VP6.2 Release [Archive] - Page 2

echo

1st March 2004, 20:09

Hi everyone. I've found that you can play back matroska's with VP6 and subtitles using mplayer. You can get a copy for windows from a link at their site: http://mplayerhq.hu
You just have to copy vp6vfw.dll to your codecs folder and add something like this to your codecs.conf file:

videocodec vp6
info "On2 VP6"
status working
fourcc VP62,VP61,VP60
driver vfwex
dll "vp6vfw.dll"
out YUY2
out BGR32,BGR24

I just modified the vp5 settings to apply to vp6 :) It works just fine from what I've seen so far and even matroska aspect ratio support works OK. ;) I guess it can be used under *nix as well but I have only tried it under windows.

regards
echo

SeeMoreDigital

1st March 2004, 20:26

Well I have to say I like this new version.

My 1pass CBR encodes look very good when using the new VP62 profile. But I still don't think that 2pass VBR is working as well as it could.

I apologise if this has already been mentioned, as I must confess, I have not read 'all' the posts.

Cheers

kilg0r3

2nd March 2004, 10:23

Yes Il ike it too, but I start to hate my slow computer (Athlon 1800+), which makes 4-7 fps max on a 688*312 clip.

@On2Tech

Q1: Is there any headroom for speed improvements?

Q2: How much does the hq mode (39MB) differ from the considerably faster 'good quality fast encoding' (40MB) mode? The file difference indicates that they differ no too much. But I cannot say whether the file size of two otherwise identical one-pass encodes is a good indicator?

thx, kilg0r3

On2Tech

2nd March 2004, 15:11

@kilg0r3

A1: There's always room for more optimization, but I can't make any forecasts or promises at the moment.

A2: best quality gives a pretty sizable jump in quality. File size alone isn't really a good indicator, you have to look at the results to judge. When quality really matters, the investment in time is worth it.

Rest assured we'll keep you all posted on any improvements!

wata

3rd March 2004, 13:00

have a vp62 encode foreign movie with subs, subs fail to display
anyone manage to display subtitles with media player classic

kilg0r3

5th March 2004, 09:18

I am an VP6 novice. However the few 'tests' I have done so far confirm sagitaire's much more sofisticated ones.

I am mostly interested in high-bitrate encodes around 2000kbits. So I encoded the Harry Potter 2 trailer with various one-pass settings. The hardest was fixed quant 6 with a sharpness value of 10. It might be for my bad eyes, but even at these settings I couldn't spot too many artifacts, that had not already been in the source. And, you could still see the grain. And if I say grain, I dont mean the in famous flickering or wavering of the background.

Furthermore the image produced by VP6 is the most visually pleasing/natural. There seem to be no color problems or any visible (at least for me ;)) loss of detail. Of course, the crispness increses at lower quants. It is just that the compression works in a way which seems to be very in line with the (/my) human visual system.

After all the preaise I'd like to reformulate my question from above. My problem is that my machine encodes resized but unfiltered letter box content with only about 7 fps. So I consider using the 'good quality fast encoding' mode. I'd be perfectly willing to sacrifice some disc space in order to compensate for the lesser quality of the compression algo. However I would also be very happy if somebody could provide me with a rule of thumb how much this would be.

@ On2tech
If I encode at fixed quant but diferent quality modes, will I get the same quality but only different filesizes, or, will it be a mixture of these?

Cheers, kilg0r3

bond

5th March 2004, 11:37

guys, lets stop this discussion now!
or transfer it to the another thread...you wish, we play ;)

plz continue the kilobit discussion here (http://forum.doom9.org/showthread.php?s=&threadid=72116)

On2Tech

5th March 2004, 12:39

Originally posted by kilg0r3
And now to something completely different ...

After all the preaise I'd like to reformulate my question from above. My problem is that my machine encodes resized but unfiltered letter box content with only about 7 fps. So I consider using the 'good quality fast encoding' mode. I'd be perfectly willing to sacrifice some disc space in order to compensate for the lesser quality of the compression algo. However I would also be very happy if somebody could provide me with a rule of thumb how much this would be.

@ On2tech
If I encode at fixed quant but diferent quality modes, will I get the same quality but only different filesizes, or, will it be a mixture of these?

Cheers, kilg0r3

Thanks for the kind words. Unfortunately I don't think there is a general rule of thumb for how much best quality helps over good quality. It is very material dependent. Sometimes best quality is only a little better than good quality, other times its a lot better. That being said I would say its rare for best quality to make more than a 10% file size difference. For an entire movies I doubt you'd see even a 5% difference.

