VP6 version 6.1.0.2 released!

[C0mPr355]

Hello gentlemen and ladies :)

We have just released an updated version of the VP6 encoder.

VP6 6.1.0.2. http://www.on2.com/

the new version fixes some of those pesky bugs that you have all been having. Specifically:

-badframes.stt has been addressed and eliminted.
-The interlace problem with half fields being displayed has been fixed.
-some slight speed improvements on the low end.
**note that the two-pass improvements that are stated on the website are mainly from the 6.0.9.2. version of the codec.Those improvements are obviously included in this new version;)***

Enjoy and please put all feedback in this thread :)

c0mpy

[Sirber]

any news on VBR quality on 2-pass?

[slavickas]

thanks for new version, now i hope that everybody recodes with new version (i got >10megs badframes.stt file , by just watching winamp5 tv)

btw there is illogical recomendation in doc
>Low Datarate Compression
>Peak Bitrate Set to 80-90%.
if i correctly understood actually it will limit b\max bitrate, so why not set proper bitrate and peak to 100%?

[C0mPr355]

right now this was mainly to get rid of the bugs mentioned above. It is not really any type of update to a better two-pass mode since 6.0.9.2. We are working on the two-pass mode improvements in the VBR area and so on. As for

- as for the peak bitrate setting that has more to do with streaming than local playback. So it in effect refers to how much of the bitrate is streamed back to limit bandwidth overusage. But duly noted is your find. It is actually an error in printing...the Peak bitrate option is not available in VBR encodes since there is no streaming involved. I will have that updated asap. Thanks :)

c0mpy

[netchris]

First of all congrats to C0mPr355 and the vp6 team for giving us an amazing codec.
I made some tests with difficult music video clips (lots of motion and vivid colors-I use them as reference clips to test the codecs), and the results with yesterdays release are top quallity.I use an average bitrate of 550 kbps.
The two previous releases were problematic visually when using 2 passes (blocky frames - due to lost keyframes I guess).With the new release the results are clearly better and it seams to me that the codec its starting to show its potential. I believe C0mPr355 you are underestimating this release when saying there are no important fixes for the 2 passes algorithm (mabe on purpose?) as I found the previous release not usuable because of the blockiness.
Anyway for me vp6 is slightly better than real 10 (at least for the bitrates and the purpose that I use it) as it gives a more detailed picture and an absolutely pleasing result.

:) Thanks again keep up the good work!!!!! :)

my 2 cents
Chris

************************************************
no flames please for any grammatical mistakes :-) english is not my native language.

[virus]

Hi!

I've tried out both VP6 6.0.9.2 and 6.1.0.2 and I'd like to share my views with you. I made some 2-pass encodings (clips from 40 seconds up to 6 minutes long) at 512x288 @ 25 fps (from DVD sources) with datarate between 650 and 800 kbps (~1 CD rips) and I compared it with XviD 1.0-beta2.

The output was fine. I think VP6 is very good at matching the colours of the source. Instead in some XviD clips there was a slight shift to red (not only on faces) and the original colour range was lost. Same (and maybe worse) for DivX5. Also, under 700 kbps (that's < 0.2 bpp) VP6 wins even for overall quality. I did my tests with no spatial/temporal resampling and min quantizer set to 4. I'm just wondering how a codec with no B-frames can reach such a performance. IMHO from a purely technical standpoint VP6 is a very interesting codec. :eek:

Also, I did a test with 120 KB/s rate (983 kbps, 0.266 bpp) and compared VP6 to XviD to dig deep into the classic "XviD retains more detail". This turned out to be true, but only with MPEG quantization, as I expected. XviD w/ H.263 looked quite similar to VP6.
So I think a different quantization should be embedded in VP6 for higher bitrates (> 0.2 bpp). VP6 is optimized for low bitrates but it doesn't scale well at higher rates. You add a lot of bits and get almost nothing. Lowering the min quantizer to 2 seems to help, but not that much.

