Submitted by “With your purchase of On2, you now own both the world’s largest video site (YouTube) and all the patents behind a new high performance video codec – VP8. Just think what you can achieve by releasing the VP8 codec under an irrevocable royalty-free license and pushing it out to users on YouTube? You can end the web’s dependence on patent-encumbered video formats and proprietary software (Flash).”
I’ve heard people claim that VP8 is very good, and I’ve heard people claim that it is very bad (i.e. worse than Theora and H.264). But I’ve never seen any real evidence.
Is there any place I can find a comparison between Theora, VP8 (and preferably H.264) with the same video clip, similar to http://people.xiph.org/~greg/video/ytcompare/comparison.html ?
If VP8 is not significantly better than Theora, I don’t see any reason for them to use it, because Theora already has greater support (although not as much as H.264, but that is not really an option).
It doesn’t really matter what people think is the better codec, it matters what Google thinks.
Google have already made provably incorrect statements against Theora, which is referred to in the very page you linked:
This statement was made despite Wikipedia serving perfectly acceptable Theora video clips, such as the one on this page:
http://en.wikipedia.org/wiki/Theora
Since Google have already apparently closed their mind to Theora, even to the extent of buying On2 on the basis of their mistaken beliefs, I can’t see them backing down now (a matter of saving face). It would therefore have to be VP8 or nothing.
The only way Google can get VP8 accepted (when Theora is already both open and perfectly viable) is for Google to open VP8 and for VP8 to be appreciably better again than Theora/h264 are currently.
Edited 2010-02-23 02:26 UTC
I decided to test the Theora video link you provided on Wikipedia. While it plays fine with the original resolution, going full screen turns it into a slideshow.
I never had such slowdowns even with Flash video (btw, I’m using Firefox 3.6, screen resolution is 1920×1200).
Youtube, Hulu, BluRays, and everything else works fine as full screen on my machine. Unfortunately, Theora on Firefox 3.6 apparently does not.
It’s jerky with high CPU usage in Firefox for me as well but the downloaded file plays fine in MPC so the Firefox decoder may be to blame.
why should we have a codec that demands high cpu usage, how’s it gonna work on portable devices. people do not want theora. people want it to work on their smart phones.
(1) Theora is less CPU intensive than h264. It demands less calculation for the same quality and bitrate.
http://en.wikipedia.org/wiki/Theora#Playback_performance
(2) Even if it were actually true, I’d personally still much rather a codec that did demand more cpu over one that demands any royalty fees and restricts user choice of web access platform.
(3) A hardware decoder for Theora is in development. (See link above).
Apparently it is called leon3.
http://svn.xiph.org/trunk/theora-fpga/
This design will answer any doubts over use of Theora on smart phones. However, the more general observation would be that if a video is small enough to fit on a smart phone screen, then Theora is undemanding enough so as to not require such hardware assistance, even on an underpowered smart phone.
(4) On existing video graphics cards, many GPUs can be programmed for general purpose computing.
http://en.wikipedia.org/wiki/GPGPU
One of the things that one could use GPGPU for is … implementing a GPU hardware accelerated Theora decoder. Hence many graphics cards already have all the required support for hardware-acceleration of Theora video.
Edited 2010-02-23 13:03 UTC
There are many implementations of video players (other than the one implemented within Firefox 3.6) which play Theora just fine, even full screen, even without any assistance from hardware video acceleration.
http://en.wikipedia.org/wiki/Theora#Playback_performance
Any performance problems you may be experiencing with playback of Theora on Firefox 3.6 (compared with, say, Flash/h264) are therefore highly likely to be Firefox 3.6 problems, not Theora problems.
Edited 2010-02-23 02:58 UTC
I tried it too at 1920×1200 (xf86-video-intel) without any issue… The problem might not be theora+wikipedia but more on your side. It could be a lot of stuff though: firefox? video driver?
Hopefully these glitches get fixes for everybody.
After reading the comments, I tired downloading. Yet it works perfectly. That means somehow my Firefox setup is degrading the video performance.
From looking at On2’s codecs in the past (particularly VP3 / Theora), it seems like the codecs are pretty decent, but their implementation sucks.