For reference: the main difference between good quality and best quality comes in the way motion vectors are found and then whether or not they are used. Best Quality spends more time looking for motion vectors, and then does a better job of deciding whether or not the use of that motion vector is worth the cost of transmitting that motion vector.

kilg0r3

5th March 2004, 13:51

Hi On2Tech,
Sorry for being unclear in my questions and/or slow in grasping what you mean. I will retry a last time and then shut up. Promise.

The scenario: I start from one source of which I will create two encodes (a) and (b). The only difference between the encodes will be that (a) will use 'Good Quality Fast Encoding' while (b) uses 'Best Quality'. Else, (a) and (b) will be done using the following settings.

General: Number of Passes: 1 - End Usage: Local - Noise Red.: Off - Sharpness: 5 - Auto Keyframe: Yes - May Frames between Keys: 120

Advanced: Undershoot: 100 - Adjust Quant: No - Temp. Resamp.: Off - Spat. Resamp - Off

My assumptions: (1) The file produced by (a) will be larger than the one produced by (b). (2) The image quality of the two clips (a) and (b) will be next to identical.

My question: Are these assumptions correct?

Your answer: Yes_[ ] No_[ ] :)

On2Tech

5th March 2004, 15:03

Originally posted by kilg0r3
Hi On2Tech,
Sorry for being unclear in my questions and/or slow in grasping what you mean. I will retry a last time and then shut up. Promise.

The scenario: I start from one source of which I will create two encodes (a) and (b). The only difference between the encodes will be that (a) will use 'Good Quality Fast Encoding' while (b) uses 'Best Quality'. Else, (a) and (b) will be done using the following settings.

General: Number of Passes: 1 - End Usage: Local - Noise Red.: Off - Sharpness: 5 - Auto Keyframe: Yes - May Frames between Keys: 120

Advanced: Undershoot: 100 - Adjust Quant: No - Temp. Resamp.: Off - Spat. Resamp - Off

My assumptions: (1) The file produced by (a) will be larger than the one produced by (b). (2) The image quality of the two clips (a) and (b) will be next to identical.

My question: Are these assumptions correct?

Your answer: Yes_[ ] No_[x] :)

My answer: Yes_[x] No_[x]

but I feel I need to qualify this answer. Even in the ideal case constant quantizer doesn't equal constant quality unless the quantizer is effectively lossless. Instead constant quantizer is merely a cap on the worst the quality can be.

There is another thing in vp6 which messes up the works. That is in effect when you specify fixed quantizer every type of frame is given exactly a fixed quantizer. But there are 3 types of frames. The choice of frame type is automatically determined by the compressor based on statistics gained in part from the quality of the motion vectors found. As such Best Quality can ( and frequently does ) produce more of one type of frame then good quality. The quality can then vary in a way not intended by your choice of a fixed quantizer.

The result is that it is possible that when using best compression you can see any of these situations:

1. b is bigger and a good amount better than a
2. b is smaller and better than a
3. b is a good amount smaller and about the same quality as a
4. b and a are about the same size and quality

I answered your question yes however because I think 2 and 3 are the most frequent results.

kilg0r3

5th March 2004, 18:56

Thank you very much for you immense patience.

Cheers. kilg0r3

virus

6th March 2004, 18:39

@On2Tech

I'd like to draw your attention (you may already know of that) on that link posted by thegeby in the "H.264 status and definition" thread:

http://www.ebu.ch/tech_texts/tech_text_d96-2003.pdf

looks like European broadcasters are not happy with current AVC licensing. Don't you think that VP6 may successfully enter the competition for a place in the DVB standard for HDTV content? You offer source code licensing and have a codec with low decoding requirements (=cheap hardware)... have you ever take any step in this direction?
From a technically standpoint, what can prevent VP6 from being the next-gen solution for HD (lack of bidirectional coding, low efficiency with "near" keyframes, too high saturation bitrate...)?

Mug Funky

8th March 2004, 15:42

hmm. sorry to put a dampner on things, but i can't seem to be able to download this...

says "cannot write to the database" after i fill in the fields and stuff.

if it's a 50% compression improvement over VP6.1 like it says, then i'd luurrve to try it out!

garbel

8th March 2004, 20:48

Try this :)

http://www.on2.com/software/vp6_vfw_codec.exe

garbel

C0mPr355

8th March 2004, 21:34

for those of you having trouble displaying subtitles with your VP62 encoded content:

1)Please make sure to use the newest media player classic, as it has its own subtitle loading abilities.
2)Your subtitle files are in srt format or at least saved into that format. If you need help saving subtitles from VOBs get subrip.
3)Make sure that your player is in direct show format and it is using VMR9 or VMR7 format (under the options>output screen).

This is the easiest way to play back subtitles with a VP62 encoded file.

let me know if you guys have any trouble with this.

