[MPlayer-users] Mencoder 1.0pre4 and DVD rip

[D Richard Felker III]

On Sat, May 08, 2004 at 02:31:08PM -0700, Loren Merritt wrote:
> On Sat, 8 May 2004, Petr Kopecky wrote:
> 
> > I'm just trying to encode some DVD with mencoder but the quality is very
> > poor.
> > I've been always using 3-pass encoding with this parameters (for example):

I would recommend against using 3-pass mode (the frameno thing)
because it's known to break a/v sync sometimes. Also, adding -mc 0
will improve a/v sync.

> > mencoder movie.dumpstream -aid 129 \
> >     -ovc frameno -o frameno.avi -oac mp3lame \
> >     -lameopts cbr:br=128:aq=0
> >
> > mencoder movie.dumpstream -aid 129 \
> >     -o Nazev_filmu.avi -oac copy -ovc lavc \
> >     -lavcopts vcodec=mpeg4:vbitrate=1000:vhq:vme=1:vpass=1 \
> >     -vop scale=512:288
> >
> > mencoder movie.dumpstream -aid 129 \
> >     -o Nazev_filmu.avi -oac copy -ovc lavc \
> >     -lavcopts vcodec=mpeg4:vbitrate=1000:vhq:vme=1:vpass=2 \
> >     -vop scale=512:288
> >
> > The encoding speed was about 20fps a the quality of avi was really good.
> > Now with the latest mencoder (1.0pre4) the speed is approx. 55fps and
> > the quality is bad.
> >
> > Are there any significant changes since 1.0pre3?
> > What parameters should I use now?
> 
> Don't use vme=1. The default (vme=4) is much better quality, and
> usually faster.

Yes, NEVER use any vme except 0 (very low quality but useful for
realtime encoding) or 4 (default). The others are SLOW AND LOW
QUALITY. They have not been maintained in well over a year (maybe two
years?) so they're also buggy!!!

> Also, mbd=2 is always better than mbd=1 (aka vhq).

Yes.

> Recommended fast options: mbd=2:trell:cbp

I would recommend testing with and without cbp. My experiece has been
that it lowers psnr.

> Better, but 3x slower (and maybe tailored for my specific movie):
> mbd=2:trell:cbp:qprd:v4mv:mv0:aic:cmp=2:precmp=2:subcmp=2:dia=6:predia=6\
> :vmax_b_frames=1:preme=2:last_pred=3

IMO aic has no use with mpeg4. I also recommend NOT using B frames,
but people will disagree on this. Even if they do give a slight psnr
increase, they have quite a few disadvantages that aren't immediately
obvious.

> :naq:lumi_mask=0.05:dark_mask=0.01\
> :vqcomp=0.7:vqblur=0.3

These options have nothing to do with encoding speed, but rather
quality tradeoffs.

Rich