On2 VP8

the Mad Duke · 13th September 2008, 16:04

Did anybody get to beta test it? I don't see a personal use version on their site. I hope they release one.
http://www.on2.com/docs/graf/splash_on2.html

nm · 13th September 2008, 18:00

Looking at the sample clip, it's interesting how they got x264 output to look like MPEG-4 (A)SP with all that macroblocking and somewhat annoying keyframe pumping. Would be nice to see the settings used. VP8 output on the other hand looks pretty much like x264 a year or two ago (without AQ and Psy RDO).

Atak_Snajpera · 13th September 2008, 18:08

"...VP8 50% better than h.264" Hahahaahahah Good joke. Besides PSyRDO shows that in fact PSNR value is USELESS!!!!

nm · 13th September 2008, 18:36

Quote:

Originally Posted by Atak_Snajpera

"...VP8 50% better than h.264" Hahahaahahah Good joke.

Note that they only claimed up to 50% less data / fewer CPU cycles, which could also be "25% worse most of the time" in real world scenarios

e-Pawel · 13th September 2008, 18:37

Quote:

Originally Posted by Atak_Snajpera

"50% better than h.264"

That's what already VP7 was claimed to be... yeah in deed, good joke

Dark Shikari · 13th September 2008, 20:34

Indeed, a good joke. VP7 was slower than H.264 and significantly worse quality than VC-1 in most of my tests (let alone H.264), so VP8 will likely be equally a joke.

IgorC · 14th September 2008, 02:21

3 years ago On2's devs made everything possible for that VP7 could compete with x264 (revisions 2xx). They constantly were updating their codec from VP7.0.0 to VP7.0.9.
I hope On2's devs will try to do the same for vp8. Of course now VP7 has no chance to be comparable to x264. Since then there was only misc. update vp7.0.10.

I think the most important problems that VPx codecs have unstable local quality while overall quality isn't bad. VP7 has a problem with scene change. Key frames are highly quantized and rest of sequences recover details in temporal sequence. As a results there is no constant quality. Another issue is too strong deblocking filter. It's good only for very low bitrate.

To compete with x264 they must plan long time development to achieve solid and well tuned codec.

Just my opinion.

And good luck for a small company On2. They aren't MS to spread millions for VC-1 and they must face big standard.

Sagittaire · 14th September 2008, 05:41

Quote:

Originally Posted by Dark Shikari

Indeed, a good joke. VP7 was slower than H.264 and significantly worse quality than VC-1 in most of my tests (let alone H.264), so VP8 will likely be equally a joke.

I don't think. VP7 was excellent codec.

Dark Shikari · 14th September 2008, 07:26

Quote:

Originally Posted by Sagittaire

I don't think. VP7 was excellent codec.

By excellent you mean it decoded slower than ffmpeg's H.264 decoding did... using a 2004 build of ffmpeg?

Tagert · 14th September 2008, 10:28

Big claims indeed

Wonder when they will be able to back it up by releasing their awesome codec

Manao · 14th September 2008, 11:01

Bond : golden frames are indeed like references in h264, but limited to 2 references max : the previous reference, and another one, that can be updated whenever they want. It seems it can be updated piecewise too, which makes it a bit more flexible than 2 references with h264. As for the usefulness of that, I don't know. It must help in some contrive cases, but beyond that... In any case, it will help less than bframes & bpyramids would, but they can't use those because they are afraid of patents (though bframes should be patentless by now, I think).

Their description of their loop filter and its adaptivity to motion doesn't impress me. h264 loop-filter also adapts to motion (though the way it adapts to motion seems different from VP8's).

Their visual results for h264 seems a bit strange, since I seem to see a good old IDCT mismatch drift during the GOP, which can't happen with AVC. As stated by nm, it more looks like ASP than AVC.

As for the PSNR curves, they must always be taken with a (huge) grain of salt, wherever they come from. Since they don't say which encoder and which settings are used, however, the curve most probably are correct. They just don't mean anything and can be ignored.

Dark Shikari : sagittaire is right, when it came out, VP7 was lighter than existing AVC decoders, and more than hold its ground quality-wise. Our encoder, at that time, only managed to consistently beat it when we included High profile.

