On2 VP7

[SpaceV]

Hi ON2 Guys,

I just read VP7 is demoed in Asia, when can we play with it?

Also the new Ninelanes Settop box webpage is showcasing VP4/5!!

Why is that, isnt that bad for ON2 that they demo 2-3 year old codecs??? I looked at their demo VP5 streams, nothing to be pround of
in my optinion. You guys should tell em to demo your best, the settop box is able to decode VP6 !

[virus]

I did a Google search and learned that there's a glycoprotein called VP7 :D

For the rest, the only real hint can be found in the On2's Investor Presentation... "actively developing VP7 for 2004 release". I don't expect the devs to give away any detail soon, so I guess we just have to wait.
My humble opinion is that VP7 would be required to match VP6.2 quality at 10-15% less bitrate at least, in order to keep up with the crowd of upcoming H.264 implementations (complete and well-tuned ones, not just buggy demos ;)). Whether this is actually possible remains to be seen.

cheers :)
virus

[SpaceV]

I expect it to be at least 20% if not 25-35% better than VP6.2

If not, it does not deserve the name VP7 but should be VP6.3.

Wonder what else is new....

Please any ON2 guy give us some scoop, we wont tell anyone ;)

[the Mad Duke]

Originally posted by SpaceV
I expect it to be at least 20% if not 25-35% better than VP6.2

If not, it does not deserve the name VP7 but should be VP6.3.

Wonder what else is new....

Please any ON2 guy give us some scoop, we wont tell anyone ;)

From what I've heard and read it is at least a 30% "improvement" over VP6.2. It is currently being shown to select companies in China. JMO but I wouldn't expect a personal use VP7 any time soon.

[virus]

Originally posted by the Mad Duke
From what I've heard and read it is at least a 30% "improvement" over VP6.2.
quoting from on2.com [about VP6.2]:
"Achieves same quality as VP6.1 at up to 50% lower data rates"... well, imho this is ridiculous BS. Let's not listen to their marketing department anymore, please ;)

IMHO a perfectly-tuned H.264 codec can bring the same quality as VP6.2 with maybe 15-20% less bitrate, maybe even 25%. But for sure not more. I don't believe that a tool capable of shrinking a VP6.2 encode down of an "at least 30%" amount with the same quality will be available even in 2008... let's imagine in 2004...

virus

[aketon]

I hope that they are going to add b-frames in VP7! B-frames are going to give a real push to their codec! Have you ever tested VP6.2 with any mpeg-4 codec with b-frames disabled??? You will see some very interesting results!:D VP6.2 is by far in this cases the best codec!;)

[SpaceV]

I really hope they release it to the public as they did with VP6
and why shouldnt they?? There was no harm done by releasing VP6.

I think they even gained from feedback, especially from sites like Doom9.

They want the codec to be used, they need a big base and attract some
attention so people see there is more than DivX and WMV etc...

VP6 will be used for the new Sims2, hope it reaches a big crowd that way.

[Bulletproof]

I liked VP6, I hope VP7 will be competitive against H264 that would be great, the more competition, the better. Maybe ON2Tech can shed some more light on VP7?

[On2Tech]

We're pleased that everyone's taken notice of our recent developments. Right now unfortunately all that we can offer is word that there will be additional information on our web site in the relatively near future about VP7. You can imagine how excited we are about the new version of our codec, and I'd love to share more information directly with you, but being a public company, we have to be sensitive about the manner in which information like this is disclosed.

Thanks,

On2Tech

[Sirber]

If you want I can include it to RealAnime ;)

[aketon]

Is there going to be a linux version of VP7???:D

[aketon]

http://biz.yahoo.com/prnews/040913/nym113_1.html

"We believe that the VP7 codec will be proven to be by far the best codec in the world"


I hope so!:D

[JohnV]

Waiting for some proofs. A closed presentation showing some businessmen how good it is compared to the competitors and the normal marketing hype doesn't exactly assure me. Last year until now MS has hyped that WM9/VC9/VC1 (what ever you want to call it) is better than h.264 in different exhibitions etc. I wonder how long they are going to continue now that higher quality implementations start to appear.. ;)
Not even VP7 streams/decoder is publicly available, so makes me wonder whether it's ready yet for public dissection. ;)
Well, marketing hype is one thing, but for their sake they better walk the walk also. ;)

[the Mad Duke]

Originally posted by JohnV
Waiting for some proofs. A closed presentation showing some businessmen how good it is compared to the competitors and the normal marketing hype doesn't exactly assure me. Last year until now MS has hyped that WM9/VC9/VC1 (what ever you want to call it) is better than h.264 in different exhibitions etc. I wonder how long they are going to continue now that higher quality implementations start to appear.. ;)
Not even VP7 streams/decoder is publicly available, so makes me wonder whether it's ready yet for public dissection. ;)
Well, marketing hype is one thing, but for their sake they better walk the walk also. ;)

Keep in mind that On2 is a very small company that is publicly traded on the AMEX stock exchange. They compete against an alleged monopoly like M$FT and a huge conglomeration like MPEG LA. It is to their best interest and to the share holders of the company to show their new improved codec to potential clients before releasing it to the public. I have followed the Duck Corporation almost since it's inception and personally would not doubt their claims to the performance of VP7.

[Doom9]

well, they still have my email address, so if they want to have their best offering in the next codec comparison.... (but don't get all excited about that now, right now I don't even have a PC at home)

[stephanV]

If its anything better than VP6.2 (which would be logical to assume), it would already be a great codec. :)

Ah well, patience is a virtue :)

[Tommy Carrot]

Originally posted by the Mad Duke
I have followed the Duck Corporation almost since it's inception and personally would not doubt their claims to the performance of VP7.

I don't know, their marketing department likes to exaggerate a bit, to say the least. Probably some of you can remember their infamous statements about VP3 outperforming mpeg4, VP4 delivering the same quality than mpeg4 at half bitrate, VP6 outperforming H.264 (it's easy to say this when there is no decent implementation out there), etc.

So i take their statements with a grain of salt. I certainly hope it will be true, but we'll see.

[IgorC]

Dont forget. Vp 6.2 is the best video codec untill H.264 will be realese

[Sirber]

Originally posted by IgorC
Dont forget. Vp 6.2 is the best video codec untill H.264 will be realese What is that comment? What if he likes more XviD quality? Maybe it's the best codec for you but it's certainly not for me :o

[the Mad Duke]

Originally posted by Doom9
well, they still have my email address, so if they want to have their best offering in the next codec comparison.... (but don't get all excited about that now, right now I don't even have a PC at home)

Perhaps the fellow doing the test for PC Magazine on the Best Quality thread might want to contact On2 to see if they will let him compare VP7. It can't hurt to ask. The worst they can so is no.

[JohnV]

Originally posted by the Mad Duke
I have followed the Duck Corporation almost since it's inception and personally would not doubt their claims to the performance of VP7. Hehheh, show me one marketing department of any company in this area of business that doesn't exaggerate at all..

On2 marketing has exaggerated before quite a lot, like Tommy Carrot pointed out (so does practically all companies to some extend so that's nothing special). Anybody who has followed this business should know that what marketing departments say should be taken with grain of salt.

I'm suspicious, because On2 doesn't show even demo streams comparing VP 6.2 and 7. Obviously they are trying to get deals for example in China as quickly as possible. Maybe public presentation doesn't fit to their strategy for some reason..

[Sagittaire]

Originally posted by JohnV
Hehheh, show me one marketing department of any company in this area of business that doesn't exaggerate at all..


another example ... lol

Originally posted by Gej
We have tested several clips produced by the studios and mapped PSNR against bitrate for both codecs. Results to be posted on DivX Labs soon. And we never said x2 more powerfull but 30% smaller bitrate at same PSNR than WM9.

perhabs with VP6 or H264 Ateme but certainely not with MPEG4 ASP and DivX ... :devil:

[virus]

a small note, just to keep the gossips going... :D

If you browse On2's clients and partners list (http://www.on2.com/clients.php3) you'll find a company named RealNetworks. I guess some of you already know them :)

Now, what is the relationship between On2 and RN? Some sort of anti-MPEG alliance? or maybe anti-M$ tactics? or...?

:devil:
virus

[slavickas]

Originally posted by virus
a small note, just to keep the gossips going... :D

If you browse On2's clients and partners list (http://www.on2.com/clients.php3) you'll find a company named RealNetworks. I guess some of you already know them :)

Now, what is the relationship between On2 and RN? Some sort of anti-MPEG alliance? or maybe anti-M$ tactics? or...?

:devil:
virus

vp4 plugin for realplayer?

btw there is more interesting threads than repeating over over "don't believe in marketing hype". i just know vp6 is good codec, vp7 should be better DOT

[JohnV]

Well, On2 is a pretty small company, and although I'm a MPEG-4 supporter, anything that is away from Microsoft is a good thing imo. :D
The last thing I want is a Microsoft media monopoly and MS control of media usage. I certainly don't want to be "forced" to use some average quality MS controlled media in the future.

Though companies like DivxNetworks isn't helping MPEG-4 much either since all it seems to care about is its own implementation of "DivX". MPEG-4 compatibility concept is totally secondary to DXN.

Hopefully VP7 is good enough codec that it drives the competition and development of all codecs further.

[the Mad Duke]

Originally posted by virus
a small note, just to keep the gossips going... :D

If you browse On2's clients and partners list (http://www.on2.com/clients.php3) you'll find a company named RealNetworks. I guess some of you already know them :)

Now, what is the relationship between On2 and RN? Some sort of anti-MPEG alliance? or maybe anti-M$ tactics? or...?

