[whatwg] H.264-in-<video> vs plugin APIs

Mike Shaver wrote:
> b) bandwidth concerns (but even if Theora took _double_ the bandwidth,
> and _all_ the content was converted overnight, that's still only a 25%
> increase in bandwidth, plus a few percent for Chrome when it ships
> <video> as well)

Actually, looking at
http://en.wikipedia.org/wiki/Youtube#Format_and_quality_comparison_table
I'm pretty confident that Theora (once Theora 1.1 with
encoder-improvements has landed) can deliver usable quality at the
resolution/bitrate combinations YouTube uses (e.g. 320x240 at 200 kbit/s
or 480x360 at 512 kbit/s). It may not touch H.264's quality, but it'll
for sure give more pleasing results than the Sorenson-flavor H.263 FLV
content YouTube grew big with. In doubt one can also dedicate some bits
otherwise spent on audio to video (128 kbit/s audio for Medium/High
quality is plenty given how well Vorbis performs). It appears quite
possible to come up with encoding parameters "to get the job done with
good enough quality" that don't require additional combined bitrate.

I of course cannot comment on what resources Google/YouTube have
available, but I feel compelled to point out that freeing ~25% of users
(Firefox and Opera being Ogg-only) from the need of using Flash on
YouTube may very well have a strategic value to Google to enhance the
effect of their open-web strategy. Anyway, that's of course something
Google will have to discuss internally.

bye,

Maik Merten

Received on Saturday, 13 June 2009 08:18:38 UTC