[whatwg] Another Theora vs H.264 comparison

[Maik Merten]

Hi David,

that's an interesting comparison because it's a bit different from what
Greg Maxwell
(http://people.xiph.org/~greg/video/ytcompare/comparison.html) and I
(http://people.xiph.org/~maikmerten/youtube/) did. The comparisons done
by Greg and me try to answer the "how does Theora compare to current
online streaming setups" question (and it appears H.264 setups used by
e.g. YouTube try to implement a tradeoff between encoding speed and
quality, making it a somewhat narrow race between those H.264 setups and
Theora), while your comparison uses the best currently available
encoders for both formats without encoding time constraints (from what I
understand x264 is indeed a very fine encoder for H.264 - you're using
ffmpeg as a simple frontend to this encoder, so it's not actually ffmpeg
encoding).

AFAIK there are still some open items for a final Theora 1.1 release

* better quantization matrices
* per-block choice of quantizers
* "smoother" bitrate management
* overall tuning of the visual model

all possibly being of use for your test vectors, so indeed: "we'll see".


Maik

P.S:

off-topic: I like the "Ogg" name because it's unique, easy to pronounce
and even works as a verb, which are nice properties to have when looking
at "MP3" or "H.264" or "MPEG-4 AVC", so I guess it can be a matter of
taste ;-)

David Gerard wrote:
> (please excuse the faint odour of dead horse around this subject)
> 
> http://www-cs-faculty.stanford.edu/~nick/theora-soccer/
> 
> The test files are actually from xiph.org, which strikes me as less
> than ideal even if they're entirely fair.
> 
> 
> - d.