Now, about the codec flaws...:
1) I'm going to strongly support the request (already made) of a Registry hack to set the default bitrate (in kilobits please!). I use DVDx 2.2 which has no GUI to enter the value. And with an integer KB value you've only a granularity of 8192 bps, not really the best...
2) sadly, with 2-pass I got oversized files up to 3% for 40-seconds clips, and even 0.5-0.6% on a 6-minutes clip. This is too high compared to XviD/DivX, and not so good for fitting a movie into a CD.
3) 2nd pass best quality is quite slow (got 7 fps on a P4 2.4GHz with VOB decoding/resize overhead, and 9-10 from uncompressed source)
4) where is the "General" profile you talk about in your site?

Also, I found a badframes.stt file in VDub directory, generated by version 6.1.0.2 (got a lot of badframes.stt with 6.0.9.2, now only one... guess we're improving :))

Overall, I think that VP6 is a very good product.
If On2 manages to improve it (oversize/interface especially), it may even be excellent.
I look forward to newer releases.

[slavickas]

>oversizing
i maybe wrong but i think 1KB in vdub = 8000bits, not 8192 (it's 2.4% difference)

[Shandra]

:up: Cool, going to check it out tomorrow ;)

And again - I seem to have missed it within the guide, the threads, etc. How do you predict the File Size? If someone can give me a helpfull slab on my heads back and feel that it is out of place here (as me the blind one is asking for the obvious) PM me please ;)

[Sirber]

Originally posted by netchris
Anyway for me vp6 is slightly better than real 10 :devil: :devil: :devil: ;) :D

Both are very similar (to me) at ~500kbps and I don't know which one I prefer :(

[Tommy Carrot]

Bug-report. :D

I think i've found the culprit behind the quality fluctuations. There are periodical "spikes" in the bitrate graph, in other words, some frames are much lesser compressed than the rest without any reason, even when i set constant quantizer. Here is what i'm talking about:

http://www.freeweb.hu/videodigitalizalas/carrot/rc_error1.png http://www.freeweb.hu/videodigitalizalas/carrot/rc_error2.png

The motion is rather constant in these scenes, so these "spikes" shouldn't be there.

Anyone else experienced this problem? I think this can be the reason behind the rate-control problems.

I've win98SE system and athlon xp 1700+, if that helps.

[Sirber]

I've seen taht too, so I thought the small one were b-frames... :confused:

[Tommy Carrot]

Yea, but vp6 doesn't have b-frames, so this must be a bug.

[SpaceV]

in some old post, one of the ON2 guys admitted recreating
the slow playback isue with HDTV resolutions.
Is that fixed in this release?
Thanks.

[Bulletproof]

Originally posted by Tommy Carrot
Bug-report. :D

I think i've found the culprit behind the quality fluctuations. There are periodical "spikes" in the bitrate graph, in other words, some frames are much lesser compressed than the rest without any reason, even when i set constant quantizer. Here is what i'm talking about:

http://www.freeweb.hu/videodigitalizalas/carrot/rc_error1.png http://www.freeweb.hu/videodigitalizalas/carrot/rc_error2.png

The motion is rather constant in these scenes, so these "spikes" shouldn't be there.

Anyone else experienced this problem? I think this can be the reason behind the rate-control problems.

I've win98SE system and athlon xp 1700+, if that helps.

Those are probably missed keyframes, I sent On2Tech a file with many missed keyframes using the newest version of the codec, but he hasn't said much about it.. Or it could be the temporal/spatial resamplers doing that, not positive though..

[Tommy Carrot]

Originally posted by Bulletproof
Those are probably missed keyframes, I sent On2Tech a file with many missed keyframes using the newest version of the codec, but he hasn't said much about it.. Or it could be the temporal/spatial resamplers doing that, not positive though..

No, those are not scene-changes, those frames shouldn't be any larger than the rest. They are quantized less than the other frames, without any reason, that's why they are larger.

