On Wed, Jan 03, 2001 at 10:59:18AM -0800, [EMAIL PROTECTED] wrote:
> Hey, this looks spiffy. I have something of a usage question though, in
> regards to the 'gender' (distributed ripper/encoder) program I'm
> scripting. I'd LIKE to add audio normalisation as an option, but since
> ripping and encoding tasks happen concurrently and asynchronously (but
> with file locking) I'm not at all sure how to handle a sane
> normalisation default, or even if there IS such a beast.
Normalization has to take into account the maximum amplitude each
wave reaches before scaling each value down or up to the desired level.
normalize computes the power of each small block of the the sound file
in RMS (Richard Matthew Stallmans , sorry, I couldn't resist :) :) :) )
looking for the highest value so all blocks can be proportionally scaled.
So, there is no way to have normalization ocurring concurrently
to acquisition of the sound and encoding. You'll have to record everything
first, then normalize everything and then encode the normalized file.
It is not possible to do normalization through a pipe, for an
example.
You can normalize each file to a predefined aplitude value so you can
do it once each file was "ripped" instead of doing normalization in batch
mode. The author of normalize chose from his experience 0.25 dB to be a
good power default.
I use this default to record my CDs (made out of my own midi files and
mods mostly) and I am glad with it.
>
> -- Ferret
>
