HTML5 video element codec debate reignited
Posted Feb 5, 2010 19:00 UTC (Fri) by gmaxwell
In reply to: HTML5 video element codec debate reignited
Parent article: HTML5 video element codec debate reignited
Yes. As a format H.264 has a lot of attractive things in it and among those tools in H.264's collection is basically a superset of Theora. If your technical comparison points are only quality-per-bit and not decoder computational complexity, decoder implementation complexity, licensing, or... "Assume an ideal spherical encoder in a frictionless market" H.264 has more signal processing tools in the decoder, so it will win that kind of comparison.
Though it's important to note that what we can actually compare the quality-per-bitrate of is encoders, not formats.
Quoting Timothy Terriberry,
"There are _so_ many things you can do wrong in an encoder that do much more harm to quality than a clever optimization scheme or a complicated, patented scheme can improve it. Most encoders do some or all of them, and the original VP3 encoder was no exception. With Thusnelda we're doing things a lot smarter than we used to be, and that will only get better."
Libtheora doesn't beat the quality-per-bit of best H.264 encoders (such as x264), and won't if they also keep improving, but it does very well against some very popular ones. So thats another layer to question of 'good enough for the proposed uses and whether or not other tradeoffs make Theora preferable': "You're already using an inefficient H.264 encoder, Theora isn't worse than that. Why not use Theora and help contribute to a world without format royalties?"
to post comments)