theora - Apr 2006 - encoders

hi everybody!
i have two questions:
1. which encoders do you recomend? i have tried most, and i am pretty unhappy 
with all of them :( . thoggen only works on linux and is very unstable 
(crashes always during the encoding process) and most frontends to 
encoder_example dont work or are very buggy. ogmrip also doenst look very 
promising since it seems to convert dvds into other lossy formats first and 
then into theora. 
the only decent encoder is vlc. it works fine, but the encoded movies have a 
lot of "freeze-frames" (the movies 'stutter'). is this a
libtheora or a vlc
issue? because right now i am back to encoding ogm files with vorbis and xvid 
codecs...
i am also thinking of writing a proper graphical 'ripper' myself,
although
right now i definitely do not have the time( i also miss other feauteres like 
ripping multiple audio streams into one file). if i should do so, would you 
recommend using gstreamer-theora or a different encoder as a base?
2. why is theora so slow? why is theora-mmx in an exrta branch (this makes it 
hard to setup)? it would be much easier if one could decide at compile time 
whether one wants mmx extensions or not... most modern day computers have mmx 
(all of the x86 family dont they?)

thank you very much for your work!
i hope theora gets more public attention in the future.
greetings

hannes

Hi Hannes
> thoggen only works on linux and is very unstable
> (crashes always during the encoding process)
I haven't had any problems with Thoggen. It has always worked well for
me. Tim is porting Thoggen to gst 0.10 very shortly and this should
really help on the stability etc. As I understand things, gstreamer etc.
is portable so you could make a port for a different OS.
> would you 
> recommend using gstreamer-theora or a different encoder as a base?
I am a big fan of the gstreamer framework, but if you have stability
problems with Thoggen, I don't see how these would go away if you
created another gstreamer ripper.
> 2. why is theora so slow? why is theora-mmx in an exrta branch (this makes
it
> hard to setup)? it would be much easier if one could decide at compile time
> whether one wants mmx extensions or not... most modern day computers have
mmx
> (all of the x86 family dont they?)
In my opinion mmx is not the way that theora should be going. I have
used theora-mmx and was very thankful for the speed increase. That said,
my computer can do SSE etc. As I understand things, many chips for many
architectures have a wide variety of supported optimisations.

Dirac immediately went for liboil ( http://liboil.freedesktop.org/wiki/
) and I was surprised that I haven't seen more of it for the Xiph
codecs. Not only does liboil optimise regardless of which processor you
have, but it checks what your processor can do without user interaction.
It also has the obvious advantage of centralising optimisation work.
Theoretically all the mmx work could be incorporated into the liboil
libraries and then it could, as I see it, become the default branch. If
the processor couldn't handle it, liboil wouldn't use it. I think that
as the BBC develops liboil for Dirac it will become an even better
option for Xiph.

I believe that there is a maintained liboil branch but I see very little
about it and haven't seen any official releases of it (compared to the
mmx packages which were a fairly official release).

Please excuse any mistakes I may have made on technical points, but on a
conceptual level I firmly support liboil over any processor-specific
optimisations.

Aaron