[MPlayer-users] Re: Problem with mencoder

D Richard Felker III dalias at aerifal.cx
Fri Aug 15 22:46:46 CEST 2003


On Fri, Aug 15, 2003 at 08:46:58AM +0200, Stefan Seyfried wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> D Richard Felker III <dalias at aerifal.cx> writes:
> 
> >> just use "-oac copy" or "-oac pcm" in the first pass (which can go to
> >> /dev/null), so you also only need to encode audio once (in the second
> >> pass).
> > 
> > It would be nice if this were possible, but actually, because of
> > bugs/misbehavior in mencoder and/or lame, A/V sync will not come out
> > the same both times unless you use EXACTLY the same encoding options.
> 
> even for audio? I didnt know that. I can imagine that there are problems
> with oac copy (codecs behaving differently ...) but what about pcm?
> Shouldnt mplayer be able to maintain A-V sync there and later with lame?

"Should" is the key word. If you want to see for yourself, just encode
a movie once with -oac pcm and once with -oac mp3lame, and watch it
skip/dup different frames... :(

> > And that means the frame numbers and stats from pass 1 will be totally
> > bogus for pass 2, since different frames will be skipped/duplicated,
> > so quality will be much lower...
> 
> Well, since i'm in PAL land, i dont get many of these skipped/duplicated
> frames, dont have problems with pulldown and other hassles ;-))

It's not a matter of whether you get many. In fact, you don't get many
with NTSC either. But even with ONE different skipped/duplicate frame,
frame numbers will be off, so the first scene of a new frame (which
should be encoded I and probably takes lots of bits) will be confused
with the last frame of the previous scene, messing up rate control.
Actually the skips/duplicates should balance back out after a while,
in theory, so it should only mess things up for a short while after
the discrepency.....but I have no idea whether this actually happens
in practice. The A/V sync code in mencoder is very very broken,
especially with lame...

Rich



More information about the MPlayer-users mailing list