CruNcher · 14th September 2008, 19:52

Yeah about the PSNR results they dont say alot tough for H.264 basicly with inloop deblocking it means higher PSNR and most times looks awfull tough with VP8 it doesn't have to be the same because they use a complete different deblocking aproach which seems really interesting

, though 50% is a heavy claim thats like they wanted to say: "hey look we drove ahead of MPEG before they even get started with H.265"

IgorC · 15th September 2008, 23:35

Wonder if they will go to VFW?

Sharktooth · 16th September 2008, 02:06

i wonder if they will release a free-for-personal-use encoder.
then speculation will end...

LoRd_MuldeR · 16th September 2008, 02:28

Quote:

Originally Posted by Sharktooth

i wonder if they will release a free-for-personal-use encoder.
then speculation will end...

Plus a free-for-personal-use decoder library, of course

CruNcher · 16th September 2008, 08:58

Maybe this also implies they gonna Open Source VP7 now

slavickas · 16th September 2008, 09:27

Quote:

Originally Posted by CruNcher

Maybe this also implies they gonna Open Source VP7 now

well they promised to release VP4 as open source and it never happened ( big IINM

)

Inventive Software · 16th September 2008, 09:58

VP4's decodable in ffmpeg, yes no?

As is VP5, yes no?

LoRd_MuldeR · 16th September 2008, 16:19

Quote:

Originally Posted by Inventive Software

VP4's decodable in ffmpeg, yes no?

As is VP5, yes no?

Most likely this is thanks to the reverse engineering capabilities of the ffmpeg developers, not thanks to On2...

smok3 · 16th September 2008, 14:43

i think they missed all the trains, at least i can't see what could happen, even if they would develop the world-best-codec.

13th September 2008, 16:04	#1 \| Link
the Mad Duke Registered User Join Date: Mar 2004 Posts: 41	On2 VP8 Did anybody get to beta test it? I don't see a personal use version on their site. I hope they release one. http://www.on2.com/docs/graf/splash_on2.html

13th September 2008, 18:08	#3 \| Link
Atak_Snajpera RipBot264 author Join Date: May 2006 Location: Poland Posts: 7,815	"...VP8 50% better than h.264" Hahahaahahah Good joke. Besides PSyRDO shows that in fact PSNR value is USELESS!!!! __________________ Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper

13th September 2008, 20:34	#6 \| Link
Dark Shikari x264 developer Join Date: Sep 2005 Posts: 8,666	Indeed, a good joke. VP7 was slower than H.264 and significantly worse quality than VC-1 in most of my tests (let alone H.264), so VP8 will likely be equally a joke. __________________ Follow x264 development progress \| akupenguin quotes \| x264 git status ffmpeg and x264-related consulting/coding contracts \| Doom10

14th September 2008, 02:21	#7 \| Link
IgorC Registered User Join Date: Apr 2004 Posts: 1,315	3 years ago On2's devs made everything possible for that VP7 could compete with x264 (revisions 2xx). They constantly were updating their codec from VP7.0.0 to VP7.0.9. I hope On2's devs will try to do the same for vp8. Of course now VP7 has no chance to be comparable to x264. Since then there was only misc. update vp7.0.10. I think the most important problems that VPx codecs have unstable local quality while overall quality isn't bad. VP7 has a problem with scene change. Key frames are highly quantized and rest of sequences recover details in temporal sequence. As a results there is no constant quality. Another issue is too strong deblocking filter. It's good only for very low bitrate. To compete with x264 they must plan long time development to achieve solid and well tuned codec. Just my opinion. And good luck for a small company On2. They aren't MS to spread millions for VC-1 and they must face big standard. Last edited by IgorC; 14th September 2008 at 02:24.

