Re: [tlug] MP3 normalization

[Travis Cardwell <travis.cardwell@example.com>]

Hi Bruno,

On Wed, Dec 18, 2013 at 3:57 PM, Bruno Raoult wrote:
> - One CD for home, where the volume levels of the 2 songs are similar (even
> this is difficult to explain, but let say I don't want to have to change the
> volume between songs to hear them correctly at the songs loudest parts, but
> I would like to hear the Bolero as the original, from very quiet at
> beginning to loud at the end ).

Technically, normalization of the songs should be able to achieve this
goal.  In practice, songs with widely different dynamic range will
sound like they have noticeably different volume levels to the human
ear even when normalized.  When you want to hear the original, I
recommend not processing the track at all, as it is likely already
mastered by a professional.

You can learn about normalization in general via the following search:

https://www.google.co.jp/#q=audio+normalization

> - One CD for my car. Here, I would like to be able to listen to Bolero, even
> at the beginning, but still the loud part not destroying my ears. So the
> volume amplitude of the Bolero should be limited. Let say for instance I
> would like the beginning to be the same volume as the original middle part
> of the music, the end not being changed, and everything in-between being
> relatively correct to both ends. This means I would like to have the Bolero
> to look louder and louder during the playing, the end having a similar
> volume to "Rehab".

Processing the songs so that the dynamic range is compressed should be
able to achieve this goal.  This is easier said than done, however,
especially when trying to process a set of songs with widely different
dynamic range in search of consistent output.

You can learn about dynamic range compression in general via the
following search:

https://www.google.co.jp/#q=dynamic+range+compression

Here is a pretty accessible explanation:

http://documentation.apple.com/en/finalcutpro/usermanual/index.html#chapter=52%26section=5%26tasks=true

> PS. I believe radios do this kind of manipulation permanently, especially
> when broadcasting classical music.

Classical music tends to have high dynamic range, so compression tends
to be very noticeable.

Pretty much any audio utility should be able to do normalization and
dynamic range compression, as both are fundamental effects.

As Alberto has already pointed out, Audacity can do both:

* http://audacity.sourceforge.net/
* http://wiki.audacityteam.org/wiki/Amplify_and_Normalize
* http://wiki.audacityteam.org/wiki/Compressor

On Linux, I use sox for such tasks:

* http://sox.sourceforge.net/
* http://sox.sourceforge.net/sox.html (see the "norm" effect for normalization)
* http://sox.sourceforge.net/sox.html (see the "compand" effect for
dynamic range compression)

The following page shows some possible "compand" parameters for
various types of audio:

http://forum.doom9.org/showthread.php?t=165807

Dynamic range compression can be difficult to master.  It will likely
involve a bit of guess and check for each song, at least until you get
a feel for the effects of the parameters.  I recommend ending the
effect chain with a "compand" and tweaking the parameters until you
get what you want, and then only adding a "norm" effect at the end
when satisfied.

Sox is much more advanced than Audacity.  You may be interested in
investigating other effects such as "mcompand" or "loudness" when
working on a particularly tricky track [1].

Travis

[1] Warning: NSFW Hip-Hop
http://www.youtube.com/watch?v=gcJbsfD_Dug
(One needs good speakers and good ears to even hear much of this track!)