:devil:
virus

I think the relationship between Real and On2 has been dissolved for some time now. Real licensed On2's VP4 & 5 codec gaining certain exclusive rights of use. This is just my conspiracy theory but I think Real kept On2's hands tied with this contract and had no intent to use the VP codecs while they actively developed Real Player 9 with intent to market it to companies that On2 would have gone after had they not given Real exclusive rights. The following is a long read from On2 that gives a little bit of detail.:

In July 2001 the Company entered into a Software Development and License
Agreement with Real Networks, Inc. Under terms of the agreement, we agreed to
develop a software plug-in to RealPlayer that would allow consumers to play
files encoded in our VP4 and VP5 compression format using the RealPlayer. In
conjunction with the development of the codec software, we also agreed to
develop a software plug-in to the RealSystem Server that would enable
broadcasters to stream VP4 and VP5 encoded files and a software plug-in to the
RealProducer that would enable users of the RealProducer to encode content using
VP4 and VP5. In order for us to develop the plug-ins described in the
RealNetworks agreement, RealNetworks was required to provide us with certain
development software as well as fulfill certain other obligations. RealNetworks
failed to deliver such software and perform such other obligations. After
honoring the contract from its inception and attempting to resolve the situation
with RealNetworks for over 9 months through business discussions, in May 2002 we
notified RealNetworks that we were terminating the agreement. RealNetworks
responded and disputed our right to terminate the agreement and further claimed
that we were in breach of the agreement. We responded by reiterating our
positions. We have not had any contacts or discussions with RealNetworks since
delivering this response. As a result, we are not generating any revenues from
the sale of RealNetworks related products.

The Company believes that our May 2002 termination of the RealNetworks
agreement was effective. However, because the arrangement with RealNetworks was
exclusive in certain respects, should it hereafter be determined that such
termination was not effective, we could be restricted from supporting other
software platforms, including certain components of our own software platforms
and, although we do not believe that it would be the case, we might be found to
have breached the agreement with RealNetworks. Whether we might have liability
to RealNetworks depends on a variety of factors, including whether RealNetworks
met its obligations to us, whether it caused any breach that we might have
committed and whether it otherwise has liability to us in connection with the
agreement.

[virus]

Originally posted by the Mad Duke
I think the relationship between Real and On2 has been dissolved for some time now.
so, just a webmaster's fault? a logo that should have been removed long ago is still there? that's disappointing... I hoped for some more anti-M$ stuff ;)

[the Mad Duke]

Originally posted by virus
so, just a webmaster's fault? a logo that should have been removed long ago is still there? that's disappointing... I hoped for some more anti-M$ stuff ;)

I think both companies are anti M$FT but just not together.:D

Okay On2, release that VP7!

[Atamido]

Originally posted by IgorC
Dont forget. Vp 6.2 is the best video codec... Ever tried to play VP6 video on anything other than a Windows system? Any codec that you want to call the 'best' should probably have at least a little cross platform compatibly.

[pernas]

Originally posted by Pamel
Ever tried to play VP6 video on anything other than a Windows system? Any codec that you want to call the 'best' should probably have at least a little cross platform compatibly.
VP6 works well with MPlayer on other platforms than M$ :D

[Sirber]

Any specs for VP7? Please, please! no 50% better than VP6 BS :devil: Thanks :D

Note: BS means BBQ Sauce ;)

[Sharktooth]

Im really interested in VP7 too.
IMHO VP7 will be on par if noth slightly better than current h.264 implementations.

[IgorC]

you can find some interesting information in the offcila site about VP7. Have a look :

Program Will Be Closed for Launch of VP7 NEW YORK - PRNewswire-FirstCall - Dec. 22
NEW YORK, Dec. 22 /PRNewswire-FirstCall/ -- On2 Technologies, Inc., The Duck Corporation , today announced that the number of downloads of its TrueMotion VP6 Personal Use codec has exceeded 550,000. The program will be closed to new licensees today in anticipation of the launch of the TrueMotion VP7 codec.

The company said that the successful download program would be critical to the marketing of VP7 and On2's new encoding tools. On2 encoding tools will target users of the company's TrueCast Server and will also include products for several of its decoder distribution products. The company believes that these products will be launched over the course of 2005.

"The use of VP6 and the information we have received about customer behavior should pave the way for a successful launch of VP7 and the launches of encoders for several of our large distribution partners that allow On2 to market the encoder," said Douglas A. McIntyre, On2 Chairman, President and Chief Executive Officer. "Every VP7 license will be a paid license, including a nominal charge for the VP7 personal use encoder. We believe the market will view VP7 as the leading codec in the world and it will command a premium price."

About On2 Technologies, The Duck Corporation

On2 Technologies is a leading technology firm at the forefront of video compression. The company revolutionized digital media delivery with the creation of its advanced full-motion, full-screen, TrueMotion(R) VP4/VP5/VP6 compression and streaming technologies. On2's TrueMotion codecs are used extensively in the video-on-demand, videoconferencing, Internet media, surveillance, and store-and-play markets. They operate with On2's own TrueCast Server and Player software, as well as third-party player and server products. On2's software is used by such leading companies as Sony, NTT, AOL, BBC, ABC News, Nullsoft, SeaChange International, China Netcom, NHK, and Nanwang Multimedia. The VP6 decoder has an installation base of millions through its inclusion in AOL 9 Media Player, Winamp 5, Viewpoint Media Player, and others. On2 licenses its software for use in set-top boxes, consumer electronics devices and wireless applications and offers high-level video encoding, customized technical support, and consulting/integration services. Located in New York City, the company has an office in Clifton Park, NY, and operations in Cambridge, UK. To contact On2 write to sales@on2.com or visit http://www.on2.com/.

On2 Technologies, Inc.

Web site: http://www.on2.com/

[Sharktooth]

Nominal charge for the VP7 personal use encoder?
Uhm... how much it is this "charge"? (BTW i dont think it's a good idea...)

[slavickas]

Originally posted by Sharktooth
Nominal charge for the VP7 personal use encoder?
Uhm... how much it is this "charge"? (BTW i dont think it's a good idea...)

yeah :(,although it shouldn't be (very) pricy, atm vp6 costs 99$ for commercial use for personal it should be way lower

[bond]

i wonder if vp7 will be backwards compliant?

[LordRPI]

Those guys in Clifton Park are insaaaaane :)

[DeathTheSheep]

So, this was just, what? 3 or 4 days ago, right? Additionally, On2 has removed the capablility of free VP6 consumer download.

According to the above documents, therefore, this is all in order to kick off their VP7 codec, but this for a "nominal fee."

How soon, then, do you think they will release the new codec? Or have I simply muddled the aforementioned facts?

Cheers, baa

[BoNz1]

Originally posted by pernas
VP6 works well with MPlayer on other platforms than M$ :D

Using a binary codec from windows does not count as being cross platform IMO.

[pieter1976]

This is what I red on the on2's website

On2 will also unveil its new TrueMotion VP7 codec to the U.S. market at CES. TrueMotion VP7 performs up to 35% better than VP6 at the same datarates and offers a very substantial improvement over the performance of H.264, Windows Media, Real 10, and DivX. In the weeks following CES, On2 will make content encoded in TrueMotion VP7 available on its web site.

I can't wait

[Sirber]

cool! I'd like some juicy specs :D

[Sharktooth]

Well, i wish on2 at least will release a Decoder-only package for free.
VP7 is a really promising codec...

[the Mad Duke]

Originally posted by Sirber
cool! I'd like some juicy specs :D

"On2 Releases VP7 Preview'

www.on2.com

[bond]

read more here (http://www.on2.com/vp7.php3)

they provide self encoded samples and a player able to decode them (anyone knowing if its dshow based?)
also of course the marketing claims, like every codec developer loves to spread: 50% better than vp6, and of course better than existing other codecs...

as vp6 wasnt available for free anymore i even more doubt that vp7 will be :(

[Sharktooth]

Yeah... but those samples are impressive...

[pernas]

Too bad i'm at work with no access to port 554..:(

How do they look compared to older (TC) VP6 samples @ 300/500kbps ?

[Sharktooth]

... better ... :D
They look really great.

[pernas]

Originally posted by Sharktooth
... better ... :D
They look really great.
Great !! ;) (...screenshots...anyone?)

[bond]

here a list of the additions made since vp6:

- 4x4 dct transform
- second order coding of DC information
- new “wide pass band” motion filter
- adaptive loop and new post-processing filters
- Improved two-pass rate targeting and datarate control
- Two-pass spatial resampling
- Light Level Change Adaptation
- Block Adaptive quantization and filtering
- context modeling in the entropy encoder

[Irwin]

Originally posted by pernas
Too bad i'm at work with no access to port 554..:(

How do they look compared to older (TC) VP6 samples @ 300/500kbps ?

I too.
I wrote to support on2 and now http download available.

BTW fourcc is ONYX

[Sirber]

Download to what? VP7 codec?

[Irwin]

sample video files !

[Sirber]

From the samples I've seen, I wanna try it T_T. If it's so great on anime at that bitrate, I absolutly need to try it :D

[Sagittaire]

From OneTech:

http://www.on2.com/vp7/Salmon.gif

OneTech don't use the best seeting for Nero H264 : in fact H264 Nero with best setting is always better than VP6. VP7 seem really better than VP6 too ... perhabs little better than H264 Nero but perhabs not !!?

