[MPlayer-users] wmv, rm, mov, vp6. Which is best?

[D Richard Felker III]

On Sat, Mar 27, 2004 at 09:47:34PM +0100, Philipp Wollermann wrote:
> Hi,
> 
> Amadeus wrote:
> 
> >>>Is it considered a fact, that xvid is better than libavcodec's mpeg4?
> >>>     
> >>>
> >>No, the opposite.
> >>   
> >>
> >
> >In that case I am glad I have encoded everything with lavc mpeg4=)
> >
> Could someone please explain *why* lavc mpeg4 is better than xvid? I 

Subjective testing shows that xvid-encoded files appear to have
significantly more 'mud'/distortion artifacts during motion, while
lavc has somewhat more blocking. Subjective opinions aside (IMO the
former looks much worse), blocking is much easier to repair with
postprocessing that 'mudding' is.

If you want better evidence, do a casual subjective viewing test of
your own, a formal double-blind test, and/or psnr tests and see how
the two compare.

> always thought it was the opposite - xvid is a codec which is actively 
> developed,

And lavc isn't actively developed??? Where do you get ideas like this?
Why not rtfa (ffmpeg-devel)? New features to improve quality/bitrate
are being added all the time.

> while lavc is a sort of "all-in-one" thing. Just thinking 
> about it - shouldn't a project which is fully focused on something be 
> better than a library which has an mpeg4 encoder only amongst several 
> other things?

How does this make sense? Oh, right, you weren't trying to make sense.
FYI: all of these formats are 75-90% identical. So a _smart_ developer
is going to make one library that does them all, rather than just
supporting an arbitrary one.

When lavc has full h264 (next gen codec) support, and xvid is still
stuck on mpeg4, maybe you'll change your view...?

> I don't want to offend someone here so please correct if 
> I'm wrong ... I'm just a little confused about this.
> 
> >Doom9 is known for having stupid misconducted tests anyway, like the
> >review of lavc using ffvfw...
> > 
> >
> I don't think that Doom9 is so bad. Or does someone know of any better 
> site about this topic?

I haven't seen any good objective comparisons. Doom9 tested ffvfw
(very outdated, broken, nonstandard version of lavc) and concluded
that it was bad. AFAIK ffvfw doesn't even use the normal lavc rate
control code, and rate control is perhaps the most important part of
optimizing quality at fixed bitrate. (In fact, I believe it was using
xvid's rate control code. ;)

Rich