VP8 vs. h.264

[Svartalf]

And what about mobiles? Those need H.264 and even have acceleration support for it (meaning longer battery life.) That means no VP8. And that means Flash + H.264 and raw H.264 is still the best answer.

Yes, you might not like it. But it's just the way it is

Here's a hint RealNC (repeated again...)...

Mobile devices, the ones you're actually referring to, do not have h.264 specific hardware at all. They use a DSP, like the DaVinci from TI, to do the work. My Droid (and actually any OMAP2/3/4 device you care to mention that's used for media playback...) only needs a VP8/WebM codec blob for the DSP to play back "hardware accelerated" playback. Nothing special. Same goes for pretty much any of the Snapdragon based or Armada based ARM mobile devices.

The line you and others keep trotting out in this regard is not valid and repeating it over and over again like this makes you look foolish. Seriously.

[BlackStar]

Heh... The problem with this would be that you'd have to frame in a bit of Javascript to detect whether or not they had HTML5 available or Flash and then select the right blob for the video player (There's a few actionscript embellishments that comprise a Flash player on a web-page, typically...) Not too much of a problem, but we've seen the care unto which many "webmasters" use to make a site- I'm sure at least some'll be too lazy to do it "right".

Truth is, webmasters are 95% likely to copy YouTube's "embed video" code snippet and be done with it. The rest might actually have to read an online tutorial or something - no, teh terror!

[Remco]

Truth is, webmasters are 95% likely to copy YouTube's "embed video" code snippet and be done with it. The rest might actually have to read an online tutorial or something - no, teh terror!

Speaking of YouTube's embed code, I wish they'd get the iframe method out of beta. Every day of object/embed soup is an extra day somewhen in the distant future that Flash will have to be supported to avoid breaking the web.

[smitty3268]

Heh... The problem with this would be that you'd have to frame in a bit of Javascript to detect whether or not they had HTML5 available or Flash and then select the right blob for the video player

Javascript-free version that automatically falls back to whatever a browser supports: http://camendesign.com/code/video_for_everybody

Test page + table listing browser compatibility: http://camendesign.com/code/video_fo...body/test.html

[Veerappan]

It's been said before, but I'll reiterate:

While VP8 when it was released was behind x264 in speed/quality, progress is being made. The encoder/decoders are getting faster, and work is being put into enhancing the quality of the encoded videos. It may never actually surpass x264, but hopefully it'll get close.

One thing about vp8 that seems to often get ignored is that the VP8 decoder was also written to be easy to decode and avoids floating point computation when possible. grepping the libvpx source for float gives me a total of 11 lines with float declarations/computations.

The performance may not be there yet, but it'll get there. If you check the webm mailing list, there's at least 1 guy who's working on a project to increase the level of multithreaded speedup on the CPU, and I'm currently working on an OpenCL implementation of the decoder (which might bleed over into the encoder in some common parts).

There's already plenty of ARM/x86 asm code in the project (w/ C fallbacks of course), but to believe that VP8's encoder/decoder won't improve would be silly.

[popper]

Here's a hint RealNC (repeated again...)...

Mobile devices, the ones you're actually referring to, do not have h.264 specific hardware at all. They use a DSP, like the DaVinci from TI, to do the work.

My Droid (and actually any OMAP2/3/4 device you care to mention that's used for media playback...) only needs a VP8/WebM codec blob for the DSP to play back "hardware accelerated" playback. Nothing special. Same goes for pretty much any of the Snapdragon based or Armada based ARM mobile devices.

The line you and others keep trotting out in this regard is not valid and repeating it over and over again like this makes you look foolish. Seriously.

"ONLY" is a very ambiguous word here....
as you should be Fully aware of, being a Linux Games developer...

true it "only" NEEDS some DSP code To Be Written , and that's the key key here, up until recently, even AVC/H.264 capable High Profile Level 4.1 code blocks Did Not Exist from commercial HW vendors, you must know that ?

id also argue that currently Only OMAP4 DSP, NEON SIMD and perhaps that oddball sse extension (i forget the name right now)as used by some ARM vendors Are Capable of full decode at full generic frame rates and screen sizes of the near future are worth writing code for.

even the new ASIC/DSP that you find in Sandy Bridge will have problems with VP8 until later, AMD Do Not have anything worth talking about in the mobile space as regards full HW ASIC/DSP decoding VP8 come to that dont forget (closed UVD3 doesn't cut it).

the long/short of it, will You be the One to write the required ASIC/DSP code and place it under LGPL on https://github.com

the core x264/ffmpeg/ffvp8 devs are already writing their CPU code and people can use it right now, the same cant be said for the OMAP4 DSP developers to date never mind using the MKV/webm container.

[madbiologist]

true it "only" NEEDS some DSP code To Be Written , and that's the key key here, up until recently, even AVC/H.264 capable High Profile Level 4.1 code blocks Did Not Exist from commercial HW vendors, you must know that ?

<snip>

the core x264/ffmpeg/ffvp8 devs are already writing their CPU code and people can use it right now, the same cant be said for the OMAP4 DSP developers to date never mind using the MKV/webm container.

There is a demo of VP8 (seem to be in a WEBM container too) running on an OMAP4 at http://e2e.ti.com/videos/m/applicati...ic/240443.aspx and at http://blog.webmproject.org/2010/10/...ti-omap-4.html

This also received a bit of publicity - eg. http://www.electronicspecifier.com/T...ssor-Video.asp and http://www.netbooknews.com/12658/web...rocessor-demo/
I also found at least one twitter message about this at https://twitter.com/jamiebennett/status/27538421004

[popper]

well its interesting that you managed to find this single TI demonstration from Oct 16th 2010, around 2 and half months ago now.

so indeed at least someone on the internal TI OMAP™4 team have finally got around to actually writing the required OMAP™4 vp8 DSP code for this demonstration (assuming its accurate and they didn't PR innovate to much OC ) to try and show it at least functioning.

but more to the point, after having a quick look, TI Still dont seem to actually List or sell this OMAP™4 code block in their product lists as yet, in fact i still cant seem to find Any commercial Vendor that actually lists a Vp8 code block for actual sale or release any time soon.

[madbiologist]

For the sake of completeness, according to the Mobile Firefox developers "Firefox 4 beta 3 for Android and Maemo supports both WebM and Ogg Theora video." - see https://support.mozilla.com/en-US/qu...#answer-127019

Of course what we really care about is hardware acceleration of the above, so as to experience good frame rates and low power consumption. But of course hardware acceleration is not much good without some software to utilise it.

And as somebody said, VP8 and H264 are used for more than just web video.

[smitty3268]

but more to the point, after having a quick look, TI Still dont seem to actually List or sell this OMAP?4 code block in their product lists as yet, in fact i still cant seem to find Any commercial Vendor that actually lists a Vp8 code block for actual sale or release any time soon.

Give it a little time. VP8 was just released a few months ago, it's still beta even in Firefox, one of the leading proponents. Even h.264 wasn't supported in devices for quite some time after it was announced. 2011 should be when it all starts coming together, from browsers to youtube to Flash to hardware acceleration from devices. The device manufacturers have already pledged support so there is no reason to believe they are lying or planning not to do it.