This past week, the internet picked up — and subsequently sprinted like a cheetah with — an unsourced and highly unsubstantiated rumor that Google will open source the VP8 video codec, recently procured through their On2 acquisition. I wager that the FSF is already working on their press release claiming full credit should this actually come to pass. I still retain my “I’ll believe it when I see it” attitude. However, I thought this would be a good opportunity to consolidate all of the public knowledge regarding On2’s VP8 codec.
Pictured: All the proof you need that VP8 is superior to H.264
Update: The preceding comment is meant in sarcastic jest. Read on
The Official VP8 Facts:
- VP8 is superior to H.264. Source: On2’s website, many PSNR graphs.
- VP8 is free from patent worries. Source: On2’s website (“no patent-pool royalty hassles”).
- VP8 delivers objectively better visual quality than H.264 at the same bitrate. Source: Video encoded in On2 VP6, created by On2, posted on On2’s website.
- VP8 can be implemented on low-power hardware. Source: DSP Design Line article authored by On2’s CTO.
- VP8 uses golden frames (just like every On2 video codec since VP3). Source: On2’s website.
- VP8 uses loop filtering (another cornerstone of every On2 video codec since VP3). Source: On2’s website.
- VP8 is scalable to many cores. Source: On2’s website.
- VP8 produces objectively better results than even x264. Source: Again, On2’s website. Sorry, my x264 pals, but they have charts and graphs to back them up, so, you know…
That’s pretty much all we know about On2 VP8. Notice a common theme? That’s right: Everything we know about On2 VP8 comes from On2’s own publicity material. As of yet, we have no samples encoded in the format, and we certainly don’t have any public encoders or decoders.
So, yeah, you can write me off as being consistently annoyed when the VP8 topic crops up. Take a look at the subtitle of this blog: “Topics on multimedia…” Codecs are my business (well, hobby, anyway) and my multimedia pals and I have cataloged around 200 different video codecs on our little wiki. Call us codec snobs but we like to know technical details surrounding codecs. Or, failing that, we like to have samples we can study along with maybe a binary decoder and/or encoder so we can evaluate the suitability of the codec for certain tasks. Thus far, all we have to go on is On2’s own word about how awesome their technology is. It’s a bit odd to see so many people taking a (non-Apple) company’s claims at face value.
Do you believe the marketing material? As a baseline, you’re invited to read the VP6 and VP7 whitepapers which provide similar objective proof of those algorithms’ superiority over H.264 and other standard codecs.
Different Requirements
From most of the stuff I have been reading (articles, blog posts, and ensuing comment debates), people are lapping up the marketing material. I was incredulous until I realized that most of these observers are simply interested in different things than I am.
I — and, I suspect, many of my multimedia colleagues — are salivating for some more solid technical details surrounding this codec. It’s what we live for. I need to recognize that most outside observers take the view that a video codec is a video codec is a video codec, i.e., they’re all more or less the same and fairly interchangeable. The hope is that On2 VP8 will, at long last, provide patent and license purity that all relevant stakeholders (Google, Apple, Microsoft, Mozilla) will be able to agree on. Quality? That would be nice too, but it seems to be a foregone conclusion that VP8 offers more than enough quality, especially when considering that Theora is based on VP3, which many consider good enough, and VP8 has a higher number than VP3. Ergo, VP8 must be 5 times better than VP3/Theora. Or 5 more better. Or something. And besides, On2’s own marketing materials explicitly state that VP8 is better than H.264.
Flash-Killa
It would be disingenuous to omit the Flash-killer angle driving so much of the fervent anticipation surrounding the VP8 speculation. VP8 goes open source => all major browsers adopt it overnight as a standard video codec for HTML5 video => blight of Adobe Flash is eradicated the following week, since Flash’s only use is as a naive video player. Things move just that quickly. I’ve covered this ground already, i.e., how all of this HTML5 stuff is totally going to crush the stuff I work on at my day job.
What Is It, Really?
There’s too much speculation out there surrounding On2 VP8… but that won’t stop me from adding my own: Since, to my knowledge, VP8 has never actually been licensed or used outside of On2, the engineers could still be tweaking it behind the scenes, or even overhauling large segments of the algorithm. The list of publicly discussed features doesn’t explain a whole lot about the overall codec operation.
