WMV9 AP Information Discussion [Archive]

zambelli

17th May 2006, 04:42

I selected the Windows Media Video 9 codec in Nic's WM Encoder. And I was using CBR encoding. Perhaps that is the reason...but shouldn't there be an improvement regardless?
I don't know what your content was. If you're encoding, say, progressive 512x384 content, the difference between Main and Advanced Profile might be negligible in that scenario.

See if you can find an old video encoded with the old WMV9 codec. Encode the same source with the WMP11 codecs and see if you can spot the difference.

And my testing methodology was simple. Take a clip, encode it in WMV 9 AP. Then Encode it in WMV 9 (those were the names of the "profiles"). Watch each, see which one looks better. I also compared it to an xvid and x264 and nd avc encode of it. The latter three were better (by a large margin) than both WMV 9 AP and normal...
Were the AVC encodes also done with default settings? Make sure you're not comparing apples to oranges. There are plenty of WMV9 AP settings to tweak too.

Sirber

17th May 2006, 12:38

Will there be a CLI encoder?

dohcmark8

17th May 2006, 13:18

The WMP11 beta is out... http://www.microsoft.com/windows/windowsmedia/player/11/default.aspx

GmorG McRoth

17th May 2006, 13:52

dohcmark8

17th May 2006, 14:04

I's there any GUI that whuld allow me to play with those advanced encoder settings? Editing registry like outlined here (http://www.microsoft.com/windows/windowsmedia/howto/articles/codecadvancedsettings.aspx) is not the most user friendly way.

Not that I've seen yet. But I think WME Studio Edition will... dont quote me on that though.. im just guessing.

zambelli

17th May 2006, 19:50

Will there be a CLI encoder?
I'm working on an update to wmcmd.vbs to be used with WME9.

zambelli

17th May 2006, 19:53

Sirber

17th May 2006, 20:01

I'm working on an update to wmcmd.vbs to be used with WME9.I ment, any command line console win32 text mode application to encode with VC1? With AVS as input and WMV as output? ;)

siddharthagandhi

17th May 2006, 21:13

"See if you can find an old video encoded with the old WMV9 codec. Encode the same source with the WMP11 codecs and see if you can spot the difference."

I have a clip. I compared it. There was almost no difference or improvement. I heard that the official beta has improvement over the leaked one. I'm testing that as we speak.

And yes, AVC encodes were done with default settings. Even an AVC encode with a bitrate 500 lower looked better. Now that's saying something.

zambelli

17th May 2006, 22:30

I ment, any command line console win32 text mode application to encode with VC1? With AVS as input and WMV as output? ;)
Microsoft is not planning on releasing one, if that's your question.
A command-line encoding app should be easy to implement. Just look at Nic's encoder implementation. It's pretty much based entirely on the CUncompAVIToWMV sample from the SDK.

McoreD

18th May 2006, 02:53

I's there any GUI that whuld allow me to play with those advanced encoder settings? Editing registry like outlined here (http://www.microsoft.com/windows/windowsmedia/howto/articles/codecadvancedsettings.aspx) is not the most user friendly way.

This is something I made for myself while testing WMV9AP.
http://microsoftuse.temp.powweb.com/mcored/downloads/dvd2wmvassistant-1.3beta1-bin.zip

You need .NET 2.0. Open the EXE, go to WMVA Profile Editor and you are good to go from there.

@zambelli,
I was wondering why you call "WMV9 Advanced Profile (WVC1) video codec BETA" in the first post. Isn't WVC1 finalized yet? or are you just calling it beta because WMP11 is beta?

stax76

18th May 2006, 05:59

Am I right to assume that WMP won't install the automation interfaces so I better wait until updated interfaces are released with the encoder before I consider starting to code.

CruNcher

18th May 2006, 08:02

Ok for all of you who don't wan't to install WMP 11 but only the updated codecs here we go ;)

1. Download the WMP 11 beta
2. Extract the main .exe
3. now inside the extracted folder you gonna find wmfdist11.exe, execute it.
4. Start WME9 and enjoy the Advanced Profile ;)

@zambelli

Microsoft is also actively working on making a standalone WMV9 AP codec package available for download to users who want to test the beta codec but don't want to install the Player beta.

you allready did but you allways try to hold it back whyever M$ likes todo that. In the end everyone here knows why you do it you wan't the installations of WMP 11 as high as possible :P but not everyone is a avg user, so if you lie to us this way you won't make friends here zambelli.

here is a WVC1 example made by Remedy (most probably encoded by the Microsoft Studio staff as they partners) 720p@60 fps 15 Mbps
http://www.alanwake.com/movies.html

Test Results:

First test results show that the new Main Profile is much better now then it was before, especialy in detail preservation but the difference to the new WVC1 i couldn't see that @ 8 mbps for now, gonna try some higher bitrates later source was 1920x1080p@25 fps hmm i can see a slight textural difference looks that the new MP@HL did better here then AP @ this Bitrate. To be Continued...