In fact VP7 and Nero H264 are certainely comparable codec ... by far better than other codec (MPEG4 ASP, WMV9, RV10 ...)

[IgorC]

i haven´t any problem to download samples
http://www.on2.com/vp7/http/harry_vp7_400.avi

[Sirber]

How do you play this avi?

[IgorC]

Originally posted by Sirber
How do you play this avi?

i couldn´t . now i´m downloading VP7 player.

[Sirber]

can't wait for VFW :(

[Sharktooth]

Unfortunately vfw = $$$ ...

[IgorC]

http://www.on2.com/pdf/vp7_white_paper.pdf

it said that vp6.2 was better than Nero Avc (default Max. Definition profile). ¿¿??. i can´t agry at all.

[Irwin]

and Wmv9 is better then nero h264 !?! (psnr) I can´t agry at all too. They don't read "Codec shoot-out 2004".

Anyone have starsky and hutch good quality trailer ? - maybe little 400kbps nero h264 test.? ;-)

[Sirber]

Or they don't use the best settings...

[Irwin]

in this pdf:
Nero Digital H.264
- Defaults for maximum definition AVC profile
- High-quality two-pass
- Cartoon source
- Psycho-visual enhancements set to high.

I'm looking for starsky and hutch trailer for test this.

[Sagittaire]

Originally posted by Irwin
in this pdf:
Nero Digital H.264
- Defaults for maximum definition AVC profile
- High-quality two-pass
- Cartoon source
- Psycho-visual enhancements set to high.

I'm looking for starsky and hutch trailer for test this.

default setting aren't the best : for exemple deblock = 0 without adaptative is better for metric, psy low is better for metric than psy high ... ect ect ect

[babayaga]

Originally posted by Sagittaire
default setting aren't the best : for exemple deblock = 0 without adaptative is better for metric, psy low is better for metric than psy high ... ect ect ect
No, the settings they used are not the best but I'm not sure it explains the curve. Maybe they have a problem making the measurements (this could be difficult) ...

We know precisely the R/D curve for Football since the sequence was available for download. VP6.2 and our points are not where they are drawn.

Anyway, Bobololo did a small test encoding one of the available trailers to have a precise idea of VP7 capabilities (hp3, not exactly the same but the complexity should be very similar)
I was a bit anxious before looking at the comparison. Now I can peacefully work on some asm optimizations without fearing too much :)

[Sirber]

What is hp3 codec? :confused:

[babayaga]

Originally posted by Sirber
What is hp3 codec? :confused:
lol it's not a codec. It's a movie about a young orphan in a wizard school :D

[Sirber]

http://members.lycos.nl/diezijngek/magicgallery/Bunny@pancake.jpg
;)

[IgorC]

hp3 = harry poter 3 and Prisoner of Azkaban
i´ve ripped it just yesterday with Nero AVC with best settings :cool:


400 kbits of VP7 didn´t impressed me so much. maybe cause of low bitrate?
Can anybody upload 2 samples (just a couple of seconds) approx. 700 kbit/s VP7 and Nero H.264 at their best repectivly settings?

[Gabriel_Bouvigne]

Those funny marketting guys are using psnr to compare codecs, but setting psy enhancements to high (thus lowering psnr) in competitors.

It reminds me the Ms wma statements.
Wma7 - cd quality @64kbps !
Wma9 - cd quality @48kbps !!
And comparisons by Ms were done against a 2 years old version of Lame with intentionnaly badly choosen parameters.

Here it seems that the same scheme is followed: choose samples carefully, select unoptimal options for competitors, and tell the world that you made the best encoder.

[Sagittaire]

Here it seems that the same scheme is followed: choose samples carefully, select unoptimal options for competitors, and tell the world that you made the best encoder.

In fact VP7 is certainly by far the first or second codec for metric performance (if VP6 vs VP7 test is correct VP7 is IMO the best) ...

[Gabriel_Bouvigne]

In fact VP7 is certainly by far the first or second codec for metric performance (if VP6 vs VP7 test is correct VP7 is IMO the best) ...
Well, competing on the psnr field is a choice, but in this case why intentionnaly lowering the psnr results of competitors?

[DeathTheSheep]

I beleive that VP7 is one of the best codecs in the world. I thought the same thing about VP6, even after AVC (but to a lesser degree). The decoding of VP6 content was exceptionally fast.

I could not decode XVid, AVC, WM9 (surprise), RV10, or even experimental snow on my laptop (166mhz). VP6 decoded with ease (Main Profile).

Complexity is a big thing these days, because the world is filling with Movie-enabled cellphones (stupid cellphones :devil:), Pocket PCs (GO PPC!!! :cool:), etc, all with mobile processors.

Heh, even though VP6 wasn't supported on any mobile device I've ever worked with, It was still cool 'cuz it was just as easy as MPEG1 decoding (as far as I could tell) on my ancient laptop.

Cheers, baa

PS:
Seriously, ON2, everyone I know has a pocket pc. Me, bro, sis, cousin, tons of friends... Why ain't it supported on 'em?

[DeathTheSheep]

@vfw people:

VP7 does indeed support VFW, the standard everybody knows and (not everybody?) loves.

Features:
....
-Support for Video for Windows (VFW) and Direct Show

[On2Tech]

Can we take this opportunity to address a couple of issues.

The Salmon clip is one that was provided for us by a client for testing and it has formed part of our test suite for several years. It IS tough, just look at the data rates vs PSNR numbers and you will see that all the codecs find this clip hard. In our testing, the Nero/Ateme codec (in common with all the other H.264 other we have tested in the past), performed relatively poorly on this clip, though it did better in some of our other tests.

With all codec tests the results depend on a number of factors including but not limited to, the characteristics of the clip itself, the data rate, the image size and the codec settings. We have tried to get the best we can out of each of the codecs in our tests (given our time constraints) and the results are offered in good faith. As and when we have time to do more extensive tests, or repeat tests using the modified settings suggested by various posters, we will do so, and will post the modified results.

Within a fairly short time frame there will be an opportunity for you to test VP7 out for yourselves. We are confident that you will all see a sizeable improvement, as demonstrated by the graphs we have shown and all the other testing we have done. How big will that improvement be? Well, as always that will depend on the factors I mention above. If you are coding really hard material or animated material (especially at the low bandwidths or small image sizes that are of particular interest to certain of our clients) then it will likely be HUGE. If you are coding more straightforward material at high rates, the difference will be smaller.

One last point. This is only a preview and we are still working towards a formal beta. Some features in both the encoder and decoder are still being worked on or tuned, so we fully expect to get further improvements. For example, for the clips and graphs in the preview ALL post processing is disabled! I.e. what you see is the raw performance of the codec without any help from post processing at all.

On2Tech

[Sirber]

looking forward to it :D

[DeathTheSheep]

If you are coding really hard material or animated material (especially at the low bandwidths or small image sizes that are of particular interest to certain of our clients) then it will likely be HUGE.

Oh my goodness. I don't know what to say... Everyone who knows me here would know that that is what I live for (kind of). I'll also test this thing (and pretty darn thouroughly) if this is true...

If I understand correctly, the improvement will be rather significant with the following:

size=320*240
material=anime
framerate=7-24 fps
bitrate=excuciatingly low

Now, this is what I like.

PS: Oh, any info regarding PocketPC decoders? That would definately interest many people (I'm talking *tons*):cool:

[CruNcher]

Originally posted by On2Tech
Within a fairly short time frame there will be an opportunity for you to test VP7 out for yourselves. We are confident that you will all see a sizeable improvement, as demonstrated by the graphs we have shown and all the other testing we have done.

I think i can say for all users here, that we really looking forward to that :)

[JohnV]

Originally posted by On2Tech
We have tried to get the best we can out of each of the codecs in our tests (given our time constraints) and the results are offered in good faith. As and when we have time to do more extensive tests, or repeat tests using the modified settings suggested by various posters, we will do so, and will post the modified results.On2Tech Yes, of course you try your best. It's just a bit strange (but not surprising) that in your test even VP62 does so well against Nero AVC, which is a bit suspicous considering 3rd party results.
My personal opinion is that we should let this kind of comparisons be handled by 3rd parties, but I understand that companies need to hype and advertize..

[Sharktooth]

Well at least they have a VFW codec... and it works with any app...
However, on2tech said someting like: give us what parameters to use and we will update the graphs...
So, give him the parameters and we'll see :)

[JohnV]

I'd like On2 to give us the source video, so we can all check.

[SpaceV]

yeah ON2 should give you guys the
source video, you compress it and send the comprressed
file back to ON2 so that they can update their graph.

That makes the most sense.

Did some one sent ON2tech a msg already??

[Lefungus]

Giving valuable sample videos would be a too nice gift to Nero :)
After all these two companies compete in the same field.
Anyway, the marketing talk is always debunked once people are able to test it themselves.

[neo_anderson]

So, when will VP7 codec be revealed to us?

[Gannjunior]

Any news about this supercodec?
i'd like to test it!!:D

ciao!!

[Sirber]

nothing yet... :(

[LordRPI]

Wasn't C0mPr355 posting in the new-animated-source-for-codec-shootout thread just recently?

[Leo 69]

What a heck is their AVC-AUDIO :confused: Is it HE-AAC ?