They come up with a new codec every year or two. Once they release a new codec (or as soon as they start working on a new codec), they stop improving the previous encoder. They don’t take the time to really push one codec, before abandoning it.
Other encoders tend to be in development for years, being continually improved. Look at the improvement in h.264 encoders over the past few years, or MPEG-4 ASP encoders in the years before that. The best MPEG-4 ASP codecs used to be able to beat the best h.264 codecs, but not anymore.
VP3 is a good example. The Theora 1.1 reference encoder showed large improvements in quality just by fixing issues inherited from VP3’s encoder.
http://web.mit.edu/xiphmont/Public/theora/demo.html
In theory, VP3 should have been competitive with MPEG-4 ASP. In practice, it was competitive with the early encoders, but quickly lagged behind.
From what I’ve heard of VP8, it has the same problem. It should theoretically be comparable with h.264, but the encoder isn’t nearly as mature.
Without a commitment to keep developing it, it’ll end up lagging behind h.264. If Google open-sourced the VP8 encoder, I don’t think they’d have any trouble attracting developers to it.
That is the impression I always got with Vorbis and Theora – that there is the capacity to improve it but there is a lack of investment by Mozilla, Novell, Red Hat and other vendors to really push it to the next level. Not just about performance but they need to improve integration; I should be able to create something in Quicktime and then export it to Theora; on Windows I should be able to rip music and compress video using Windows Mediaplayer and Movie Maker respectively.
I think your problem there is in identifying the vendors which need to do something. It is not any more up to “Mozilla, Novell, Red Hat”, as they have done their bit. Properly implemented, Theora 1.1 works just fine, and is entirely competitivie with h264.
No, the parties who now need to take further action to solve your issues are … those software vendors who produce Quicktime, Windows, Windows Mediaplayer and Movie Maker.
I don’t think this is going to happen … AFAIK the software vendors involved have self-interests in h264. Perhaps the way forward for you would be to use software other than Quicktime, Windows, Windows Mediaplayer and Movie Maker.
Edited 2010-02-23 06:45 UTC
I don’t think showing that wikipedia, a site known for its text hosting, can host a few videos in theora really proves anything about its feasability in being used on the most popular video sharing site.
Its like saying a honda civic is the fastest car proctor and gamble owns, therefore F1 racers are stupid for driving anything else.
Hardly.
Surely it can be said (in anyone’s estimation) to be a suitable codec for a video sharing site if Theora can provide 4,399 video clips at competitive filesizes and bitrates (54,826,751,050 bytes in all) suitable for hosting on a website such as wikimedia:
http://commons.wikimedia.org/wiki/Commons:MIME_type_statistics
(Hint: for Theora, the media type is VIDEO, and the mime type is application/ogg).
Good try at a criticism, I suppose, but perhaps you need to look for another angle. Better luck next time Bill, Shooter of Bul.
PS: Heaven knows how many Theora videos are hosted and served at Dailymotion’s openvideo site:
http://openvideo.dailymotion.com/en
http://openvideo.dailymotion.com/us
but surely that site also qualifies as a video sharing site.
http://blog.dailymotion.com/2009/05/27/watch-videowithout-flash/
http://www.fsf.org/blogs/community/dailymotion-theora
PPS: Well I guess that answers my own question … openvideo.dailymotion.com has over 300,000 Theora videos.
Edited 2010-02-23 09:24 UTC
Well, its nice to have some facts, but those are not enough to disprove google’s claim which was about the necessary bandwidth to run all of their videos through it. There isn’t a video site that either A) displays all of its video via Theora, or B) has the amount of bandwidth of Youtube. IF there were a site that met both requirements, that would disprove google. As would the following:
For the same files of the same quality
Videos average size in Theora = X.
Videos average size in Flash(or other format) = Y.
By deduction you could see which would cost more in bandwidth.
