theora - Sep 2009 - Many improvements and a few problems

Hi,

I must say I'm very impressed with improvments in the latest versions of 
the encoder, especially in 2-pass mode. I was trying encodes of videos 
with sudden changes from no or moderate motion, to high motion scenes. 
Most samples of Theora quality I saw on the net were very slow motion 
usually. These high-motion videos were especially hard for Theora and it 
quickly introduced unacceptable artifacts.

Now the encoder reacts better, and much better in two-pass. Just used 
ffmpef2theora nightly. Blockiness even at quite low bitrates almost 
desappears or gets to an acceptable level. Even tried a 720p50 HD video 
(I thought Theora would be very bad at that) and the result was quite 
good at around 9 Mb/s.

Now a couple possible problems:

1) One-pass can still improve I think, with a better and quicker 
adaptation to higher motion or detail parts, even going beyond the 
target bitrate for a moment, maybe. I tried -soft-target but I don't see 
a great difference.

2) Latest versions produce different OGG streams which challenge some 
players and splitters. I think the change has been somewhere between the 
stable ffmpeg2theora (march 2009) and this summers builds:

.OGV produced now are not recognized by MPC-HC's internal OGG splitter, 
which worked well for all files previously. If Haali splitter is 
installed, it can handle most files, but in some of them duration is 
incorrectly detected and the stop after a few seconds.

Has something changed in the OGG stream or in libogg or something?

> .OGV produced now are not recognized by MPC-HC's internal OGG splitter,
> which worked well for all files previously. If Haali splitter is
> installed, it can handle most files, but in some of them duration is
> incorrectly detected and the stop after a few seconds.
>
> Has something changed in the OGG stream or in libogg or something?
Nothing has changed in the Ogg bitstream or the Theora format.
However, the new encoders make use of more freedom allowed by the
format, so code that just handled the subset used by previous versions
of the libtheora encoder might be confused by newer streams.

This should not affect duration determination though, as the container
layer (which you get time from) hasn't changed, but MPC-HC might have
custom code that decodes at the same time.

(I assume that by splitter, you mean demuxer ?)

> 
> 2) Latest versions produce different OGG streams which challenge some 
> players and splitters. I think the change has been somewhere between the 
> stable ffmpeg2theora (march 2009) and this summers builds:
> 
> .OGV produced now are not recognized by MPC-HC's internal OGG splitter,
> which worked well for all files previously. If Haali splitter is 
> installed, it can handle most files, but in some of them duration is 
> incorrectly detected and the stop after a few seconds.
> 
> Has something changed in the OGG stream or in libogg or something?
can you try if it works with out use ffmpeg2theora --no-skeleton
it might be that mpc-hcs internal splitter can not deal with skeleton
streams, if so, it should be fixed, please point the mpc-hc developers
at http://wiki.xiph.org/Ogg_Skeleton

j