128 kbps quality equal to that of compact disks even to expert listenershttp://forums.overclockers.ru/images/smiles/icon_hitrost.gif

[Sharktooth]

Originally posted by Leo 69
What a heck is their AVC-AUDIO :confused: Is it HE-AAC ?


http://forums.overclockers.ru/images/smiles/icon_hitrost.gif
No, it's on2 proprietary audio codec. Dont confuse it with h.264/AVC, it's a completely different thing.

[huang_ch]

Just wondering, why no comparison with RV10...

[huang_ch]

Originally posted by Sharktooth
No, it's on2 proprietary audio codec. Dont confuse it with h.264/AVC, it's a completely different thing.

So you mean on2 also has a audio codec, right? Any information of how this audio codec works compare to others ?

[Sharktooth]

From what i've "heard" (i tested the on2 software that included AVC for audio encoding) is something like mp3pro, maybe slightly better.

[huang_ch]

Originally posted by Sharktooth
From what i've "heard" (i tested the on2 software that included AVC for audio encoding) is something like mp3pro, maybe slightly better.
WooWoo~~~ Though may be not as good as HE-AAC... but at least much better than MP3, on2 brings us a rather big surprise....:D

[Sirber]

Their AVC is quite old. It's been there since VP4 or before.

[huang_ch]

Originally posted by Sirber
Their AVC is quite old. It's been there since VP4 or before.
:o I just become know on2 since VP6... Is it possible to use VP6/7 video codec seperately and plus a RA10 or OGG and combine into MKV or other containers?:D

[*.mp4 guy]

Yes that is easily possible, just use AviMux Gui to mux the avi file and the Vorbis (*.ogg extension) audio into a matroska (*.mkv). That is of course assuming there will be a vfw interface, then it shouldn't be a problem(directshow might work too?).

[Sirber]

Originally posted by huang_ch
:o I just become know on2 since VP6... Is it possible to use VP6/7 video codec seperately and plus a RA10 or OGG and combine into MKV or other containers?:D It's called RealAnime :p

[Sharktooth]

I wonder if on2 guys could enlight us with some VP7 news ... :)

[Doom9]

it's not quite finished yet. Besides that, you already know that it's not going to be free.

[Sagittaire]

test in progress with VP7:

Movie, CGI and Anime with VCD (384*288), DVD (720*400) and HDTV (1280*720) resolution ...

[SpaceV]

Sagi, you got an encoder?

Doom9 what means not quite finished yet?
Can you be more precise please, when do you think it will be done?

Or are you under a NDA?

Also whats happening in China? EVD just got reaffirmed by the Gov as standard, is that still VP6?

Also whats going on with XM sat radio using VP6/7??

[CruNcher]

SpaceV yes some testers allready have it just be patient

[Sagittaire]

very very little test with VP7:

source: HPII 720*576 MPEG2 MP@ML
pre-process: complete YV12 pre-process for all codec
post-process: PP4 for XviD, PP2 for VP7 and No PP for H264 in YV12

avs script for encoding:
VDM in fast compress mod (YV12)
CLI main profil for Nero H264 (YV12)


source=Mpeg2Source("G:\Mes dossiers\B.A\Harry Potter\azerty.d2v")
source=Trim(source,70,3145)
source=Crop(source,0,76,-0,-76)
source=LanczosResize(source,720,304)
Return(source)



avs script for PSNR test (YUY2):


# --> Video Opening <--

source=Mpeg2Source("G:\Mes dossiers\B.A\Harry Potter\azerty.d2v")
source=Trim(source,70,3145)
source=Crop(source,0,76,-0,-76)
source=LanczosResize(source,720,304)
source=ConvertToYUY2(source)

video=DirectShowSource("G:\Mes dossiers\B.A\Harry Potter\VP7.avi",fps=25)
video=ConvertToYUY2(video)

# --> PSNR analysis <--
Compare(video,source,"","PSNR-VP7.txt")



PSNR test

XviD for reference in q8
H264, Ultra high, VHQ4, VHQ bvop, H263, QPel, GMC, trelli, bframe 2/1.50/0.00

Minimum Average Maximum
Mean Absolute Deviation: 0.0156 1.7862 3.1250
Mean Deviation: -0.0380 +0.8608 +1.6260
PSNR: 35.0292 40.2419 65.2736
Overall PSNR: 39.7265



One2Tech VP7:
with best setting ... ;-)

Minimum Average Maximum
Mean Absolute Deviation: 0.0007 1.3482 2.6535
Mean Deviation: -0.2085 +0.0557 +0.2913
PSNR: 36.3714 42.5899 79.4781
Overall PSNR: 41.4765



Nero H264 :
with best setting ... ;-)

Minimum Average Maximum
Mean Absolute Deviation: 0.0061 1.3608 3.0858
Mean Deviation: -0.4652 +0.1647 +0.7735
PSNR: 34.5288 42.5690 69.2643
Overall PSNR: 41.4328



http://multimediacom.free.fr/Video/VP7.JPG
blue: VP7
red: H264
green: XviD


little visual test (250frames) with FFVI reencoding sample

Source -> FFV1 (http://multimediacom.free.fr/Video/Source-FFV1.avi)
VP7 -> FFV1 (http://multimediacom.free.fr/Video/VP7-FFV1.avi)
H264 -> FFV1 (http://multimediacom.free.fr/Video/H264-FFV1.avi)
XviD -> FFV1 (http://multimediacom.free.fr/Video/XviD-FFV1.avi)


[French On]
le MPEG4 ASP est aux fraises ... lol
[/French Off]

[Sirber]

PP2 for VP6? A line for VP7...

Is your test about VP6 or VP7?

[Leo 69]

I'd better stick to X264. ;) As far as I can see,it is not that worse than VP7 and it's free-of-charge.Besides it is better than current Nero encoder.

What about visual quality of the VP7 encodes, Sagittaire ?

[Sagittaire]

Originally posted by Leo 69
What about visual quality of the VP7 encodes, Sagittaire ?

Nero h264 (with best setting) at this time is very better than x264 (with best setting) thus for PSNR test VP7 is very better than x264 ...

Very little sample actually in upload (250 frames in FFV1 lossless reencoding for source and each codec) for visual test. IMO H264 is visually very little better (more constant quality) than VP7 for this trailer ... perhabs too low variability setting (70%) for VP7

[Sagittaire]

little visual test (250frames) with FFVI reencoding sample

Source -> FFV1 (http://multimediacom.free.fr/Video/Source-FFV1.avi)
VP7 -> FFV1 (http://multimediacom.free.fr/Video/VP7-FFV1.avi)
H264 -> FFV1 (http://multimediacom.free.fr/Video/H264-FFV1.avi)
XviD -> FFV1 (http://multimediacom.free.fr/Video/XviD-FFV1.avi)

[huang_ch]

Originally posted by SpaceV
Sagi, you got an encoder?

Doom9 what means not quite finished yet?
Can you be more precise please, when do you think it will be done?

Or are you under a NDA?

Also whats happening in China? EVD just got reaffirmed by the Gov as standard, is that still VP6?

Also whats going on with XM sat radio using VP6/7??

Do you mean China EVD is using VP6?

[huang_ch]

Originally posted by Sagittaire
little visual test (250frames) with FFVI reencoding sample

Source -> FFV1 (http://multimediacom.free.fr/Video/Source-FFV1.avi)
VP7 -> FFV1 (http://multimediacom.free.fr/Video/VP7-FFV1.avi)
H264 -> FFV1 (http://multimediacom.free.fr/Video/H264-FFV1.avi)
XviD -> FFV1 (http://multimediacom.free.fr/Video/XviD-FFV1.avi)
What's the final bitrate for each file?

[Sagittaire]

bitrate isn't important ... but I use for reference XviD with best setting for PSNR in q8 and 450 Kbps:

- XviD: 6882 Ko
- VP7 : 6948 Ko (+1%)
- H264 Nero : 6811 Ko (-1%)

In fact VP7 is very better than VP6 for anime source ... test in progress

[bond]

Sagittaire, you always talk about "h264". i assume you know that h.264 is a coding standard and not a codec, so which codec do you use when talking about h264?

[Sagittaire]

Originally posted by bond
Sagittaire, you always talk about "h264". i assume you know that h.264 is a coding standard and not a codec, so which codec do you use when talking about h264?

... and it's true

edit: Nero H264

but x264 is too a very good H264 version too ... :D
(result with ateme decoder)


Minimum Average Maximum
Mean Absolute Deviation: 0.0058 1.3949 3.0194
Mean Deviation: -0.6451 +0.1716 +1.1371
PSNR: 34.5454 42.3492 68.5553
Overall PSNR: 41.2319


Only 0.25 dB less than Nero H264 main profil with best setting ...

[SpaceV]

yes the last info I got is that VP5/6 is the codec of choice for
the chinese EVD standard. But this info is a year old, meanwhile there was the arbitration because ON2 did not get any money from E-World.

So I am not sure what is going on but since I have not heard that the EVD standard was modified I assume that it is still VP5/6.

Maybe the ON2 people on this board can comment on the situation.

[Sirber]

wasn't the "chinese EVD standard" a marketing joke?

[azsd]

EVD in china = divx player in world....
Now some DVDs with standard MPEG-2 struction will attached an encoded vp6 files in disk witch can both play in EVD players and classic DVD players.
is seems an fake...

[huang_ch]

Originally posted by Sirber
wasn't the "chinese EVD standard" a marketing joke?
:o Indeed, I'm a Chinese, maybe it is not so properly to critisize this but I really feel someone are too near-sight, always try to make a standard without real core technologies and fail many many times. That already hurts many people's heart like me:angry:.

