[MPlayer-users] vhq

[D Richard Felker III]

On Mon, Jan 13, 2003 at 10:03:48PM +0000, Ian wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Moritz Bunkus wrote:
> 
> >Don't change the values between passes. ALWAYS use the VERY SAME
> >settings. Only change them if you REALLY know what you're doing. As
> >you're asking these questions you do not belong to the latter group,
> >so your answer is: 'don't change them'.
> 
> Just realised I asked the wrong question - sorry!
> 
> What I really mean is I am using an experimental "3-pass" method where I 
> do a first quick grab at up to 9000 kb/s and mp3 the sound with no hq or 
> qmax/min, then I do a regular 2-pass compression down to about 900kb/s 
> with -oac copy, hq and qmin=2, qmax=30.
> 
> What I want to know is where is the best place to do the hq and 
> qmax/qmin - at the first stage, second stage or both.
> 
> From what I have been reading here perhaps the answer is both.  Might 
> reduce the initial 9000 kb/s though ;)
> 
> I'll try all combinations of course, but it would be interesting to get 
> some technical opinions!

Is the first pass a live capture? If so you should use constant
qscale=2 or even 1 for maximal quality. Otherwise reencoding at the
second step will reduce quality quite a bit. There's no point in using
vhq during this pass unless you have cpu time to spare...but if you do
it could make the file somewhat smaller.

Another approach you could try, but keep in mind this is BLACK MAGIC
and may not work well at all... But if it does work it could save lots
of time. Capture the first time vhq:vqscale=1:vpass=1, then use the
log file from the capture to jump right into the second pass of the
twopass process. Again this is not a recommendation, but something
very experimental you could try, which *might* work.

Rich