[MPlayer-users] Re: funky newbie problem

[D Richard Felker III]

On Fri, Dec 06, 2002 at 08:22:18AM +0000, Diego Zuccato wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> D Richard Felker III wrote:
> 
> > With that said, I've had relative success encoding it myself. The
> > commands I used were the following, but you could probably go ahead
> > and use bitrate 800 on the first pass if you don't care about the
> > pass1 filesize coming out too big.
> Which file ? divx2pass.log ? Its size shouldn't depend on bitrate, IIUC.
> Outfile ? Can safely be set to /dev/null (it speeds things up quite a
> lot...).

The actual avi file from pass 1. Personally I like to keep it to
compare quality with pass 2 and judge whether my second pass was setup
well. Speaking of which, right now I'm in the middle of a 2pass encode
that looks like the second pass is going to turn out a lot worse than
the first. Apparently high qcomp is good for pass 1 but not so good
for pass 2.

> > mencoder lavc_enc_prob-original.avi -vop scale=656:352 -lavcopts \
> > vbitrate=720:vhq:vratetol=500000:vqcomp=0.9:v4mv:vcelim=9:vpass=1
> > 
> > mencoder lavc_enc_prob-original.avi -vop scale=656:352 -lavcopts \
> > vbitrate=800:vhq:vratetol=500000:vqcomp=0.9:v4mv:vcelim=9:vpass=2
> Err... Did I miss something here? I thought that first and second pass
> should have to be IDENTICAL, except for vpass and outfile. What's the
> meaning of a bitrate change?

Well I wanted a final bitrate of about 800-820 kbit/sec. In the first
pass, high vratetol means you have to set bitrate lower to keep from
going to big. Like I said this is irrelevant if you don't care about
comparing the quality (at constant size) between passes like I did. In
the second pass, however, the 2pass code tries to match the requested
bitrate; rate tolerance doesn't seem to do much.

Anyway, for 2 pass encoding, it's very important not to change options
that affect encoding strategy, like bframes, hq mode, 4mv, single
coefficient elimination, adaptive quantization, etc. between passes.
However, changing the bitrate or qscale (min/max/const) settings might
not be such a bad idea. In fact some people seem to think it's a good
idea to encode pass 1 with constant qscale=2 to get a uniform measure
of frame complexity across the whole file, then use normal rate
control in pass 2. I have no idea if this is a good idea, though.

Rich