Also interesting to note, from the daily motion blog you linked to:
<blockquote>But wait – the video quality is lower and sound is sometimes crackly… That’s normal…for now. The new encoding formats we’re using are Ogg, Theora + Vorbis. They’re not yet as good as other common codecs such as H264, ON2 VP6,ut they comprise an open format that is not patented, is free to use, and is supported by the Mozilla foundation (http://blog.wikimedia.org/2009/01/26/mozilla-and-wikimedia-join-for…). So don’t worry – it’s going to improve soon. We have a few tricks to improve the quality, but for the moment re-encoding 300,000 videos in this format forced us to compromise.</blockquote>
So, they are cheating on the video quality side (for now). Google’s argument was to keep the same quality they currently have.
The text you have found about video quality was true when Dailmotion created the openvideo site. They said at that time, and I quote, “it’s going to improve soon”. I bolded the text above where they said it.
That was then. This is now … the time of the “soon” that they mentioned has long passed now.
The grant from Mozilla that was mentioned:
http://blog.wikimedia.org/2009/01/26/mozilla-and-wikimedia-join-for…
was made in January, 2009. This grant established the Thusnelda project. That project began getting good results by the time of August/September 2009, and the Thusnelda project eventually released the final result as Theora 1.1 at the end of September 2009.
http://news.slashdot.org/story/09/09/26/1611252/Theora-11-Thusnelda…
Theora 1.1, the current version of Theora, which was released after Dailymotion started their openvideo site, now achieves that improved performance that Dailymotion mentioned.
For excatly the same filesize and bitrate, Theora 1.1 achieves almost the same quality as h264. It is so close that most people cannot tell the difference, and one needs video analysis software to be able to tell which is which.
http://people.xiph.org/~greg/video/ytcompare/comparison.html
See under the heading of “Results”, and look at the frames for h264 and for Theora (the h264 one was prepared using YouTubes own h264 encoder, so that is the quality and file-size that YouTube are getting). These two video clips are exactly the same filesize and bitrate.
OK, now look at the top of the page:
That statement of Chris DiBona of Google may have had some partial truth when he made it on June 13th 2009, but it is well and truly out of date and thoroughly de-bunked now.
Edited 2010-02-23 23:00 UTC
Now that is credible proof. Saying” wikipedia uses it” was not.
I suppose that is a fair enough comment, except for the fact that if you want to know the slightest thing about a subject, Wikipedia is probably the first place one would actually look.
Note also that the page I actually linked to:
http://en.wikipedia.org/wiki/Theora
not only gave you a Theora video clip to look at, but it also actually told you this entire story in far less words than I used.
Coincidentally, the words I quoted above appear in the Wikipedia article right next to the example Theora video clip.
BTW, does anyone know how to pronounce “Ptalarbvorm”? Personally, I thought “Thusnelda” was bad enough.
PS:
http://xiphmont.livejournal.com/48207.html
Interesting, no?
Edited 2010-02-24 01:11 UTC
Beware of generalisations.
All that comparison tells you is that Theora is close to YouTube’s h.264 encoding setup, and that you likely won’t be able to notice any difference. It’s not actually quite as good (yet – the latest pre-release versions are getting closer), but personally, I think it’s good enough for most web video.
YouTube’s encoding setup is far from the best available. If you can use x264, with all the high-profile features enabled, and all the quality settings turned up, Theora won’t even come close. Nor will any other codec, or any other h.264 encoder.
That’s not what people are actually using, of course. YouTube uses x264, but Apple uses QuickTime (which, compared to x264, is garbage). Lots of commercial software uses MainConcept’s encoder (again, not very good compared to x264).
Nobody uses high-profile mode, because that would make the content unplayable on mobile phones (including the iPhone), or even some software implementations (older QuickTime versions, for instance).
Large video sharing sites like YouTube could not get away with encoding video at 10FPS, so they use lower quality modes. They aren’t even trying for the best quality available – they’re going strictly for “good enough”.
The upshot of this is that h.264 for web video is basically cripppled. Even in that crippled state, it’s still easily good enough, and Theora can be competitive with that.
Assuming the encoder is fast enough. It’s not. Yet.
Beware of generalisations. Don’t count your chickens.
http://www.osnews.com/permalink?410651
x264 cannot touch Theora in a legal context (when it comes to being proof against attack from patent trolls), and believe it or not YouTube does prepare video for the web, and not for any other purpose.
