Re: [transcode-users] Re: BUG - audio/video alignment issue

[Phil Ehrens <phil@xxxxxxxx>]

Erik Slagter wrote:
> On Thu, 2005-03-24 at 23:33 -0800, Florin Andrei wrote:
> > On Thu, 2005-03-24 at 23:19 -0800, Florin Andrei wrote:
> > > On Wed, 2005-03-23 at 23:17 -0800, Florin Andrei wrote:
> > > 
> > > > Find a DVD where the PTS difference between video and audio is larger
> > > > than 200ms. Then go through the steps above.
> > > > The result will be a DVD with video and audio not in sync!
> > > 
> > > Weird: tcprobe will report the same PTS difference on the original VOB
> > > and the one generated by mplex (provided that the difference was passed
> > > to mplex unchanged). However, the first one is in sync, while the second
> > > is not.
> > 
> > Ok, i think i've found evidence that the culprit is mplex.
> > 
> > I started to increase the -O parameter, since with the difference
> > indicated by tcprobe, the audio in the generated VOB was too early.
> > Well, no matter how much i increase -O, it does not get in sync. I used
> > ridiculous values ("-O 1000ms" while the original indicated difference
> > was 295ms) and the relative position of audio and video did not change:
> > when played with xine or mplayer, the generated VOB has the same time
> > shift between audio and video.
> > tcprobe will happily report 1000ms difference on the generated file, but
> > the players do not agree.
> > 
> > It seems like mplex can only handle differences up to a certain value
> > (around 200ms or so), after which it will not create a correct file. Yet
> > it seems to "lie" somehow, it's tricking tcprobe into thinking it
> > creates the correct PTS difference, but it cannot fool xine and mplayer.
> 
> I suspect some players (or with some options) do not take the timestamps
> into account at all; the only thing afaik that mplex does, is to change
> the timestamps.
> 
> Also afaik mplayer -mc 0 would (also) discard the timestamp info. Maybe
> interesting to test.
> 
> I still think it's strange because I've corrected a/v sync issues with
> mplex various times. The only difference with other posters is that I
> found out the a/v sync difference by trying.
> 
> Are you sure, btw, the a/v sync issue isn't the other way around, e.g.
> you'd have to use negative values for -O?

I have tried to correct av-sync offset using the mplex -O option,
using mplayer to check the result, and the results from the -O
option were random... seemingly working for small values, but
actually going out of sync after a few minutes, and apparently
ignoring large values.

Maybe mplayer doesn't properly make use of whatever correction
mplex makes?

Phil
-- 
Phil Ehrens <phil@xxxxxxxx>      | Fun stuff:
The SilverLake Linux Users Group | http://www.ralphmag.org
3428 Winslow Drive               | http://www.yellow5.com
Silverlake, CA 90026             | http://www.slug.org