[Sirber]

MPEG2 and VP6? I don't see the point. Does anyone have EVD specs?

[edit]

http://www.lsilogic.com/technologies/industry_standards/enhanced_versatile_disc_evd.html

Seems MPEG2 only...

[tmpy222]

As it was said in the article, they don't like to be controlled by a cartel of so called "patent" holders. So they make their own standard like they did with SVCD. The licenses they pay on EVD is one fifth of what they pay on DVD. Last I read, they are paying like $20 license on a $29 DVD player... I call that highway robbery for licensing.
What do you mean there is no "core" technology? Are you saying modern vp6 codec isn't "core" enough to compete with outdated mpeg2? I guess vp7 would have been better but it just came out. They are also HDTV compatible, I read. I was thinking they didn't choose WMV9 because nobody wants to be controlled by Microsoft either. :)
Anyway, I think what they are doing is a "long term" solution to the monopoly of patent holders cartel. It's cheaper to pay a single proprietary company than to 18 proprietary companies like MPEG LA, 3C, 6C...

The more choices, the better for the consumers like us. :)

[Sirber]

I'm saying it's not said anywhere it's using VP6. It sais it's using MPEG2.

[damrod]

i don't know but if they release dvd with both mpeg2 and vp6 encoding
it can be a good idea ....you can play the vp6 in 720p or 1024p
or choose for ordinary people the mpeg2 stream :)

[Sirber]

LSI Logic is providing E-world with HD encoding and decoding technologies based on the popular DoMiNo™ architecture. EVD is the first HD format created by using MPEG-2-based technology and a standard optical red laser. To deliver the industry's best quality HD and SD images, the DoMiNo architecture uses LSI Logic's award-winning PerfectView® MPEG-2 encoding algorithm.;)

[the Mad Duke]

The 1st generation EVD players used the LSI chip with MPEG2 compression. The next generation EVD players are now what are subject to speculation. A company called Analog Devices has also designed a chip for EVD. It's called the Blackfin and it has the On2 VP6 codec ported to it.
The arbitration between E-world and On2 should be over shortly. Anybody that wants to know what codec is being used in the next generation EVD players should find out soon.

[hdonly]

Any more VP7 test results?

Sounds like VP7 may exceed Ateme/Nero H.264?

[Sirber]

From my tests, I have blocking on my anime test clip at 536kbps. Issue has been reported.

[huang_ch]

Oh, I don't want to get this VP7 topic into a EVD topic. :o
Back to VP7, Siber, why you choose 5xxKbps to test? I now use 500Kbps with RV10/RA10 for Animes, and when I first saw VP7, it just claims to get about XX% improvement than existing codec, so I hope to get Animes encoded to below 400Kbps or even 300Kbps with VP7. :D Would that be a dream? :D

[Sirber]

536kbps was to be able to add a 64kbps HE-AAC audio stresm to have 600kbps :p

Currently, quality-wise with settigns I think it's best (IMHHO):

RV10 - Ateme > VP6 > VP7 > x264

RV10: 2-pass VBR, EHQ Extreme, Streaming RC
Ateme: long time ago... in the beta stage :rolleyes:
VP6: 2-pass VBR, Spacial resampling, automatic PP
VP7: same as VP6, no resampling (not avalible yet), unknown decoder settings
x264: 2-pass VBR, 8 ref, 3 bf

But, IMO now, 536 is not enough anymore. Too much used to CQ65 :o

[huang_ch]

Siber, you mean x264 is better than VP7 currently?

[Sirber]

opposite

vp7 > x264

but with the latest build (149), they are more equivallent :)

[DeathTheSheep]

Well, I'm a fan of extremely low bitrates. I like anime too, and I always encode it to 320*240 resolution. Remember my old thread about B-frames? You get the picture. ;)

So, this codec, having been claimed to perform exceptionally at low bitrates*, with animated content*, and at lower resolution*, is an absolutely perfect codec for me to experiment with.

So whom do I contact? On2Tech? On2 Customer Support? I don't want to send an email the wrong way, so I just came to ask from those of you who've obtained a beta/alpha trial. Thanks.


[edit]PS:
If you want, I could post some actual PSNR data, screenshots, and visual comments about each codec at different bitrates/res. I'd also say a bit about the pros and cons of each, thier speeds, their stability, their playback efficiency, etc. (if you'd like), my beloved Doom9 kindred. Cheers to us all.

DTS

[huang_ch]

Originally posted by Sirber
opposite

vp7 > x264

but with the latest build (149), they are more equivallent :)
But... in your previous post, you mentioned that Real>VP6>VP7... why? I think/guess/expect that VP7 should be much better than VP6/Real...:confused:

[Sirber]

My test clip is very difficult. I had issues with VP7 on it that affected quality. These issues has been sent to on2. VP7 lack "spacial resampling" in VBR mode, which has been used in VP6. I had also some blocking due to the new sharpen mode. I need to test it more to find the best settings for my clip.

[Sagittaire]

VP7 uses a sharpness slider control with a range of 0 to 7, as opposed to the 0 to 10 range in VP6. The best PSNR results are usually achieved with this value set to 0. Using the slider at higher values will provide sharper images but may also result in a few more artifacts (blocking/ringing artifacts).

On most clips, setting strPostProcessingLevel to 2 provides the best PSNR results, but if you are testing an extremely low datarate or animated clips you will find that 4 provides more visually appealing results.


For anime source:
Sharpeness = 0
Use PP4 for strPostProcessingLevel in reg

[Sirber]

Cool! Thanks. But with sharpenss at 0, the iamge get's a little too blur, so I will try 1 or 2.

[Sagittaire]

@ One2tech
your mail box is full ...

[Capirossi]

where I find vp7? :cool:

[Sharktooth]

Actually it's in beta testing stage, so you won't find it.

[pieter1976]

what futher improvements can we expect on vp7 over the beta version.
what will be the name vp7.1?

[SpaceV]

what I would like to know, are the On2 guys responsive
to your test results, do they agree, do you get updated versions
of VP7 ?
do you see results and improvements, is there progress?

please let us know at least that much.

I am very interested in VP7.

Thanks!

[Sharktooth]

Originally posted by pieter1976
what futher improvements can we expect on vp7 over the beta version.
what will be the name vp7.1?
??? Beta means it's still not released. So it will be VP7.

[Sirber]

Currently at VP7 beta 3. On2 guys are very responsive and helpfull. I did not test beta 3 yet, I'm encoding an anime serie :D. It should be over at noon. This build is supposed to work with avs2avi, so it will work via RealAnime. I will test beta 3 as soon as I get back from work.

[SpaceV]

Thanks for the feedback.
Did you see progress, quality/speed improvements over the
previous betas?

Hope we see the final VP7 soon.

[Sirber]

Yes. I had square problem, they are mostly gone. I also had a "too much keyframes per second", which has been fixed too. Good thing: first pass is too fast :D. I'll post more later.

[the Mad Duke]

VP7 has been released.

[eqbal]

Originally posted by the Mad Duke
VP7 has been released. ]

download link?

[Doom9]

it's payware now.. $19.99 for the codec, $29.99 for the codec and encoding app. You can buy it at www.on2.com

[zulu]

http://www.on2.com/duckutils/duck_license.php3?class=vp7

[Palikrovol]

It has any DirectShow decoder or it has only a VFW decoder?

[Sharktooth]

No free DS decoder available :( Why?

[Sirber]

DShow decoder and VFW encoder.

[the Mad Duke]

"The new codec software is available as a Video for Windows (VFW) plugin or integrated with the On2CompI Encoder application.

On2 is offering 10-day trial downloads for both encoding products so customers can test the software before deciding to purchase."

[eqbal]

VP7 , it's too slow for me.
AND what's the best setting for encode movies (with 2hr length) in VP7

[Sirber]

In my case VP7 is faster than VP6 :confused:

[eqbal]

Originally posted by Sirber
In my case VP7 is faster than VP6 :confused:

can you tell me your VP7setting for movie?

[IgorC]

VP7 1-st pass is very fast, but second is very slow. Why sharpness 0 is better (not like VP6.4 5-10)?

[Sirber]

Originally posted by eqbal
can you tell me your VP7setting for movie? I only compress animeVP7 1-st pass is very fast, but second is slow. Why sharpness 0 is better (not like VP6.4 5-10)?In my tests, 0 is too blusy and 5 is squary, so I'm testing at 2.

[eqbal]

setting for:
Mode?
Bitrate?
Max frames Btw Keys?
Sharpness?
Adjust Quantizer?
???????????????????

[IgorC]

Originally posted by eqbal
setting for:
Mode?
Bitrate?
Max frames Btw Keys?
Sharpness?
Adjust Quantizer?
???????????????????
http://forum.doom9.org/showthread.php?s=&threadid=90784 .

[SpaceV]

Hi All,

I downloaded VP7 and did a test with DVDx 2.3 (sourceforge.net).

I set it to 750 kbits and ajusted the quantizer min/max to 2/30.
I go for 1 pass best quality.

I started the encode and it tells me a projected filesize of 4.8 GIGA byte! So I let it encode a couple of minutes from the movie
and the output size is 190 MB !

After I stopped the encoding I went back to my codec settings and just changed VP7 with VP6.2 and used the same settings and the projected output size is around 600MB as compared to 4.8 GB.

What is wrong with VP7 or what am I doing wrong.

Anyone know what is going on.

Help please.

