[MPlayer-users] 3-pass encodes with ogg vorbis?

[D Richard Felker III]

On Mon, Jan 19, 2004 at 09:54:54PM -0800, * wrote:
> --- D Richard Felker III <dalias at aerifal.cx> wrote:
> > On Mon, Jan 19, 2004 at 06:09:20PM +0000, Goran Hane wrote:
> > > Just a thought, but why don't use normalize for audio-normalization, just
> > > run normalize audiodump.wav after the audio is extracted and then oggenc
> > > it?
> > 
> > NEVER blindly normalize audio. It will hurt quality severely if your
> > samples frequently go near +-32767, since frequency-space codecs
> > cannot reliably reproduce these values without clipping. If your
> > normalization program is smart enough to keep the peak sufficiently
> > below the limit, it's probably ok, but you should do double-blind
> > listening tests to confirm that the quality isn't reduced. IMO it's
> > MUCH safer to adjust volume at playback time rather than during
> > encoding. With ogg/vorbis you can even insert a tag in the file header
> > to automatically amplify the output.
> > 
> > Rich
> > 
> I don't understand how adjusting at playback would be any different
> than during encoding, unless the idea is that is changeable.

That's because you don't understand DCT. Get a math/engineering text.
:)

> Also I can't seem to find anything to insert that in ogmtools?

I think it's just a particular comment tag or something, but I've
never used it myself.

> Also I should mention, that I think transcode uses a linear rescale
> (or that's what it sounds like anyway) and mencoder seems to use dB,
> so I've been converting with:
> 
>     mencoder rescale = 20*log(transcode suggested volume rescale)

20?? Where did you get that number? It will definitely clip the audio,
ruining quality. 10 is the correct number for your formula.

Again, NEVER try to normalize audio before encoding unless you FULLY
UNDERSTAND EXACTLY what it involves.

Rich