[theora] is this FUD or not?

[Maik Merten]

On 24.03.2010 03:37, Tom Sparks wrote:
> I have just got an email was trying to source videos for theora test media
>
> here is the message:
> "Intrinsically H.264 properly encoded should be about 30% more efficient than Theora."
>
> is this FUD or not?


That's not exactly FUD, but a bit oversimplified. It is true that H.264 
is a more advanced format (it is one codec-generation younger than 
Theora), but the exact performance characteristics very much depend on 
the H.264 profile used (H.264 is subdivided into several profiles, each 
defining a different subset of encoding techniques allowed).

See http://en.wikipedia.org/wiki/H.264#Profiles for an overview.

Mobile devices usually don't support the most efficient H.264 profiles - 
the iPod e.g. only allows the "Baseline" profile and e.g. cell phones 
have similar restrictions. The set of coding features that works on a 
reasonably wide range of devices is somewhat narrow, so its is difficult 
to harvest the whole power of the H.264 format if one tries to target as 
many devices as possible (as is usually true for web video).

So is the "30%" figure accurate? Well, the exact number is of course 
subject of the selected H.264 profile.

Of course things also depend on the actual encoder used (a format is 
useless without an encoder) - and there are "good" encoders (those may 
e.g. be slow, some very well-tuned H.264 encoders exist) and "less good" 
encoders (those may e.g. be much faster; some GPU accelerated encoders 
aren't producing the same quality as their CPU-based counterparts - but 
at least they have usable speed).

My personal guess is that Theora fits somewhere into that spectrum, 
meaning depending on the selected H.264 profile and the actual encoder 
used one may get similar results with a good Theora encoder.


Maik