Writing this article motivated me to re-read and carefully study that DSP Design Line article and I must confess that there are some interesting tidbits in there. One item that caught my attention was their concept of SIMD without dedicated SIMD instructions. It’s nice to see On2 getting back to their roots– this concept forms the basis of On2’s very first video codec, Duck TrueMotion 1 (effectively VP1).
The DSP Design Line article is far more fascinating than any of the literature on On2’s site. The article paints a picture of a powerful and flexible codec that is suitable for low, medium, and high bandwidth applications and whose decoding algorithm can be scaled to operate realtime on low power, mobile CPUs all the way up to beefy, multicore desktop machines, all while offering efficient and accurate compression. If the codec offers all of these technical benefits, plus truly free licensing terms and is satisfactorily patent-free (or indemnified), I look forward to singing VP8’s praises.
Pingback: Tweets that mention VP8: The Savior Codec | Breaking Eggs And Making Omelettes -- Topsy.com
That photo is certainly not proof that one is better than the other. It’s nothing at all. First, note that the picture on the right has less noise, but also less detail than the picture on the left. Second, and more importantly, it’s not about what one frame looks like, it’s about perceived quality vs size of the entire video. As an example, consider this: FLAC sounds much better than any MP3, but you’re paying for that in file size.
(ha) I should read the whole article before posting, I suppose!
The VP6 whitepaper you link to is from 2004, the VP7 one is from 2005.
A recent blog post by an x264 developer claimed that when it was released (in 2005 according to Wikipedia) VP7 was “dramatically better than practically all H.264 encoders at the time”.
So when you say we should gauge the current marketing material on that basis, surely that means it is going to be true?
Fix “naive” -> “native”. Hi Mike! =)
Mike,
VP8 has been licensed outside of on2. One of the last companies I founded Vusion licensed it. We had a player that decoded it and encoders that encoded it quite well. It’s a very nice codec and it is in fact superior to h.264. Just thought i’d clear that up. Cheers!
-Aaron Crayford
VP8 is at the mercy of apple and microsoft, i am sure that chrome,firefox and opera will come to an agreement, but regarding the other two … things are misty
And again, flash lives and will continue to live because of it’s RIA features, its unique development model and great authoring tools.
It is much more needed to put flash down, not only video, but also audio, animation, vector graphics, pixel manipulation api, web services(remoting/rtmp/wsdl), real time messaging protocol… maybe in 15-20 years when probably most of the big browser creators will decide on a common ground.
I think it is much more possible for Adobe to open source the flash player than it is for microsoft,apple,google,mozilla and opera to adhere to the same standards, have the same features fully and “correctly” implemented and features that he flash player has and that the others don’t have.
Everybody hates flash today for three reasons:
1) Too many low quality intrusive ads.
2) Being closed source(yes, what if Adobe, from tomorrow decides that he needs to update the flash player and by implementing complete non backward compatible reader … this will mean that they will put a road block of web accessibility, the very purpose of the web)
3) Security issues(and again, because it is a closed source program, adobe might some times decide not to fix a critical bug or they decide not to report it correctly)
I used on2 vp7 extensively at my last company until switching to H.264. On2 tried to sell us on vp8, and provided us with a beta encoder/decoder SDK. All the PSNR and SSIM tests we ran found H.264 (x264) to be superior for our use case. Also, interestingly, they wanted pretty steep licensing costs compared to what we were payimg for vp7. The reason they gave us for the higher costs? They had to pay royalties to the MPEG-LA.
I won’t take this serious until Google dog foods VP8 into YouTube.
Why else will people use this codec if Google itself won’t!
This codec may be the edge that Google needs in video. The company may not have the complete set of puzzle pieces yet. (See http://www.theseed2020.com/gbt/ and the free discussion of Catch Media.) Google could emerge as a plumbing, back office, and support services player.
Ben Kent, SSNBlog.com
I’m pretty sure Adobe would use VP8 as well in a next version of Flash, especially if it is free and open-source. Sure for plain embedded video streaming, HTML5 video could take over some bits, especially as Youtube could be switched over to it. Though, who can say HTML5 video will have all the same customized player buttons, overlay ads, overlay annotations, overlay subtitles and all those other types of interactive features?
