[MPlayer-users] Re: Problem with mencoder

[D Richard Felker III]

On Sat, Aug 16, 2003 at 12:18:42AM +0200, Oliver Fromme wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> 
> D Richard Felker III wrote:
>  > On Fri, Aug 15, 2003 at 07:46:30PM +0200, Oliver Fromme wrote:
>  > > Please don't remove 3pss encoding.  Without it, it is difficult
>  > > (though not impossible) to estimate the bitrate when using VBR
>  > > mp3 audio, which I always do.
>  > 
>  > No it's not. VBR always comes out the same average (withing a +- 10
>  > kbit margin) for movie-length encodings.
> 
> That's already too much.  For a 100 minutes movie, 10 kbps
> difference means about 7 Mbyte in file size.  But I want to
> calculate the file size as precisely as possible, so it
> will fit on an CD-R and use all the space.
> 
>  > So just make a table of what
>  > bitrates you get with each VBR setting, and then use those to estimate
>  > video bitrate.
> 
> Those differ a lot.  The average seems to be around 145 kbps
> with my settings, but for some DVDs it goes up to 160 and
> beyond, in particular when the movie contains much music
> (but not always).  On the other hand, I've also had DVDs
> for which the VBR mp3 encoding ended up at 120 kbps.  With
> the same mp3lame settings.

Hmm, it's never seemed to vary that much for me. One other option is
to use abr instead of vbr (it's usually accurate within +- 2
kbit/sec), but IIRC the lame docs say the quality might not be as good
as true vbr at the same average.

>  > Or....since you have to encode audio in both passes when you don't
>  > have 3pass mode, you could just look at the resulting audio bitrate
>  > from the first pass, and then change vbitrate parameter slightly in
>  > the second pass to compensate.
> 
> Hm, I'm not sure if that would work well with 2pass mode.
> It is my understanding that the 2pass mode requires that
> the parameters for video encoding are exactly the same in
> both passes, otherwise the information from the divx2pass
> log file would be inaccurate.

Same parameters are required for everything BUT vbitrate. Some people
even recommend using vqscale=2 for the first pass, and only switching
to vbitrate for the second. So it's definitely acceptable to adjust
vbitrate by 20 kbit/sec or so between the two passes!

Rich