[On2Tech]

Originally posted by Tommy Carrot
No, those are not scene-changes, those frames shouldn't be any larger than the rest. They are quantized less than the other frames, without any reason, that's why they are larger.

They are a special kind of frame that we intentionally encode at a higher datarate in order to bring back some detail otherwise lost to sub pixel motion estimation. It generally gets placed in slow motion scenes or pans.

Do you find the effect disturbing when the clip plays? I know single stepping through the frames it becomes evident.

[Tommy Carrot]

Originally posted by On2Tech
They are a special kind of frame that we intentionally encode at a higher datarate in order to bring back some detail otherwise lost to sub pixel motion estimation. It generally gets placed in slow motion scenes or pans.

Do you find the effect disturbing when the clip plays? I know single stepping through the frames it becomes evident.

Well, i would prefer without it. :) The quality jumpings are quite noticable during playback imo, and it can distract me. If you really think it benefits the quality i think it should be a selectable option.

[Sirber]

quality jump is a very big problem. YOu should keep the same quality between 2 key frames, or else it's very anoying to the eye.

[Doom9]

Do you find the effect disturbing when the clip plays? I know single stepping through the frames it becomes evident.I haven't noted anything like it in the latest codec comparison but I do recall that especially XviD exhibited such an effect in the SPR test (the night scene) and it bugged the crap out of me. The last comparison was the first one where there was no "quality readjustment" (that's how I call it) effect was readily visible anymore.

[Shinobu]

Yes please don't use this, or make a configurable version of it ^^.
I'm often work with constant quality or quantitser on all codecs, but when i do that in vp6 the quality jumping make me mad, it's too bad because without that vp6 would probably be my archivage codec (at least for movies ^^)....

++

[unmei]

first off, i can't really give a differenciated opinion about the new release. I just wanted to let you know i'm really pleased with VP6, but i have not yet done serious tests. Just from some test encodings (not carefully comparative) i found VP made a HUGE jump in quality since VP3 which i last tried. I think it now easily compares to RV (not tried 10 so far, but 9EHQ) and this along with the comfy vfw interface that RVx lacks so much it's become my codec of choice when i ever have a reason not to use Xvid (low bitrates..). However, something i experienced worried me a bit. I made a transode from a really crappy 320x180 hi compressed MOV (sorenson) which i fed through avisynth with fluxsmooth(insanelystrong) and other HC denoising. This i fed to both xvid and vp6 and let them outmax (i targeted for 800kbit). xvid outmaxed to 323 kbit/s, vp6 to 603 kbit/s. A huge difference, but even weirder as the xvid looks like a 'perfect' encode while Vp6 clearly shows compression artifacts. My only explanation is Vp6 cannot reach really high quality regardless of how many bits it consumes (it looks good in original size, but when in fullscreen (== 4x zoom) the effect is clearly visible). Anyway, I don't think it's a real issue the way it presented to me, since no-one should actually outmax a codec at such small resolution, it's much more intelligent to use bigger picture size. But the fact that it cannot reach high quality levels is still something to keep in mind.

[Kef71]

Originally posted by On2Tech
They are a special kind of frame that we intentionally encode at a higher datarate in order to bring back some detail otherwise lost to sub pixel motion estimation. It generally gets placed in slow motion scenes or pans.

Do you find the effect disturbing when the clip plays? I know single stepping through the frames it becomes evident.

Heh, I actually though it was a bug! ;)

Personally I find it quite disturbing. Especially when there is a lot of detail in the picture.

I’m not really sure using higher quantizers so you can use spikes of lower quantizers is better than to use a medium quantizer all the way if you understand what I mean. IMHO the whole experience just gets more ” flickering”… ;)

Kef

[justin]

couldn't you get better quality by taking out those spikes and making huger keyframes and bigger b frames??

[Sirber]

there is no bframes in VP6 :o

[DevilsChild]

I've done quite a bit of testing with the latest VP6 on DVD sources and haven't had any trouble with quality "fluctuations" or "flickering". Quite frankly, I don't know what everyone is complaining about. This codec is ROCK SOLID.

