. J. Moore and B. R. Glasberg: A revision of Zwicker's loudness model. Acustica 1996, pages 335-345). If there is interest in the community, I might look at the old sources and pack them into some kind of "loudness-library" (might be a useful feature in compressors anyway). Basic concept of loudness perception is spectral loudness summation, i.e., broadband sounds (e.g. distorted music) are perceived louder than narrowband sounds of same level (e.g. classical music). Long term loudness is dominated by the loudest few percent of a sound (but not by the global peak). Giso James Stone wrote:

On Wed, Nov 05, 2008 at 11:15:47PM +0000, Folderol wrote: I was just going to ask: is it possible to modify an MP3 (without fully decoding/recoding it) in such I way that only the level is changed ? And second question (to Vorbis experts): does Vorbis have a 'hard limit' on signal level in the way a linear PCM file (e.g. plain 16-bit) has ? Ciao, -- FA Laboratorio di Acustica ed Elettroacustica Parma, Italia Lascia la spina, cogli la rosa.

On Thu, 6 Nov 2008 08:28:51 +0000 (GMT) Luigi Curzi <luigi_curzi(a)yahoo.it> wrote: Yes. I was aware of that, but it is of no help when the files will be played on a simple player, or a machine that doesn't have a program like amarok -- Will J Godfrey http://www.musically.me.uk

On Thursday 06 November 2008 09:28:51 Luigi Curzi wrote: mp3gain can also change an mp3's loudness in a way that does not require support by the player (i.e. without replay gain tags). The limitation of this method is that the volume can only be changed in steps of 1.5 dB, as is explained here: http://wiki.hydrogenaudio.org/index.php?title=MP3Gain

On Tue, Nov 4, 2008 at 10:57 AM, Martin Wohlleben <martin.wohlleben(a)gmx.de> wrote: RVA is nice at the album level as it adjusts the overall level from album to album but keeps the relative volume of songs within the album consistent with each other. This way quiet songs stay 'quieter' while louder songs match those on other albums. Important to those of us who listen to complete CDs as opposed to playlists. Cheers, Mark