Folks,
this sentence is very scary:
“On2 tried to sell us on vp8 … they wanted pretty steep licensing costs compared to what we were payimg for vp7. The reason they gave us for the higher costs? They had to pay royalties to the MPEG-LA.”
It’s another case where the fact of being open-source it does not mean it requires intellectual property licensing from a patent owner.
If On2 owns all the IP they can release it “for free” to everyone… but if they have to sublicense something from MPEG-LA it means they DO NOT own all the IP requested.
In other words to use VP8 would be a suicide for anyone as you risk to build a product, deploy a service and then have a loyer knocking at your door asking for the royalties.
mhhh…
@Cyril: No, I mean “naive”. I.e., there are too many people who view Flash Player as software that does nothing but play back video, hence the apples/oranges comparisons to purpose-built, dedicated media players.
@Aaron: Thanks very much for confirming that VP8 is used outside of On2’s walls. Is that this company?
http://www1.vusion.com/index.html
AFAICT, Vusion offers an alternate plugin/media player called FullVu that incorporates VP8?
Aaron: I did some digging on FullVu HD. It seems to only support VP7. I don’t see any reference to VP8. But I might not be looking hard enough.
VP8 is a very cool codec, though a lot of the magic happens with post processing filters. Cool for low power stuff because it can omit this step, but sure there is a ‘analog vs. digital’ style argument to follow with people saying h.264 just looks better or whatever.
Hope to see Matroska + VP8 + Vorbis pretty soon, good for the whole web.
@George: What is your basis for claiming that VP8 is a very cool codec? Have you been able to test the production encoder and/or decoder? I’m very curious to know if VP8 has actually seen any use outside of On2’s walls yet.
Using out-of-loop postprocessing filters to cover up is a bad bad idea, it doesn’t help actual preservation and the only thing you get out of that is a blurry mess (see: WMV9 über postprocessing). It’s admitting “we look like crap and need to apply butter over the video to try to hide the artifacts”. (To On2’s credit, one idea they used in the past is apply an in-loop filter to motion-predicted data, giving some of the benefit.)
Combined with the fact that the only optimization criteria On2 has used for years has been PSNR, I’d expect a a very washed image (and that’s something you can already see in the released promo materials).
It’ll sure be a nicer base format than Theora, but only useful if it’s really patent-free (good luck with that!) and somebody bothers to write a decent encoder for it (difficult without money as it won’t ever be as good as h.264 -> less motivation from devs).
In any case if it lowers h.264 patent fees it’s good news.
Pingback: cnbeta啊。。。
Pingback: The Gospel according to Jobs: Thoughts on Flash | Thushan Fernando Uncut
Of course the hype is idiotic. However, I will say that VP7 compared quite favorably with H.264 and x264 in particular. In fact was the first codec to improve quality notably over VP3, IMHO. So, I don’t doubt VP8 may very well slightly surpass the best H.264 encoders, ALTHOUGH like all other On2 software, I’m sure it does so by using an order of magnitude more CPU time…
You have to make a fair comparison:
First, all H.264 material on the web (as in HTML5), and all material viewable on mobile phones (with “hardware” support, whatever that means) is Baseline H.264. VP8 is better than that by most measures. Even the x264 people agree.
Second, the latest Theora 1.2 encoder has lots of demo material at their website. They are very impressive considering how Theora started out. It’s getting closer and closer to H.264 quality. Theora is based on VP3. Think what a modern encoder optimized for psychovisual quality can achieve with VP8!
The use case for patent free video on the web is enormous. Online video would explode.
So let’s focus on the fair criticism: That the patent situation is a bit opaque, to say the least. Google has not made the patent search they undertook public. And the x264 devs are right it’s a bit too close to H.264 considering what happened to VC-1.
“First, all H.264 material on the web (as in HTML5), and all material viewable on mobile phones (with “hardware†support, whatever that means) is Baseline H.264. VP8 is better than that by most measures. Even the x264 people agree.”
QFT.
VP8, even in its current unoptimized state, beats the heck out of Baseline H.264, which is used by YouTube and all other video-enabled devices. Dark_Shikari’s FUD-filled article couldn’t change this fact (I’m surprised how he actually makes things sound worse than they actually are). Latest Theora development report (demo9) shows significant improvements due to psychovisual optimizations, and that’s already in the roadmap of VP8 developers.