[SpaceV]

I am using the 10 day Trial version and when the
trail box comes up with the buttons "Buy", "Install License" and "Close", I hit the close button.

[Sagittaire]

quantisation interval for VPX is not the same than MPEG4 ASP (XviD or DivX) : don't touche RC defaut setting. If you want better target bitrate use higher undershoot and don't forget that VPX use kbit/s and not kbps (1000 kbit/s = 1024 Kbps)

[virus]

Originally posted by Sagittaire
quantisation interval for VPX is not the same than MPEG4 ASP (XviD or DivX) : don't touche RC defaut setting. If you want better target bitrate use higher undershoot and don't forget that VPX use kbit/s and not kbps (1000 kbit/s = 1024 Kbps)
Wrong. Only VP6 uses the monstrosity 1 Kbps = 1024 bit/s. Instead VP7 (finally!) uses 1 kbit/s = 1000 bit/s, like all other decent codecs on the face of the planet.

As for RC precision, VP7 sould be clearly better than VP6 (with default settings), though I don't know if On2 has already completely solved the issue that makes the RC sometimes drift (and oversize) at high rates. If this problem doesn't occour, VP7 will usually hit the target very well even on short clips.

As for Sharpness, I recommend 3-4 for mid bitrates, and 5-6 for high bitrates (natural material). Keep in mind that Sharpness in VP7 is dealt with in a very different way than VP62, so they don't really compare.

[SpaceV]

1 minute of VP7 has 70 megabytes, where 1 minute of
VP6.2 with identical settings has about 7.5 megabytes.

Sagi, you are tryong to say that the quantizer and the 24 bit difference in the bitrate make that difference? Its about times 10x
larger. I think my VP7 size is as big or bigger than my original MPEG2.

There is something completely wrong here.

Did anyone try the free download VP7 codec version?

[On2Tech]

I would like to ask those of you who are sending private messages to my inbox to try and send anything related to vp7 problems and concerns (not rants or raves) to codec_support@on2.com

As I do not have the time to always check my inbox to see if it is overflowing or full, this would be a faster way for me to get back to you guys.

I only ask that you please keep this technical and civil.

Thanks

on2tech.

[netchris]

I am sorry i have to report a negative opinion, but so far i am kinda disappointed. First pass is very fast , thats good, but second pass is extremely slow (iirc slower than vp6, surely slower than nero avc). Made some tests with anime eps, 500kbps the target bitrate and the results are blocky in many parts, with ringing evident in many places. Compared to a nero-avc encode nero seems almost perfect, while vp7 has obvious flaws. I bet by tweaking i can get a bit better picture, but imo it cant compete with avc codecs. At least not now, when it matures it might be alot better....

[Sirber]

In my AMD64 3000+, VP7 is faster than VP6 at 2-pass VBR, 536kbps. :confused:

[netchris]

I am not in a position to make any claims about the speed compared to vp6,cause im too lazy to reinstall vp6 :rolleyes: My proc. is athlon xp 2800, the codec might do heavy use of sse2 which my proc doesnt have, so that could be the culprit (if i am correct about being slower than vp6).
Made some tests with video clips, and the codec is very good, it surely performed better with real people video than it did with anime content. Still i find h264 (nero to be precise, as x264 is not stable enough yet) to be better.

[huang_ch]

Originally posted by Sirber
In my AMD64 3000+, VP7 is faster than VP6 at 2-pass VBR, 536kbps. :confused:
AMD64.... How long did it take you to encode a 20-25min Anime with VP7?

[Sirber]

I only encoded my little test clip so far. First pass is at least 4 times faster than RV10 EHQ50. Second pass is around 8FPS IIRC.

[SpaceV]

ok I used VP7 with VirtualDubMod
and it seems to work although my FPS is around 2-3
on my AMD64 @2.45GHz

Sooo why does VP6.2 and all other codecs... Xvid, x264 etc...
work with DVDx2.3 but VP7 does not?

Using DVDx, the output size is about 4 times the requested datarate
with quantizers set to 4/56.
VP7=12MB for 1 minute, VP6.2 same settings 3 MB
(it was a very easy minute with lots of black)

With VdubMod the datarate is ok with both VP versions.

Any idea?
Would be great if someone could reproduce the same thing.
As I said DVDx is open source from sourceforge.net

[dragongodz]

Using DVDx, the output size is about 4 times the requested datarate
not here. just did a quick test on a 1 minute clip and all was fine using DVDx 2.3.

can you tell me what settings in DVDx you used ? export(RGB24 or YUY2), resolution etc ?

[EDIT]
i know what you mean about the speed. 6.1(advanced profile) was around 27.5fps while 7.0 was around 6.6fps on my Athlon XP 2400+. BIG difference in speed.

[Manao]

Did anybody get the same issue as i did, with the YUV to RGB vfw converter installed with on2comp ? I wasn't able anymore to use YV12 as input through virtual dub for x264. I had to deinstall on2comp.

[Sharktooth]

Yes! same issue here! Just change form the on2 yv12 handler to xvid or divx in the registry. No need to uninstall on2compi.

[C0mPr355]

space you didnt happen to have PCM audio compressing into your vid stream did you ?

That would account for a much higher file size.

[WorBry]

After all of the anticipation I'd have to say that I'm rather dissappointed.

Ran a few comparisons with VP6.2 using several Pal Type II test clips as source and my standard AVS script:

AVISource
ConvertToYUY2(interlaced=true)
ColorYUV(off_y=-16,gain_y=16) - for luma corection with Sony DV codec
KernelDeint(order=0,sharp=true)
LanczosResize(720,544)
Undot

VDubMod:
Fast Recompress
Audio MP3 CBR 128

Codec configuration:

Both: 2000kbps, 2-pass VBR (best quality), progressive, no denoise, AutoKF, maxKF 9999, no resampling. Other settings at default except for:
VP6.2 - sharp 7, quantizers min 2 (max default 40)
VP7 - sharp 5 (quantizers default min 0, max 56)

Observations:

1. VP7 2nd-pass painfully slow
2. Average PSNR - no significant differences between VP6.2 and VP7
3. No improvement after changing VP7 sharpness to 0 (as recommended)
3. CPU demand on playback with TrueCast Player and MediaPlayerClassic (PC: AMD XP2800, 1GB DDR RAM): the VP7 encodes drew 95-100% usage, more than twice that of VP6.2(40-50%)
4. Couldnt see any decoder configuration option for VP7. Is it hidden somewhere?

Cant see anything that would cause me to reach for my credit card as yet

[Sharktooth]

VP7 has been optimized for LOW bitrates... so a 2000kbps compare is obviously misleading...

[Sirber]

2mbps for HDTV is low bitrate :p

[Sharktooth]

LanczosResize(720,544)

:p

[Sirber]

http://www.gosmiley.com/frowning/weinender_smili.gif

[edit]
4. Couldnt see any decoder configuration option for VP7. Is it hidden somewhere?Registry! :D

[WorBry]

I prefer (visually) 720x544 to 720x528 for conversion from 720x576 DV at full res.

Maybe VP7 is optimized for low-bitrate, but I'd been getting excellent PSNR results with VP6.2 at around 2000kbps, in fact comparable to XviD at 2800kbps. However, on my PC at least, XviD (and DivX) have a lighter CPU demand (25%) than VP6.2 and I was hoping that VP7 would perform better in this respect.

[Sharktooth]

That's why PP is activated in both VP6.2 and VP7.
Try disabling PP and see VP6 fly... :)

[JnZ]

Hi, guys.

I've just test VP7, but it's really VERY SLOW.
I test 1000 frames from Winnetou 3 (good old film :cool: )
Settings:
LanczosResize(672,288)
Biterate 908 kbits (This biterate is for full one CD)

I also test x264 r.159 and XviD 1.0.3.
Every codec with maximum quality settings.

Encoding speeds on my configuration:

Codec First Second
------------------
VP7 ~34fps ~1-2fps
x264 ~16fps ~13fps
XviD ~40fps ~11fps


VP7 is rrrreealllyyy slowwwww brrrrrr.
On my visual comparsions results are: x264>VP7>XviD.

EDIT: I must correct FPS numbers, because my proc is overclocked at NOW :D

[Leo 69]

@ JnZ


Where have you taken x264 rev.159 ? http://forums.overclockers.ru/images/smiles/icon_eek.gif

[JnZ]

Originally posted by Leo 69
Where have you taken x264 rev.159 ?
Just downloaded from http://x264.nl/ :cool:

[SeeMoreDigital]

As Sharktooth mentioned... Where's the DSdec filter?


Cheers

[Sharktooth]

Originally posted by Leo 69
@ JnZ


Where have you taken x264 rev.159 ? http://forums.overclockers.ru/images/smiles/icon_eek.gif
Rev 159 has been just built. Tomorrow i will compile CPU optimized versions.

[Leo 69]

Just tested VP7 @ 500 kbps @ 720x320 short clip and noticed dissapointing mosquito noise. I haven't seen anything like this since I've started using H.264... This reminded me of my Xvid times :rolleyes: Yes, certainly VP7 is MUCH better than Xvid, but it isn't such a serious competitor to current Nero's H.264 realization in Recode, that's for sure.

[Sirber]

VP6 codec has much more muskito noise than VP7. Remember that VP codecs don't use bframes :)

[hdonly]

http://on2.com/duckutils/duck_license.php3?class=player

Fixes CPU complexity problem. Clips look better too!