Blight

9th March 2004, 05:26

On2Tech:
As seen on this thread and also been notified by some of my users, The VP6 DirectShow decoder does not play "nice" with the DirectVobSub filter. This is very unfortunate as it cripples the format to some effect.

Currently, the fix would be to block/unregister the VP6 decoder and fall back to the VCM -> DirectShow bridge filter, but this has a CPU hit and is far from optimal.

Any plans on fixing this issue in a later release?

C0mPr355

9th March 2004, 18:02

yes I agree on the cpu hit.

ATM we have VP62 directshow filter which seems to be having some problems displaying subtitles (I am still looking into solving this) which hits the cpu for about 48-67%.

With the VP62 blocked and the VM filter enabled subs can be displayed but the cpu hit is much larger ~50-75%.

We will look into this and see what can be done.

Thanks for the point out.

C0mPr355

10th March 2004, 17:55

ok for some bizaar reason media player classic likes it when I apply the filters in a specific way.

I have my overrides listed as so:

VP6 Decompressor (defaults with highest quality)
DirectVobSub (auto-loading version) *everything pretty much default as well.

when using this combination I seem to be able to lower the cpu hit by about 5%...now obviously this isnt something that is the greatest but it can help when using the vp6 decoder along with subtitles.

VSFilter 2.32 was used for DirectVobSub

test this out and let me know what you think. I also used the specify directory with the auto loading of vobsub to get the subtitles to auto display without having to load them with the "Load subtitle" option.

rom1v

13th March 2004, 12:34

Originally posted by C0mPr355
yes I agree on the cpu hit.

ATM we have VP62 directshow filter which seems to be having some problems displaying subtitles (I am still looking into solving this) which hits the cpu for about 48-67%.

With the VP62 blocked and the VM filter enabled subs can be displayed but the cpu hit is much larger ~50-75%.

We will look into this and see what can be done.

Thanks for the point out.
http://forum.doom9.org/showthread.php?s=&threadid=72531
I don't understand how to fix it...

rom1v

14th March 2004, 13:16

Originally posted by C0mPr355
for those of you having trouble displaying subtitles with your VP62 encoded content:

1)Please make sure to use the newest media player classic, as it has its own subtitle loading abilities.
2)Your subtitle files are in srt format or at least saved into that format. If you need help saving subtitles from VOBs get subrip.
3)Make sure that your player is in direct show format and it is using VMR9 or VMR7 format (under the options>output screen).

This is the easiest way to play back subtitles with a VP62 encoded file.

let me know if you guys have any trouble with this.
Yes it works, but VMR7~9 is not the best way to view movies...
It would be great if vobsub would read subtitles...

rom1v

14th March 2004, 14:18

Just install FFDShow and active "RAW: all suported", and subtitles are loaded with vobsub when video is VP6 :D

sayneboody

15th March 2004, 02:18

I have a clip where the camera is panning down a large tree and it looks really blurry, the color isn't right, and the leafs leave a blurry smudge as the camera is panning down. I tried encoding it at a high bitrate 3000 kbs and still get blurry leafs that smudges downward (although quit less). I use the HQ settings from your user mannual and I left sharpness at 0, since my source video is DVD quality.

Any way to fix my problem?

Sirber

15th March 2004, 04:07

can you post a screenshot?

kilg0r3

15th March 2004, 09:37

Originally posted by sayneboody
I left sharpness at 0, since my source video is DVD quality.Any way to fix my problem? Probabely using sharpness 5-7 would do some good :)

sayneboody

15th March 2004, 20:16

Here are some screenshots

http://cwebb.greatnow.com/index.htm

I tried setting the sharpness at 7 and it did help somewhat, but the clip still had some smudges even at 6000 kbps (sorry I don't have the screenshots). Is this normal even at 6000kbps? Please excuse me, I am a newbie.

Tommy Carrot

15th March 2004, 23:42

@Sayneboody: My guess it's a rate-control problem. That scene seems to be very complex, so i think the codec overcompresses it to avoid the huge bitrate-boost.

Sirber

16th March 2004, 03:36

is the scene rapid or slow?

C0mPr355

16th March 2004, 17:56

sayne if you could pm me and let me know all your settings I could help you out much better.

rom1v I have not done much with matroska filters and the like...so I guess I shall take up that flag now :)

Manao

13th April 2004, 22:32

Hi,

I tried recently this codec, on low bitrate scenario ( 300 kbits, 352x288, DV shootage ), in order to make 1-pass encoding ( 1-pass best quality, no noise reduction except my avisynth's script, sharpness 5, progressive, no automatic keyframes, and defaults for the rest ), and I got a problem : the video ( which lasted 2h38 ) had only one keyframe ( the first frame ), which makes random search quite difficult :sly:

I guess it's due to the combination low bitrate / no automatic ketframes, but I think at least keyframes should be inserted at scenechanges, to be on the safe side.

