H.265 or VP9? The fight is only beginning.
I really thought it was a done deal when Google announced their roadmap for WebRTC, where VP9 was there as a new experimental addition.
VP9 is the future of VP8. It is the next royalty free video codec proposed by Google and is in direct competition with the H.264 and H.265 line of codecs.
The way I saw it, is that while everyone is competing if the mandatory video codec should be VP8 or H.264, Google pulls out the VP9 rabbit and just sidesteps that debate altogether.
When it comes to video codecs in WebRTC, the forces that decide which ones are going to be used are the browser vendors.
- Chrome opts for VP8 and VP9 at the moment
- Firefox does VP8 and is adding H.264
- IE and Safari are leaning towards H.264, each with his own internal reasons
So what happens now, that iPhone 6 got released and its specs include H.265 support for FaceTime?
How will that affect a future debate? While Google is planning on adding VP9 support, Apple has millions of mobile devices in users hands already running H.265.
Google and VP9
Google is working on getting VP9 ready for prime time. They need to fit it in to multiple types of hardware, mostly with no hardware acceleration suitable for the task. This means that much of the effort takes place in software. A huge undertaking.
They already have VP9 decoding in YouTube for many of us, but that’s different than real time bidirectional video.
End of this year having it as an experimental option in WebRTC in the browser is probably a year ahead of any official release, and a bit more from mobile yet.
There’s also the daunting task of getting chipset manufacturers to add hardware acceleration suitable for VP9. An 18 month cycle at the very least.
Apple and H.265
Apple has a vested interest in the H.26x line of codecs. It is part of MPEG-LA, so it enjoys royalties and a barrier of entry for others in this space – for both H.264 and H.265.
Its current devices run H.264 in hardware, so adding H.265 was the obvious step in its A8 chipset for its new iPhone 6. Enabling it for FaceTime means better video quality for FaceTime than anything out there today (at the same bitrate), and also shows that H.265 is ready for mass production. It also means it can use higher video resolutions for FaceTime on its retina displays for the larger screen sizes it has on offer now.
Apple might have added VP9 hardware support in their A8 chipset, but if I were Apple, I wouldn’t say that until the debate on the video codec in WebRTC is over – why lose a strong story around H.265?
Where does that leave us?
The decision of the mandatory video codec in WebRTC is as far as ever. We might see a shift in gears from H.264/VP8 to H.265/VP9 debate, but other than that, I don’t see this one concluding anytime soon.
If Microsoft and Apple do get their act together and offer WebRTC in their browsers, I am afraid it won’t have video interoperability with Chrome in its initial releases.
Tsahi, you know very well that both VP8 and H.264 are both free and royalty free… until you get sued for using them. Seriously, VP8 was suspected to be encumbered by patents and, indeed, there have already been claims made. VP9 will be no different. The companies who have worked on video coding for decades have a huge pile of IPR.
Quite frankly, I think the current licensing cost of H.265 is nuts, but I suspect the folks participating in the MPEG-LA licensing pool for that will lower the cost eventually. If they do not, people will go use other technologies. H.264 went through similar licensing frustration when it was introduced and I expect a repeat of history with H.265 gaining wide adoption.
Paul,
Nice to see you here again 🙂
The fight here isn’t between who has patents threatening its well being, but rather on the concept of a royalty free video codec. The difference in stance and the statement that video coding technologies should be commoditized is what at stake here.
Yeah, I think everyone would like a royalty-free codec that is in the business of making any kind of video product. And, there are a few initiatives to try to make that happen. However, I’m very doubtful that anyone will actually produce something really good that isn’t stepping on somebody’s patent.
So considering reality :-), what I’d really like to see is some rational licensing fees. H.265 royalties are just too much right now. H.264 is good and Cisco, by offering a binary library, is covering the cost for just about anyone that wants to use it, thus making it totally free.
Perhaps we need to do the same for H.265 and/or put pressure on the license holders to reduce costs.
I get a little wound up when I hear the suggestion that VP8 or VP9 or whatever else is “free” when I know there is IPR on them, held by some of the same companies that have IPR on H.264 and H.265. And using any of the others totally puts companies at risk of being sued.
Paul,
You left me unconvinced on this one.
Same people who can sue on VP8 can sue on H.264. MPEG-LA patents was paid for on VP8 by Google already. The licensing around it is far more permissive than what Cisco’s openH264 binary can ever offer to those running on their own devices or when you try to fit it into an iOS device.
Today, when every device out there supports video, and most services are free, the only barrier left for many use cases is a royalty bearing video codec .
I’d like to remind you that image compression was once patented, with GIF causing headaches. Most patents have expired. The rest was worked around. We since then moved on to PNG format and others are pursuing new territories (WebP anyone?). We no long think of image compression as something requiring royalty payments. Voice is headed that route as well with SILK, Speex and now Opus. It is time we start making the shift towards video with this mindset.
Yeah, I did say everyone would like that. But, we’re not quite there yet. Honestly, I suspect it will be another 10 years.
But, I don’t think VP8 licensing is more permissive. Not every IPR holder is at the table with VP8. The same can be said of H.264, admittedly, but VP8 exposes one to greater unknowns. At least the H.264 patent holders are known.
HTML has support for digital rights management (DRM, restricting how content can be used) via the HTML5 Encrypted Media Extensions (EME). The addition of DRM is controversial because it allows restricting users’ freedom to use media restricted by DRM, even where fair use gives users the legal right to do so.
Well, I think Google wont enable VP9 encoding in Chrome WebRTC until it would be adequately quick. For now VP9 is 4 times slower than VP8 (as far as I remember our last tests), which makes it not worth using. Speeding up VP9 theoretically could be done by doing much finetuning for Intel processors, with much effort, maybe with Intel support – H264 works good on x86 partially because Intel invested much effort in IPP acceleration of H264 for Core i*. Or we can just wait while processor would become 4 times faster or have 4 times more cores.
But I guess that Google is mainly investing in VP9 encoding IP core, it is not released yet, so this 18 month hardware release cycle is not even started.
Also I want to note that we can’t be sure that Apple’s H265 implementation is any better than good VP8/H264 implementation before we have some independent tests (it could be even worse). I’ve seen some examples of early H264 implementations performing worse than H263+ and hardware implementations are always simplified and so have less quality. Actually for quality improvement for hardware encoders it still makes sense in investing in better H264 implementation (higher profile, more features, better search), at least to match software opensource x264 encoding – but this does not look good for marketing reasons.
It’s interesting how Facetime’s calls between MacOS and iPhone 6 are handled – I guess they are H264 still in both sides or in direction MacOS->iPhone 6.
Stass,
All true. Thanks for raising these issues up. My only doubt here is where in the 18 months hardware cycle is Google today – there were announcements in CES at the beginning of this year for VP9 decoders in TVs if I recall correctly, so the cycle might have already begun – I just don’t know.
iPhone 6 must have H.264 support as well, otherwise, it won’t be able to FaceTime with an iPhone 4 or 5.