[Leo 69]

Originally posted by Sirber
Remember that VP codecs don't use bframes :)

Dirty shame :D

[eqbal]

x264 is faster and have better quality (vs VP7)

[Yong]

I tried, it's sloooooow! 4646 frames need 46 min. to complete the encoding process!
(on P4 2.4GHz machine, but working on backgraound, encoding thread priority is "normal")
Decoding also slow, eat more CPU power than h264 if play wiht Media player classic, even the audio playback will skip or pause...
but no problem with Mplayer.
look good at 256kbps:p
I try to compare it with x264 @ 256kbps, may be x264 has potential to better than VP7...:cool:

EDIT: i found some encoding option were disabled(grayed-out),
trialware restriction?

[Sirber]

not implemented yet :)

[On2Tech]

Originally posted by Leo 69
Just tested VP7 @ 500 kbps @ 720x320 short clip and noticed dissapointing mosquito noise. I haven't seen anything like this since I've started using H.264... This reminded me of my Xvid times :rolleyes: Yes, certainly VP7 is MUCH better than Xvid, but it isn't such a serious competitor to current Nero's H.264 realization in Recode, that's for sure.

Our tests show that we are either equal to or better than Nero's codec in almost all situations. If you find a clip upon which VP7 is not performing as well as Nero or some other codec we'd love to get your help in resolving the issue.

Please write us at codec_support@on2.com. If you are able and willing to get us your source, coding parameters we'll attack it with a vengeance.

Even if you don't. Thanks for trying our codec. We understand testing these things take time, and appreciate that you have taken the time to download and try our codec out. Even negative comments help us make the codec better ( sometimes especially)...

We've found a number of issues and the codec is getting better and better because of the outstanding help we've gotten from people who are regular posters on this board. If you are dissappointed today, I suggest you might not be with our next release. As such, we are actively responding to these issues and watching these forums.

[On2Tech]

Encoding Speed

Our encoder speed is slower than we'd like. We've been working hard to make it run faster and will continue to do so. I think some of the beta testers can probably attest that its a lot faster than the earlier versions. We're working on this issue.

Problems running from certain apps

We have also had people report to us that the codec doesn't work in a couple of apps ( though it seems to work in vdbub). We are trying hard to resolve this issue and hope to have a fix up early this week.

Encodes on P3's, Celerons, Athlons worse Than P4's

We've also discovered and are working on currently a problem with encoding on different cpu's. On some processors the encoder is currently producing better results than others. ( We think there is a bug in either our sse optimized code or our mmx code) . It seems that our P4 routinely beat encodes using p3 Celerons, and athlon machines.

Decode Speed

We've made significant improvements to the decoding speed of our codec. ( You should be able to see some of these improvements by playing video through our truecast player). The recent release of the vfw codec did not have these improvements. The next version will. The improvements included roughly a 20% overall improvement in decode, and processor adaptive postprocessing ( we do the best post processing we can do based on the time spend decoding the frame)..

[On2Tech]

Originally posted by netchris
I am sorry i have to report a negative opinion, but so far i am kinda disappointed. First pass is very fast , thats good, but second pass is extremely slow (iirc slower than vp6, surely slower than nero avc). Made some tests with anime eps, 500kbps the target bitrate and the results are blocky in many parts, with ringing evident in many places. Compared to a nero-avc encode nero seems almost perfect, while vp7 has obvious flaws. I bet by tweaking i can get a bit better picture, but imo it cant compete with avc codecs. At least not now, when it matures it might be alot better....

Our tests on Anime have shown the exact opposite. Ie we look less blocky than Nero-AVC. Can you please let us know what material you have, what machine you've used, and with what parameters you have encoded it?

[eqbal]

what's your parameters that use to encode movie with better result than nero?
i like to know it.:p

[Leo 69]

@ On2Tech

Please post parameters at which VP7 perfoms at its best in low-bitrate
conditions (400-900 kbps) and in DVD-resolution (720x320). I'll see what I can get up to :)

Cheers

[virus]

Originally posted by On2Tech
Our encoder speed is slower than we'd like. We've been working hard to make it run faster and will continue to do so. I think some of the beta testers can probably attest that its a lot faster than the earlier versions.
yeah, the first beta was much slower than the current trial version. But the problem in VP7 is not that the codec is "slow". The problem is that the GUI currently only offers a couple of presets, and they're both oriented to high quality (and thus most of the time-consuming options are enabled).

It's clear to me that the problem is that VP7 doesn't currently offer to the user the ability to turn off some flags and get higher speeds (at the expense of quality of course, like it happens with all codecs). The solution imho is not to reduce the complexity of Best and Good modes (unless you can do it through pure code optimization, i.e. without any quality loss - that kind of speedup is always welcome :)), but the adoption of a proper speed/quality slider with 4 or 5 presets that allows the usual tradeoff between encoding times and output quality.

There has been a lot of talk here about the design of a "n00b-friendly" GUI for XviD some time ago. Several people outlined how the most simple design was the adoption of such a slider, with the codec choosing appropriate parameters internally for each preset. This is the way to go.

BTW for those who think that VP7 is "painfully slow", I can reveal that time ago On2Tech talked about an internal version capable of encoding D1 in realtime. So it's really just a matter of missing speed scalability in the user interface, *not* of On2's devs being unable of writing a codec with decent speed ;)

cheers :)
virus

[On2Tech]

Originally posted by SeeMoreDigital
As Sharktooth mentioned... Where's the DSdec filter?
Cheers

Its coming ( both encode and decode ). Its on our list of things to do. First we're trying to handle some of these other issues.

[AlexeyS]

1. On2's main problem is poor marketing. I think VP7 is the best (little better than H.264 for me), but there is no Hardware Players with VP6/VP7 support...

Look at DivX and WMA9 - these codecs really blurry and with blocking, but they're popular because DivX Network and Microsoft use strong marketing.

2. Do you plan support of 2 CPU systems and 64 bit support?

3. Decoding speed is also problem for me. It's little slower than H.264, and much slower than VP6... I even can't play 1280x720p video without lags on my Athlon FX-55 CPU.

[Sirber]

I prefer more efforts on the codec than on marketing. You need marketing when you sell crap ;)

[AlexeyS]

Originally posted by Sirber
I prefer more efforts on the codec than on marketing. You need marketing when you sell crap ;)
Yes, and 99% users watch movies in this 'crap'. I don't think it's right.

[Sirber]

Life is not perfect. :(

[Sagittaire]

Originally posted by virus
yeah, the first beta was much slower than the current trial version. But the problem in VP7 is not that the codec is "slow". The problem is that the GUI currently only offers a couple of presets, and they're both oriented to high quality (and thus most of the time-consuming options are enabled).

It's clear to me that the problem is that VP7 doesn't currently offer to the user the ability to turn off some flags and get higher speeds (at the expense of quality of course, like it happens with all codecs). The solution imho is not to reduce the complexity of Best and Good modes (unless you can do it through pure code optimization, i.e. without any quality loss - that kind of speedup is always welcome :)), but the adoption of a proper speed/quality slider with 4 or 5 presets that allows the usual tradeoff between encoding times and output quality.

There has been a lot of talk here about the design of a "n00b-friendly" GUI for XviD some time ago. Several people outlined how the most simple design was the adoption of such a slider, with the codec choosing appropriate parameters internally for each preset. This is the way to go.

BTW for those who think that VP7 is "painfully slow", I can reveal that time ago On2Tech talked about an internal version capable of encoding D1 in realtime. So it's really just a matter of missing speed scalability in the user interface, *not* of On2's devs being unable of writing a codec with decent speed ;)

cheers :)
virus

very good post ... agree, agree, agree and agree ...

like standard, better, extreme or insame for DivX ... or normal, good, best, extra or full for H264 Nero

[Gannjunior]

I agree with Virus too. " Saggie parole! " ;) Moreover: VP gui, in general, really has a very little tuning interface ( in particular if u compare it with xvid tuning possibilities...).

ciao!

[Sirber]

Yeah, I have to agree too :(.

[DeathTheSheep]

I think 3ivx is the champion at "Quality Slider" technology. 3ivx 5, yes... One could learn a lot from it, I s'pose.

[Sirber]

It's cool to have a quick quality slider, but to have access to advanced stuff is cool too. Beginner vs Expert :rolleyes:

[DeathTheSheep]

3. Decoding speed is also problem for me. It's little slower than H.264, and much slower than VP6... I even can't play 1280x720p video without lags on my Athlon FX-55 CPU.

My suggestion is to lower postprocessing. The default value is 2. In the registry, you can change strPostprocessingvalue= to 0-4, 0 being the lowest. Try it if you have lags--it just might help!


Oh, and On2Tech, I'm exteremely satisfied with VP7. Is the VP6/7 Pocket PC decoder coming out soon?

PS: Nero just made its Showtime Mobile's debut.

Good luck to On2, VP7, us, me, and the world. Cheerio lads.

[DeathTheSheep]

Originally posted by Sirber
It's cool to have a quick quality slider, but to have access to advanced stuff is cool too. Beginner vs Expert :rolleyes:

Very true. Tabs->
1st tab (Basic/Beginner's tab)=Beginner settings with slider.
2nd tab (Advanced User's tab)=the super advanced stuff with the nitty gritty junk that the sliders actually do.
3rd tab (Advanced features and settings)= to enable optional tags and settings in any arbritraty order to acheive either excellent or terrible results, depending on how they're set.

