Google Opens Up VP8, Launches New Container Format

[deanjo]

Heh... Not to hear them tell it. More to the point, you don't know WHAT they're going to do- and won't until things are done by them...unless you're privy to their stuff and breaching NDAs right now...

MPEG-LA has said that it will re-evaluate come 2015 and that is on public record. They have not said either way what will for sure happen after that. Just as they did back in 2005.

[deanjo]

One place I do think where VP8 does have a future are areas where codecs such as realmedia, Bink and possibly divx are used. It is better at least then those solutions in terms of image quality.

[Svartalf]

One place I do think where VP8 does have a future are areas where codecs such as realmedia, Bink and possibly divx are used. It is better at least then those solutions in terms of image quality.

Heh... There we DO agree on things, deanjo. And I'm waiting for a bit of spare time to tinker with the stuff to see if something along those lines can be done without too much penalties on performance. If so, it's a full-on win over using Theora for the stuff (and Theora's a good fit there as well if there's not enough performance for one's liking. Using DivX means you're owing MPEG-LA royalties, and Bink's $5k per platform (i.e. Windows, MacOS, etc...) used.

[Xilanaz]

here we go already on the Apple and MPEG-LA issue: http://www.electronista.com/articles...l.to.fee.webm/

[drag]

That won't happen for quite a long time I'm afraid if at all even with no patent issues at present.

I don't think so at all, actually.

Linux has a lot of very powerful media tools and streaming stuff right now, but it's just that nobody can really use them or support them out of the box since the codec issues make life difficult.

The tools for dealing with VP8 are already ready to be shipped. Gstreamer, support, is already here and if you install the VP8 plugins then Linux WEbkit-based browsers like Epiphany and Midori already 'just work' with Google's stuff.

To improve it they would most likely have to start looking at techniques that are already patented and that closes that road down.

They have to do the same sort of stuff to VP8 that they did for Theora's newer versions and the big improvements they got there.

Like was mentioned before VP8 encoder was heavily optimized for high PSNR scores... which look good on paper and on graphs when your trying to sell people proprietary codecs.. it's not a good approach in practice. If they tweak the current stuff and make the encoder more suited and take some 'psy' optimizations into account then it will easily beat most of the commercial H.264 encoders out there.

They also have to contend with device manufacturers and not many will bring out a new product to support an inferior format to what they presently have now.

That's Apple's crack I smell around here?

VP8 already is optimized for the DSPs and SIMD devices found in mobile devices. There is nothing magical about hardware that means that to support H.264 you can't support accelerating VP8 and visa versa.

'ASICs' are not some magical thing that must never be touched and device makers don't design chips and in actuality have almost nothing to do with the the design and fabrication of the vast majority of the chips in their devices.

What actually happens is device makers pick out a reference platform from companies like Texas Instruments and then copy that design as much as possible to a formfactor that will fit in the shape they want.

Seriously. That's how it works.

The majority of big 'handset makers' and OEMs out of Asia actually have very little technical expertise when it comes to designing integrated circuits. They take a SoC from Ti or FreeScale, take a reference platform from one of them or third party company and then figure out how to manufacture it cheaply. They don't hack on the software if they can help it (and when they do it sucks unless they license from some software development firm). They specialize heavily in the manufacturing and industrial design and that is their bread and butter.. the less they actually have to touch the chip design or software design on the devices they ship the better off they are.

It's those SoC (System on a Chip) is were 'acceleration' comes from. It's just one big 'chip' that has the majority of the functionality you need to build a embedded system. This is done to keep prices low and keep things low-power.

Look at Google's hardware partners:

http://www.webmproject.org/about/supporters/

* AMD
* ARM
* Broadcom
* FreeScale
* Imagination
* Logitech
* Marvel
* Mips
* Nvidia
* Qualcomm
* Texas Instrument
* VeriSilicon
* ViewCast


Can you pick out the companies that make the SoC's that go into the majority of Smartphones and other 'multimedia' handhelds?

Well... It's not 'ARM'. They just design processors that people license to integrate into their SoC products.

It's Qualcomm, FreeScale, Broadcom, and Texas Instruments. Not companies like HTC or even Apple.

For it to truly be adopted in large masses it still has to provide a superior solution in capabilities to the status quo.

That should not be very difficult as the status quo for streaming video is pretty lousy. The fact that I can take Theora, as lousy at is is, and get better quality then the majority of H.264 or VC-1 video out there is a testament to that.

By that time it is very likely that the MPEG-LA group will have something out to eclipse h264 and once again the gap widens between the opensource alternatives. If VP8 debuted at the same time as h264 it might of had a fighting chance in mass adoption but the market is firmly entrenched in h264 already.

The vast majority of the market is 'Flash Video' and even Adobe is supporting VP8.

We saw this with mp3 and ogg years back. MP3 was already widely adopted and ogg remains a obscure item that few devices support.

Apple coming out and dominating the player market is what killed Ogg for consumer devices. Otherwise if you actually pay attention most half-way decent players support Ogg and even Flac actually. I have no problem playing Ogg Vorbis on the majority of 'Mp3 player' I have come across, except ipods and Sony stuff.

Apple could of easily supported Ogg if they wanted to. Hacked versions of the ipods have been able to play Ogg Vorbis just fine from the first generation up. But they wanted AAC instead.

And why do you think that Apple wanted AAC instead of Ogg?

Same reason they are compaigning against open codecs like Theora or VP8.

It's because Apple is a member of MPEG-LA. They make money from people licensing H.264 and AAC patents.

