[MPlayer-dev-eng] Re: XviD B-Frame CBR and FQ encoding bugfix

[Martin Drab]

On Wed, 2 Apr 2003, Rémi Guyomarch wrote:

> On Wed, Apr 02, 2003 at 01:00:04PM +0200, Martin Drab wrote:
> > Hi,
> > 
> > I'm sending a bugfix for XviD frame encoding. In fact there are two 
> > bugfixes, both are related to encoding B-Frames.
> > 
> > First fixes the choice between CBR and Fixed Quantizer during setting of 
> > the parameters for the XviD VBR engine. The condition was totally wrong.
> > 
> > Second fixes the setting of the quantizer parameter for the XviD VBR
> > engine during CBR frame encoding. It has to be set to 0 (and not -1) in 
> > order for the XviD VBR engine to fill in the correct calculated value.
> > 
> > The patch can be applied to both CVS and 0.90 release.
> 
> Thanks !
> 
> I've included it in an updated XviD driver I will commit in a few
> days. I hope dev-api-4 will be usuable shortly, not having B-frames
> support in the 2-pass driver is really annoying...

That would be really appreciated. XviD driver is somewhat too old and 
restricted. Towards that end I wrote an upgrade of that, which leaves
the current functionality untouched and when you enable one flag-argument
in the -xvidencopts, it completely disables the bad old interface and lets 
you access "directly" (almost directly - through -xvidencopts ofcourse)
all of the XviD engine and mplayer add-on VBR XviD engine parameters.
However as the old interface had too many restrictions and 
simplifications, this my "new" one has none. So it is meant to be mostly
for developers and advanced users, as you can handle all parameters and 
flags without any restriction. Not very "user friendly".
It was written for the dev-api-3 (which is now the HEAD_CVS allready).

I was planning to send it in tomorrow, however if (as you say) you have 
some new updated XviD driver, it's probably pointless.

But anyway it would be nice if anything like direct codec arguments access 
(in some simillar way as I described above) would be available (and not 
only for XviD). Because then it gives you more freedom in experimenting 
with the arguments.

A typical example of what I mean is the "me_quality" XviD argument, which 
is perhaps good when you don't know much about what those PMV_* things are 
about, but it would also be nice to be able to select your own PMV_ flags
combination.

						Martin Drab