[Sirber]

Seems good to me :)

[edit]
"You know what programmers need: helping hands that write code.. not people asking for this and that (that's just annoying and kills your motivation faster than anyone can say please)." -Doom9True, but I don't mind feature request if they are realist :)

[webwonk]

Will Macromedia's SWF8 version of their player decode VP7 or will it just be VP6.2?

When will this be available to us?

Will it use Flix Pro or some such encoder? or will there be a DirectX version of the SWF8 video encoder?

Any info you can share would be most appreciated. Thanks and keep up the good work.

[webwonk]

Has anyone heard whether Macromedia (now Adobe) is using On2's VP6 or VP7 in SWF8?

[Sharktooth]

@ On2tech: im testing the new 7.0.6.0 release with the usual spiderman 2 trailer (PAL DVD source).
The first round is a very low bitrate test @ 400kbps (without audio).
If you have any suggestions on VP7 settings please let me know asap.
However my first impressions are good even if the "best quality" profile is still slower than Nero Recode slowest settings...

[AVmaniac]

Hi ...
i've tested tested the latest release of On2's VP7 codec and when compared with NERO-AVC and the latest x264(b216) it produced the best picture at 632kbps (tested for fitting simple cd with dualHEAAC5.1stream )on "Finding Nemo"
It seems as if the 1st pass is a bit faster but the 2nd pass is nearly equal to earlier releases, it's quite slow [ from 1-6 fps @ 720x400 avs-script VirtualDub ]

My maschine configuration:
Athlon64-3400+(So754)
1024MB-DDR400
normal process priority

[Sharktooth]

Here are the 400kbps encodes...
VP7.0.6: http://www.aziendeassociate.com/test/vp7.0.6-400.avi
x264 rev216vfw: http://www.aziendeassociate.com/test/x264_rev216-400.avi

Results could have been slightly better but since no one wants to encode at 1-2fps i tried to choose the most common settings for both codecs.

EDIT: Fixed links...

[Sirber]

* Pass 1/1: Frame 20515/181560, 49.35 MB, 0.65 FPS, ETA 69:16:54

:rolleyes:

[Sirber]

Originally posted by Sharktooth
Here are the 400kbps encodes...
VP7.0.6: http://www.aziendeassociate.com/test/vp7.0.6-400.avi
x264 rev216vfw: http://www.aziendeassociate.com/test/x264_rev216-400.avi

Results could have been slightly better but since no one wants to encode at 1-2fps i tried to choose the most common settings for both codecs.

EDIT: Fixed links...
I don't know... x264 tends to smooth but keep "constant details". VP7 have blocks and variable details (second part is blurier than x264). VP7 seems to have difficulties with scene change. Overall I'd go for x264.

[IgorC]

I noticed that VP7.06 has a very hard chroma optimization. May be codec trying to trade off between chroma(color)/luma(details).
I also tried to compare H.264 Nero, x264 cls rev216, VP7.06
Color is handled better by H.264 and x264 , but VP7.06 insignifically has better details. I can´t said what is better.

[Sirber]

I did SteamBoy for 1CD (712kbps), DVD rez and I'm not sure of the results. I have blinking squares in plain color. I'll make it with RV10 or x264 to compare. Took 2 days of encoding though...

[Sharktooth]

Dont use the "2-pass Best quality"...
Good quality mode is far more convenient, more configurable and a lot faster (even with the Encode speed = 0).

[Sharktooth]

Dont use the "2-pass Best quality"...
Good quality mode is far more convenient, more configurable and a lot faster (even with the Encode speed = 0) without loosing too much...

[Sirber]

ok.

PAR don't work with MKV, set 1.78, got 1.50 on playback.

[Sharktooth]

Matroska has PAR stored directly in the header, so it should work... unless the VP7 decoder doesnt support resizing.

[Sagittaire]

Originally posted by Sharktooth
Dont use the "2-pass Best quality"...
Good quality mode is far more convenient, more configurable and a lot faster (even with the Encode speed = 0) without loosing too much...

yes ... IMO good 1 or good 2 are certainely the best quality/speed in real encoding conditions (150 000 frames and more)


Mode Speed Mins fps OPSNR SSIM
---- ----- ----- ---- ------- -----------
Best n/a 13:29 3.8 40.7703 69.60783207
Good 0 8:5 6.3 40.6787 69.26216614
Good 1 5:39 9.1 40.6234 69.15486557
Good 2 5:20 9.6 40.5712 68.97606882
Good 3 4:17 12.0 40.2839 67.55805933
Good 4 2:14 23.0 40.1687 67.18458972
Good 5 1:30 34.2 39.9335 65.64112115

[IgorC]

It was mentioned that in VP7.06 RC control was improved and the dependence from min. quant is smaller now.
I did some encoding here at 700-710 kbit/s mixed motion ( high/low motion 50/50%). During high motion frames received very high quant. while during low motion very low. difference of quality between low/high motion is easy noticeble;
Settings : Min Q - 4, VBR - 75, Min 25%, Max 200 (as a default)
Should it better to force min. quant (6?) or adjust RC settings (min,max)?
There is no problem with RC of Xvid and Nero AVC.I can upload video.
Thanks.

[Leo 69]

When will be the v7.07 available ?

[On2Tech]

When will be the v7.07 available ?

Whilst I cannot give you an exact date, as there are still a couple of improvments we are working on, I suspect it will be one or two weeks rather than days or months.

[SpaceV]

any speed benchmarks?

[Leo 69]

any speed benchmarks?

Yes, ~1 FPS for best quality mode on Athlon 3200+. Decoding of such 720x512 video takes 35-40% of CPU. I didn't try other modes.

[Sharktooth]

"Other" modes are much faster. Depends on the speed slider.
However avoid "Best Quality" mode if you're a speed freak. Take into consideration that the difference between best quality and good quality at "0" is not so evident but the fps gain is something like 3x.

[Inventive Software]

I have something that may interest you.

When doing the 1st Pass in VDubMod, I set my settings accordingly, and I tell it where I want to put the 1st Pass Information File. I prefer to put it in a separate directory to where the output file is going. Only problem is that the file I specified to go in the separate directory isn't in there, and there is a file called vp7.fpf in the directory where the output file is. What's going on?

When trying to get the 2nd Pass going, I can't because VDubMod crashes with a kernel32.dll error. I assume this is because it can't find the 2nd Pass file.#

This is kinda bad practise, because if I specify where I want the 1st Pass file to go, I expect it to go there, and not elsewhere where I don't want it.

[On2Tech]

I have something that may interest you.

When doing the 1st Pass in VDubMod, I set my settings accordingly, and I tell it where I want to put the 1st Pass Information File. I prefer to put it in a separate directory to where the output file is going. Only problem is that the file I specified to go in the separate directory isn't in there, and there is a file called vp7.fpf in the directory where the output file is. What's going on?

When trying to get the 2nd Pass going, I can't because VDubMod crashes with a kernel32.dll error. I assume this is because it can't find the 2nd Pass file.#

This is kinda bad practise, because if I specify where I want the 1st Pass file to go, I expect it to go there, and not elsewhere where I don't want it.
We've duplicated the problem and will resolve it. Thanks for catching this problem.

[On2Tech]

any speed benchmarks?


The compression speeds for each of the different settings are similar to 7.06. There are some numbers for Sagittaire's HPII test clip in our release notes thread (also see his posting above).

These tests were run on a 3.2 Ghz P4 at 446 Kbits/sec. Note, however, that speed is somewhat material and data rate dependent (especially for best quality).

For this clip the approximate range is 3fps best to 30+ fps for good quality speed 5.

The biggest improvements in this release have been at the higher speed settings (SSIM 2 for speed 5) which is now on a par with speed 3 in the previous release.

As Sharktooth points out, the loss in quality when moving from best quality to good quality (especially at speeds speeds 0-2) is generally quite small, but they are much faster. We recognize that most people will prefer to run using these faster speed modes, which is why they have been the main focus of attention for this release.

On2Tech

[Inventive Software]

Thanks On2Tech. In the meantime, is there anyway to get around this problem? If not, I can wait fairly patiently for a newer version.

[On2Tech]

Thanks On2Tech. In the meantime, is there anyway to get around this problem? If not, I can wait fairly patiently for a newer version.

I guess you are currently stuck with running both passes from a directory you can write too. Other than that I have no suggestions.

I'm sorry about that.

[IgorC]

There are positive statements about new version Vp7.07 on some sites. VP7 was improved in continious form since the first preview version. Last VP7.07 has a good decoding speed and quality on par with H.264.
There are some comparisons on 400-700 kbps on official site. But how is well VP7 for HD content?
VP6 has a profiles. VP6.2 (sharpness profile) brings better result, especially on high bitrates, comparing with VP6.0 (stream profile). Is there any plan to create profiles for VP7.

[Sirber]

Gonna make a little comparaison at 400kbps, adding VP7 and ateme to my "x246 vs RV10" comp.

[Sirber]

VP7 did not hit correct filesize.

http://www.detritus.qc.ca/files/comp400/

[Sharktooth]

The clip is too short.

[Atamido]

1:49 @ 25fps = 2725 frames.

So, in 2700 frames, with many many scene changes, this codec couldn't figure out where to spend more or less bits? I guess I just expect algorithms to be a little better than that.

[Sirber]

keyframes are too high Q, P frames then gets ugly.

[Sirber]

Is there a way to make the codec hit correct filesize always? WHatever the bitrate I set, I get 6.9MB.

[Sharktooth]

Try tweaking the overflow...