[Apopas]

Just as much uncertainty as patent infringement by VP8.

What patents does VP8 infring?

[deanjo]

I don't think so at all, actually.

Linux has a lot of very powerful media tools and streaming stuff right now, but it's just that nobody can really use them or support them out of the box since the codec issues make life difficult.

The tools for dealing with VP8 are already ready to be shipped. Gstreamer, support, is already here and if you install the VP8 plugins then Linux WEbkit-based browsers like Epiphany and Midori already 'just work' with Google's stuff.

So support in 1% of user space is a testament of success?

They have to do the same sort of stuff to VP8 that they did for Theora's newer versions and the big improvements they got there.

Theora's never versions have a hard time keeping up to even the aging mpeg 4 codecs such as xvid or divx let alone h264 in quality and adoption.

Like was mentioned before VP8 encoder was heavily optimized for high PSNR scores... which look good on paper and on graphs when your trying to sell people proprietary codecs.. it's not a good approach in practice. If they tweak the current stuff and make the encoder more suited and take some 'psy' optimizations into account then it will easily beat most of the commercial H.264 encoders out there.

That's a BIG if there. If it was as easy as you perceive it Google or On2 probably would have had already done it before it was even released publicly.


That's Apple's crack I smell around here?

VP8 already is optimized for the DSPs and SIMD devices found in mobile devices. There is nothing magical about hardware that means that to support H.264 you can't support accelerating VP8 and visa versa.

'ASICs' are not some magical thing that must never be touched and device makers don't design chips and in actuality have almost nothing to do with the the design and fabrication of the vast majority of the chips in their devices.

What actually happens is device makers pick out a reference platform from companies like Texas Instruments and then copy that design as much as possible to a formfactor that will fit in the shape they want.

Seriously. That's how it works.

The majority of big 'handset makers' and OEMs out of Asia actually have very little technical expertise when it comes to designing integrated circuits. They take a SoC from Ti or FreeScale, take a reference platform from one of them or third party company and then figure out how to manufacture it cheaply. They don't hack on the software if they can help it (and when they do it sucks unless they license from some software development firm). They specialize heavily in the manufacturing and industrial design and that is their bread and butter.. the less they actually have to touch the chip design or software design on the devices they ship the better off they are.

It's those SoC (System on a Chip) is were 'acceleration' comes from. It's just one big 'chip' that has the majority of the functionality you need to build a embedded system. This is done to keep prices low and keep things low-power.

Look at Google's hardware partners:

http://www.webmproject.org/about/supporters/

* AMD
* ARM
* Broadcom
* FreeScale
* Imagination
* Logitech
* Marvel
* Mips
* Nvidia
* Qualcomm
* Texas Instrument
* VeriSilicon
* ViewCast


Can you pick out the companies that make the SoC's that go into the majority of Smartphones and other 'multimedia' handhelds?

Well... It's not 'ARM'. They just design processors that people license to integrate into their SoC products.

It's Qualcomm, FreeScale, Broadcom, and Texas Instruments. Not companies like HTC or even Apple.

I'm very aware of "How it works". I'm also aware of many formats that DSP's that are capable of handling and even officially supporting it still has very little to do with adoption. The end implementation however is upon the end device manufacturer. I am aware of the companies involved. They are the same players as usual, nothing out of the ordinary there. And as in the past they have supported successfully adopted formats and they have also supported dud formats.

That should not be very difficult as the status quo for streaming video is pretty lousy. The fact that I can take Theora, as lousy at is is, and get better quality then the majority of H.264 or VC-1 video out there is a testament to that.

That would be a first given a level playing field where equal bitrate and file size come into play.

The vast majority of the market is 'Flash Video' and even Adobe is supporting VP8.

So? Adobe supported VP3 as well which served well until a superior solution came out.

Apple coming out and dominating the player market is what killed Ogg for consumer devices. Otherwise if you actually pay attention most half-way decent players support Ogg and even Flac actually. I have no problem playing Ogg Vorbis on the majority of 'Mp3 player' I have come across, except ipods and Sony stuff.

Apple could of easily supported Ogg if they wanted to. Hacked versions of the ipods have been able to play Ogg Vorbis just fine from the first generation up. But they wanted AAC instead.

And why do you think that Apple wanted AAC instead of Ogg?

Same reason they are compaigning against open codecs like Theora or VP8.

It's because Apple is a member of MPEG-LA. They make money from people licensing H.264 and AAC patents.

Yet despite all those ogg capable players, the mp3 still rules the audio world (even on iPods where AAC is the default format). Be sure to bookmark this thread to revisit in a year or two. The best thing about VP8 is that as long as it is seen as competition to h264 it will keep h264 in check by adding pressure to the MPEG-LA to keep streaming web use free for non profit use much like the free browsers forced companies to make their web browsers free as well.

[deanjo]

What patents does VP8 infring?

Nobody knows yet, that's the thing. With h264 having over 1000 patents on it alone (that's not including patents from previous codec patents) and patent law being as gray as it is with the similarities between it and h264 there is a strong chance that it does infringe on some patents. They probably would have been safer off looking at a wavelet based codec implementation where there is enough differential and prior art that a strong defense could be made.

[pingufunkybeat]

So support in 1% of user space is a testament of success?

Firefox + Chrome = 1% of user space?

It's not 1999 anymore.

[Loris]

Firefox + Chrome = 1% of user space?

It's not 1999 anymore.

And Opera too!
Don't forget Opera ...even if I never use it.