When it comes to the choice of a codec for HTML5, we ARE talking explicitly about video on the web.
Theora 1.1 is already entirely competitive with h264 in that role.
Theora 1.1 is not the last word on Theora, however. Ptalarbvorm is in the works.
http://www.osnews.com/permalink?410685
Assertion without backup.
In any event, one encodes once, but one plays (and sends or transmits) over the web many times. Encoding speed is therefore far less important than the practical, usable-over-the-web quality vs file-size trade-off.
Edited 2010-02-24 12:42 UTC
Speaking of F1 racers, here is a recent Theora video you might like.
http://openvideo.dailymotion.com/video/xc8fam_force-india-f1-team-i…
You will need a HTML5/Theora capable browser to see it. Google Chrome or Firefox current version will work.
Edited 2010-02-23 09:40 UTC
Allright. But you shouldn’t forget to mention the ‘Save video as…’ context menu. You know it’s about the control and any kind of leaking hurts them badly.
http://www.google.com.au/search?q=video+downloader&ie=utf-8&oe=utf-…
12,400,000 hits for “video downloader”. Almost all of these are designed to work with Flash video.
Exactly how badly is a right-click context menu “save as” function going to hurt in that environment?
The idea is about page visits. If you download content, you can watch it several times, send to a friend and so on. If you can’t download the movie, you must go back to the site and this increases page visits statistics. This is important for sellings ads and so on.
Telling somebody: “right clik and save video” is shorter than “go to website XY, then download this by … read instructions … klick there”. The later is undoable for majority of people. My theory however is not based on technicality of it or surrounding legal issues but a simple conjecture: corporation = control freak.
Interesting article. Discusses what is likely to happen with web video, and the advantages/disadvantages for the players involved.
http://x264dev.multimedia.cx/?p=292
AV is possibly the area of software development where patents are causing the most harm. Compatibility there is absolutely essential if you want to write functional software. I’ve been documenting how bad the situation has gotten (h.264 is covered by 1,000 patents from 29 companies!):
http://en.swpat.org/wiki/MPEG_video_formats
http://en.swpat.org/wiki/Audio-video_patents
http://en.swpat.org/wiki/Campaigns_to_avoid_patented_ideas
And how are they suppost to earn money by doing this? Google is (like it or not) a commercial trying to make as much profit as possible. Google does not exist for your benefit but for it’s shareholders.
Just go to youtube or google video and watch “the corporation”, you might learn something.
There are two basic ways to improve your profits: (1) increase your prices or sales, or (2) reduce your costs.
The problem with (1) is that increasing your prices tends to reduce your sales, and increasing your sales might require say extra advertising, which increases your costs.
So option (2) is generally a better thing to do if you can find a way to do it.
YouTube using a free and open video codec (Theora 1.1/HTML5) of essentially equal performance to the one they are paying license fees for now (h264/Flash) is one way for YouTube to do that.
Another way might be for YouTube to make VP8 free and open, so that people will adopt it and write software to implement it, and then YouTube won’t have to use as much bandwidth.
Simple really.
Some of the jabs at Adobe’s intentions (can a company have intentions?) – are a bit harsh, and maybe only partially right – there has been innovation in the Flash player, but in the last 2 versions (9 and 10, and even 10.1) they’ve been doing a lot of house cleaning and adding a lot of low level kinds of improvements that go largely unnoticed. The criticism stands though – they need some new higher level stuff in there.
As far as the analysis of what it’ll take to unseat Flash in the video space – I think he nailed it. HTML5 needs one good coded that can be used by all the vendors. h.264 is the strongest case right now, and this part he got wrong – VP8 can’t compete because of the hardware part (unless he’s wrong about the viability of hardware decoded VP8).
All that said, Flash isn’t going anywhere – it does too much to be replaced by any HTML5 even if the entire spec was mature and usable today (and it won’t be for years yet). Bits and pieces will be replaced (especially once the tools come about – they are still missing), but by then there will be something new in Flash (unless he’s right about Adobe’s arrogance, but as I’ve stated, I don’t think he got that quite right).