14th September 2008, 11:01	#11 \| Link
Manao Registered User Join Date: Jan 2002 Location: France Posts: 2,856	Bond : golden frames are indeed like references in h264, but limited to 2 references max : the previous reference, and another one, that can be updated whenever they want. It seems it can be updated piecewise too, which makes it a bit more flexible than 2 references with h264. As for the usefulness of that, I don't know. It must help in some contrive cases, but beyond that... In any case, it will help less than bframes & bpyramids would, but they can't use those because they are afraid of patents (though bframes should be patentless by now, I think). Their description of their loop filter and its adaptivity to motion doesn't impress me. h264 loop-filter also adapts to motion (though the way it adapts to motion seems different from VP8's). Their visual results for h264 seems a bit strange, since I seem to see a good old IDCT mismatch drift during the GOP, which can't happen with AVC. As stated by nm, it more looks like ASP than AVC. As for the PSNR curves, they must always be taken with a (huge) grain of salt, wherever they come from. Since they don't say which encoder and which settings are used, however, the curve most probably are correct. They just don't mean anything and can be ignored. Dark Shikari : sagittaire is right, when it came out, VP7 was lighter than existing AVC decoders, and more than hold its ground quality-wise. Our encoder, at that time, only managed to consistently beat it when we included High profile. __________________ Masktools x86 & x64: Stable (2.0a48) AVCMatrices : Stable (1.3) Anisotool : Beta (1.0a5)

13th September 2008, 18:00	#2 \| Link
nm Registered User Join Date: Mar 2005 Location: Finland Posts: 2,641	Looking at the sample clip, it's interesting how they got x264 output to look like MPEG-4 (A)SP with all that macroblocking and somewhat annoying keyframe pumping. Would be nice to see the settings used. VP8 output on the other hand looks pretty much like x264 a year or two ago (without AQ and Psy RDO).

14th September 2008, 10:28	#10 \| Link
Tagert Registered User Join Date: Sep 2006 Location: Karlskrona, Sweden Posts: 72	Big claims indeed Wonder when they will be able to back it up by releasing their awesome codec

14th September 2008, 19:52	#12 \| Link
CruNcher Registered User Join Date: Apr 2002 Location: Germany Posts: 4,926	Yeah about the PSNR results they dont say alot tough for H.264 basicly with inloop deblocking it means higher PSNR and most times looks awfull tough with VP8 it doesn't have to be the same because they use a complete different deblocking aproach which seems really interesting , though 50% is a heavy claim thats like they wanted to say: "hey look we drove ahead of MPEG before they even get started with H.265" __________________ all my compares are riddles so please try to decipher them yourselves :) It is about Time Join the Revolution NOW before it is to Late ! http://forum.doom9.org/showthread.php?t=168004 Last edited by CruNcher; 14th September 2008 at 20:10.

15th September 2008, 23:35	#13 \| Link
IgorC Registered User Join Date: Apr 2004 Posts: 1,315	Wonder if they will go to VFW?

16th September 2008, 02:06	#14 \| Link
Sharktooth Mr. Sandman Join Date: Sep 2003 Location: Haddonfield, IL Posts: 11,768	i wonder if they will release a free-for-personal-use encoder. then speculation will end... __________________ MPEG-4 ASP Custom Matrices: EQM V1(old), EQM AutoGK Sharpmatrix (aka EQM V2), EQM V3HR (updated 01/10/2004), EQM V3LR, EQM V3ULR (updated 04/02/2005), EQM V3UHR (updated 17/12/2004) and EQM V3EHR (updated 05/10/2004) Info about my ASP matrices. MPEG-4 AVC Custom Matrices: EQM AVC-HR Info about my AVC matrices My x264 builds. Mooo!!!

16th September 2008, 08:58	#16 \| Link
CruNcher Registered User Join Date: Apr 2002 Location: Germany Posts: 4,926	Maybe this also implies they gonna Open Source VP7 now __________________ all my compares are riddles so please try to decipher them yourselves :) It is about Time Join the Revolution NOW before it is to Late ! http://forum.doom9.org/showthread.php?t=168004

16th September 2008, 09:58	#18 \| Link
Inventive Software Turkey Machine Join Date: Jan 2005 Location: Lowestoft, UK (but visit lots of places with bribes [beer]) Posts: 1,953	VP4's decodable in ffmpeg, yes no? As is VP5, yes no? __________________ On Discworld it is clearly recognized that million-to-one chances happen 9 times out of 10. If the hero did not overcome huge odds, what would be the point? Terry Pratchett - The Science Of Discworld

16th September 2008, 14:43	#20 \| Link
smok3 brontosaurusrex Join Date: Oct 2001 Posts: 2,392	i think they missed all the trains, at least i can't see what could happen, even if they would develop the world-best-codec. __________________ certain other member