A huge difference, but even weirder as the xvid looks like a 'perfect' encode while Vp6 clearly shows compression artifacts. My only explanation is Vp6 cannot reach really high quality regardless of how many bits it consumes (it looks good in original size, but when in fullscreen (== 4x zoom) the effect is clearly visible).
[...]
But the fact that it cannot reach high quality levels is still something to keep in mind.

Sorry, but from my experience VP6 in 2 passes is perfectly capable of reaching high quality levels at high bitrates. Compression artifacts are much less pronounced compared to XviD or DivX, and become almost unnoticeable at even medium bitrates. I'm routinely getting superior quality at resolutions of 704x and higher.

[Tommy Carrot]

DevilsChild, good for you. ;) I wasn't able to get better results with VP6 than with XviD so far, but i guess i did something wrong.

[Kef71]

Originally posted by DevilsChild
I've done quite a bit of testing with the latest VP6 on DVD sources and haven't had any trouble with quality "fluctuations" or "flickering". Quite frankly, I don't know what everyone is complaining about. This codec is ROCK SOLID.

Sorry. I have to disagree with you here. I have encoded several movies where the quality "flickering" is very noticeable. In fact, so far I haven’t been able to encode a VP6 movie without these "issues". I'm happy you like VP6. I do too. But still, there are some things that could be improved. :D

Sorry, but from my experience VP6 in 2 passes is perfectly capable of reaching high quality levels at high bitrates. Compression artifacts are much less pronounced compared to XviD or DivX, and become almost unnoticeable at even medium bitrates. I'm routinely getting superior quality at resolutions of 704x and higher.

Can't comment on that. I never use 2-pass modes because IMHO they are a waste of time. A good 1-pass mode is just as important, especially when you are streaming over the internet etc and need CBR. I also find it very disturbing that a codec wastes so many bits on making the hi-motion scenes look good and makes the low-motion scenes look like crap. I'm sure there is a use for a 2-pass mode, but in my opinion it is extremely overrated. :cool:

Kef

[Kef71]

Originally posted by Tommy Carrot
DevilsChild, good for you. ;) I wasn't able to get better results with VP6 than with XviD so far, but i guess i did something wrong.

I believe it is a matter of personal taste. In my opinion XVID does deliver better results (sharper image) using higher bit rates (1 Mbps+) but I do like the smoothness and the natural "feeling" of VP6 as well. IMHO XVID & VP6 is the best codec’s out there. In low bit rates there's no contest between XVID and VP6. VP6 rules so much it almost hurts.

To be honest, I haven't made many tests with Real Video (The "other" low bit rate format), mainly because I have no "freedom" of choosing what container I use. But hey, you never know... Maybe I will become a RV10 fan after all... I just wish Real would release a proper VFW codec...

Kef

[Sirber]

RV10 uses variable frame rate, which VFW doesn't support.

[Kef71]

Originally posted by Sirber
RV10 uses variable frame rate, which VFW doesn't support.

And RV9 as well?

[Sirber]

RV9 (EHQ) = RV10

same thing

[Maximus_G]

Originally posted by Sirber
RV10 uses variable frame rate, which VFW doesn't support.
Well, VP6 is VFW-codec AND it uses variable frame rate (temporal resampling).

[eltoder]

I've also did some testing with the lastest VP6 in two-pass mode and had no problems. The quality is great. At 600 kbps it beats XVID down to the ground, mostly because it does not have ringing artifacts on the edges, that I find really annoying (though it seems this is caused by postprocessing). At 400 kbps it still looks reasonable for HVS rip. But at 1+ mbps it really seems to look less detailed than XVID. I'll look closer on this though.

-Eugene

[Sirber]

temporal resampling is frame dropping, not variable framerate. IIRC, RealVideo frames are stored as a timestamp, not x/s.

[edit]

Sorry, but here is not the place to talk about RealVideo, there is enough thread for that. :sly: