[MPlayer-dev-eng] DivX5 & VBR encoding

[Felix Buenemann]

On Monday 09 September 2002 03:08, Kim Minh Kaplan wrote:
> Felix Buenemann writes:
> > on pass one it uses much higher bitrate (~1700kbps for me instead of
> > ~900kbps), on pass two I don't really see a difference to divx5 one pass,
> > high action scenes are still blocky.
> > I tested with latest divx5linux lib.
>
> Yes, pass one does constant quantizer (2) encoding, so you better
> direct the file to /dev/null.  My personal feelings on quality is that
> divx4_vbr is a bit better.  I am trying to figure out where this
> feeling comes from...  It looks like xvid_vbr does worse thant
> divx4_vbr on the beginning of the movie, especially when there are key
> frames.  After some time it gets to par with divx4_vbr.  Did you have
> time to test it on a real movie or just on short samples?  I think VBR
> really works when you have loads of bits (= long running videos) to
> work with.

I already noticed it's constant q2, when I looked at the stats file, however 
why are the fields kblocks, mblocky and ublocks always zero, are they unused 
yet?

Btw. my test clip was about 75sec off of the LotR DVD, teasing some scenes 
from The Two Towers, it's trailer-like containing many high motion scenes, so 
it doesn't work out as good as a whole movie, where bitrate can be shared 
much better, but still I got improvements with divx5 (divx4_vbr) and lavc 
(lavc_vbr). However divx5 is really crappy on first pass (lots o' blocking), 
whereas lavc already performs very good on first pass.

> > Does this patch disable divx4_vbr code or can it coexist until problems
> > are solved?
>
> Yes.  If vbrpass=1 or vbrpass=2 is omitted (i.e vbrpass=0) then the
> traditionnal encoding is done.  And if it does not really work then
> you can always #undef HAVE_XVID_VBR ;-)
ok, then it shouldn't hurt to commit, will you do it? Is vbrpass=0 the 
default.

> Kim Minh.
-- 
Best Regards,
        Atmos
____________________________________________
- MPlayer Developer - http://mplayerhq.hu/ -
____________________________________________