VP8: The Savior Codec
multimedia.cx
multimedia.cx
First, VP6 is a pretty decent codec, widely available for comparison. Not as good as H.264, but better than Theora and (IMO) MPEG-4 ASP. Even if Google was only releasing VP6, that would be pretty compelling. It's reasonable to expect that VP7 and VP8 are incremental improvements over VP6. No one knows exactly how much, but if a free VP6 would be exciting, then a free VP6+ would be even more exciting.
Second, Google didn't pay $120M based on On2's marketing website. They obviously evaluated On2's tech first-hand and decided that it was worth buying.
1. Will Adobe join us in introducing a brand new video codec (whether it's VP7, VP8, or something new based on them) to Flash, and if not...
2. What is the theoretical maximum quality you can wring out of VP6 without breaking backwards compatibility with the installed Flash base if you throw unlimited cash at it.
They can start up a skunkworks to build the next big codec to challange H.26_5_ if they want, but the need to do something soon if H.264 isn't going to become entrenched.
Some people will say VP6 is crap and that only better quality can win, but you'll note they only say that when something they like wins. For example, Apple fans point to the quality of iPods and iPhones as why they're successful, but point to idiot consumers, network effects and dodgy dealings to explain why Mac OS has a similar global marketshare to the perennial joke that is desktop linux.
Similarly apparently H.264 won because it's some kind of design-by-committe miracle baby, yet AAC isn't a stand out technically and coasts along on the coattails of it's video codec partners. And it's still not ousted it's inferior older brother mp3 in general usage, even with Apple's help.
Installed base matters. If Google make a video announcement that doesn't involve Adobe, I'll be very disappointed.
Perhaps Google can come in and say "it's free if you include VP8 too"
While it would be nice to see Flash Player integrate VP8, Adobe's cooperation doesn't seem critical to me.
Youtube is important because they can encourage large numbers of ordinary folks to adopt whatever solution they use, and as a large-scale example to show that whetever technology they use is feasible at that scale. But if they free a codec then the point is for all the other websites to adopt it, and for that Flash would be an invaluable partner.
Adobe says that they get near 100% update rates within a year so any new codec will get out there fast. Faster if Youtube prompts people to click the updater. Not that users of Google Chrome will need to worry, as Google has conveniently decided recently to bundle Flash and auto-update it for them without intervention.
Meanwhile, Microsoft's work in this area over the past 13 years with the WMV family of codecs has been extraordinary, coaxing more quality from a given bitrate needing less CPU to play back than any other commercially viable codec. That's why WMV codecs work on so many consumer devices, even phones. VC1 is remarkably good, only just recently challenged by H.264 encoding's most recent optimizations.
Meanwhile, if you meant file container, Microsoft's recent player technologies such as Silverlight support H.264 in .mp4 containers, as do recent Flash Player versions.
With all Windows machines supporting .WMV container and VC-1 codec, Macs still shipping with a Windows Media Player or able to play it in any QT app using Flip4Mac, Linux players supporting it "out of the box", and both Sony PS3 and MSFT Xbox playing it, not to mention most every consumer device and set top box -- if you want to send someone a video that just works, then .WMV using WMV3 or VC1 has been a decent bet.
Your newer best bet for sending a file is .mp4 with H.264/AAC, which is played by Flash Player 9 Update, Silverlight, QuickTime, VLC, etc., and more recent Windows systems.
For embedding in the web, stats say Flash is your best bet, but those stats apply to VP6 codec which looks terrible and most people don't know how to play locally and doesn't work for HD.
So your best bet for embedding is .MP4 with H.264/AAC with an HTML5 tag, and inside that tag a Silverlight player embed, and inside that tag a Flash player embed, and inside that a download link for the .mp4 file.
What do you expect from a codec that's closer to the previous MPEG-4 ASP generation (e.g. Divx, Xvid) than H.264?
VC1 is remarkably good, only just recently challenged by H.264 encoding's most recent optimizations.
Recently challenged when? Only right after Blu-ray was introduced VC-1 was somewhat better than H.264 because the VC-1 encoder implementation was better than any of the "pro encoders" that the studios used.
http://mirror05.x264.nl/Dark/website/compare.html also and there are plenty of PSNR/SSIM measurements online which show H.264 is vastly better than VC-1.
not to mention most every consumer device and set top box
iPods, iPhones? Also I have never seen a real-world VC-1 deployment on an STB. Virtual all of said consumer devices also play H.264 video.
It was only Microsoft's lobbying that led to VC-1 getting accepted in Blu-ray and (on paper but not in reality) in DVB. Even Microsoft have pretty much given up on it; H.264 is in Silverlight, IE9 and Expression Encoder 3 has its own H.264 encoder.
I know that you mentioned Theora, but you imply that the same case is applicable with VP8. You said "The same article could be applied to the Theora v. H264 tests...", i.e., Theora and VP8 both demonstrate an advantage due to unfair testing and not real-world improvements. Nobody knows that because nobody has been able to test VP8 independently thus far.
Your post implies that the author claimed the test results were not representative of the codec's true performance. The author made no such implication; he only noted that there is no unbiased verification or corroboration of those results.
The quality of the code released in the VP3 dump was abysmal, and while the Xiph people cleaned it up structurally, they didn't fix any of its warts, and they introduced new ones with their awful container format. Hopefully Google won't repeat that history.
I'm really tired of the increasing desperation from people who are now willing to hang their hopes on anything that is not H.264. It's almost like it's dogma at this point: H.264 must not be allowed to succeed because _______.
That is not all we need to see. VP8 may be a better choice than H.264, it may be a worse choice. We're not even sure Google will open the codec yet (I've only seen "inside sources confirm", did I miss an announcement?) so let's not get premature. And PSNR graphs are not the golden standard of video quality, the video is. It's possible to have artifacts that psnr graphs may not represent well, but that make the codec a poor choice.