[MPlayer-users] Fast recompress

[D Richard Felker III]

On Mon, Aug 04, 2003 at 11:37:28PM +0200, Hans Meine wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> On Saturday 02 August 2003 17:14, Tobias Diedrich wrote:
> > Virtualdub works at a higher level where it does not have access to
> > Motion Vectors and it does not have a builtin encoder to which it could
> > feed such data if it had access to it. (Unless the implemented that in
> > newer versions, which I doubt)
> 
> OK, that shows my ignorance about vdub (like most other modern WinXX progs). 
> ;-) However, I find the idea of reusing motion vectors interesting (I know 
> there are dozens of DVD-copier progs which do so - while they all sell two or 
> three different backend engines under new names), although I must admit that 
> re-using stuff like that sounds to me like introducing new sources of 
> errors/inaccuracies which probably reduce the movie's quality.

Reusing motion vectors is supposedly known to be a stupid idea, at
least according to some of the codec gurus... This makes sense to me,
since most of the time spent on motion estimation is refining the mv's
at the halfpel/qpel level, not figuring out the general direction of
motion...and you probably would need to redo this last step anyway to
get good quality.

> I am happy with mencoder/ffmpeg, except that they are not yet capable of 
> creating SVCD-conform MPEG2 streams (which would be useful 'cause most 
> current standalone DVD players have - at best - big problems with MPEG4).*

Yes, we need volunteers to add mpeg2 encoding support to lavc... And
perhaps also an alternate rate control engine that's actually capable
of staying within [s]vcd parameters...

Best solution for now is to get a good card with tvout and use mplayer
for playing movies on tv.

> *: I know the mencvcd-way but the resulting quality is much too low to justify 
> its sloooooooooooooooooooooowness. ;-/

Yes, mpeg2enc is horrible crapware... :(

Rich