Sirber

13th April 2004, 22:41

Originally posted by Manao
I guess it's due to the combination low bitrate / no automatic ketframes, but I think at least keyframes should be inserted at scenechanges, to be on the safe side. That is automatic keyframe :p

CiNcH

8th May 2004, 15:35

Hi,
sorry for bringing this topic up again but I have a question on VP6.2's Heightened Sharpness Profile. What does this Sharpness slider actually do? Does it apply a sharpness filter (like the Noise Reduction filter) or different motion estimation or DCT criteria?

Buck746

14th May 2004, 02:36

It seems to determine how much priority goes to the sharper freq's. It seems to work by commpressing lower freq's more. Is that at all helpfull?

On2Tech

14th May 2004, 14:03

Originally posted by CiNcH
Hi,
sorry for bringing this topic up again but I have a question on VP6.2's Heightened Sharpness Profile. What does this Sharpness slider actually do? Does it apply a sharpness filter (like the Noise Reduction filter) or different motion estimation or DCT criteria?

It basically alters the filters used for sub pixel motion compensation. The higher the sharpness value the sharper the filters.

virus

14th May 2004, 14:16

hi On2Tech, glad to see you're back ;)

Any updates planned for the 3-months-old version 6.2.0.10?

cheers :)
virus

the Mad Duke

3rd July 2004, 17:40

Originally posted by virus
hi On2Tech, glad to see you're back ;)

Any updates planned for the 3-months-old version 6.2.0.10?

cheers :)
virus

I just listend to an interview with the CEO of On2 and he said that VP7 will be released in the not too distant future. The engineers at On2 have done a really great job on each upgrade of their codec from VP3 through VP6 so this should something to watch out for. He didn't say if it would be free for personal use or not. Also an upgraded java player will be released next week.

virus

3rd July 2004, 18:15

Originally posted by the Mad Duke
I just listend to an interview with the CEO of On2 and he said that VP7 will be released in the not too distant future.well, thx for the info - and welcome to the forum, too :)

<rant>
It's good to have someone posting infos, since the devs simply don't want to answer to anything. Yeah, I know they cannot tell too much, but revealing a "secret" (about how the sharpness mode works, see On2Tech's post above) and then refusing to comment a bit on the development of their own codec looks somewhat silly to me. Almost 5 months of silence. Not that I pretend anything, but they should definitely look at how the various devs here (for XviD, RV, Nero etc) talk with the userbase and bring attention to their codecs. I don't see the same level of interest from the VP6 devs. Needless to say, I'm no longer motivated in doing beta-testing for them, that's for sure ;)
</rant>

ah, and of course, a non-"free for personal use" VP7 does not interest me, at all.

cheers
virus (in bad mood today... ;))

bond

3rd July 2004, 18:30

Originally posted by the Mad Duke
I just listend to an interview with the CEO of On2 and he said that VP7 will be released in the not too distant future.will it be h.264 compliant? or yet another not compliant h.264 clone?

/rant :D

stephanV

3rd July 2004, 18:42

im sorry, but that was almost an insult... what do you mean with clone?

theres a world beyond MPEG4 and "compliancy"... maybe you should try it once.

bond

3rd July 2004, 18:48

weee :D

you shouldnt take things too serious

stephanV

3rd July 2004, 18:56

uhm... yeah right...

Sirber

4th July 2004, 01:09

For sure VP7 will be 30% better then VP6 at 50% less bitrate. They have a pretty strong marketing department :D Just kidding :D

Any specs yet?

On2Tech

7th July 2004, 12:58

Originally posted by virus

<rant>
It's good to have someone posting infos, since the devs simply don't want to answer to anything. Yeah, I know they cannot tell too much, but revealing a "secret" (about how the sharpness mode works, see On2Tech's post above) and then refusing to comment a bit on the development of their own codec looks somewhat silly to me. Almost 5 months of silence. Not that I pretend anything, but they should definitely look at how the various devs here (for XviD, RV, Nero etc) talk with the userbase and bring attention to their codecs. I don't see the same level of interest from the VP6 devs. Needless to say, I'm no longer motivated in doing beta-testing for them, that's for sure ;)
</rant>

ah, and of course, a non-"free for personal use" VP7 does not interest me, at all.

cheers
virus (in bad mood today... ;))

I'm sorry you are frustrated with us. Your assistance has proven extremely valuable to us and I hope that you'll help us test drive our new products when they are ready.

I continue to watch this forum closely and comment when I can. I assure you that it is our intention to help our user-base achieve the best results that they can with our codecs and that when I see questions I can answer I do so.

Unfortunately, I can't talk about future products/releases or what we plan to do with them.