GodofaGap

18th May 2006, 09:58

Q: Does the new Advanced Profile codec have a different FourCC?
A: Yes, it does. In order to differentiate it from SP/MP codec (FourCC: WMV3) and the non-VC1-compliant AP codec (FourCC: WMVA), and avoid issues with legacy decoders, the VC-1 compliant WMV9 Advanced Profile codec uses a new FourCC code: WVC1.

Are these 2 fourcc's for VC1 (WMV3 for SP/MP and WVC1 for AP IIUC) private to Microsoft, or can we expect future implementations of VC1 by other companies to use these fourcc's as well? (Or perhaps both?)

edit: bah i messed up the fourcc for AP...

zambelli

18th May 2006, 10:50

you allready did but you allways try to hold it back whyever M$ likes todo that. In the end everyone here knows why you do it you wan't the installations of WMP 11 as high as possible :P but not everyone is a avg user,
?!?!?
Riiiiight. "Forcing" hundred Doom9 users to install WMP11 in order to get WMV9 AP is really going to boost Microsoft's market penetration. I'll go tell the marketing people that the MTV Urge service is a huge waste of time - because clearly the real draw is WMV9 AP.

Cruncher, I don't underestimate anybody on this forum. I know very well you can extract the WMF SDK runtime from the install package. However, you're entirely speculating about our reasons for not endorsing that install scenario. Conspiracy theories are entertaining, but that's not how businesses are run.
so if you lie to us this way you won't make friends here zambelli.
What has gotten into you? I'm trying to be informative and help you guys get the latest codec, and that makes me a liar? If I say "we're working on making a standalone codec download available", I mean exactly that. The world doesn't revolve around your personal interests. There are Windows Server 2003 customers out there who want WMV9 AP for their streaming needs - WMP11 isn't targeted at W2K3, so we need to figure out an alternative way to provide them with a codec update. These decisions are made for a reason and I assure you they're not the least bit diabolic as you perceive them to be.

P.S. I'm not here to "make friends" and turn people over to Microsoft's products. I've been a member since the forum was started, before I ever worked for MS. The fact that I can now act as a liason between you guys and MS is something that I hope you can appreciate.

First test results show that the new Main Profile is much better now then it was before, especialy in detail preservation but the difference to the new WVC1 i couldn't see that @ 8 mbps for now, gonna try some higher bitrates later source was 1920x1080p@25 fps hmm i can see a slight textural difference looks that the new MP@HL did better here then AP @ this Bitrate. To be Continued...
I am very interested in hearing such feedback. That's exactly why the beta is being released publicly.

zambelli

18th May 2006, 11:02

Are these 2 fourcc's for VC1 (WMV3 for SP/MP and WMVA for AP IIUC) private to Microsoft, or can we expect future implementations of VC1 by other companies to use these fourcc's as well? (Or perhaps both?)
My guess is that 3rd party encoder implementations will want to use the same FourCCs if they're not planning on providing their own decoder implementation. That way they can leverage Microsoft's decoders. However, if they want to force people to use their decoder implementations too, they might use a different 4cc code to avoid filter conflict.

It's difficult to regulate these things. Remember how Canopus had a different 4cc for their DV video, even though it was perfectly compliant with every other DV decoder on the market? We'll see how things turn out for VC-1. Hopefully it won't be a FourCC mess. :)

GodofaGap

18th May 2006, 13:37

Ok, I feel I am doing something wrong...

I used the clip from this (http://forum.doom9.org/showthread.php?t=90784&highlight=quality+challenge
) infamous ( ;) ) thread, but comparing PSNR scores WMV9 AP appears to do worse than DivX3 (900 kbps) :confused:

I've used Nic's wmenc to encode the file, but the fourcc seems to be WVC1 so that should be ok... visual checking the impression is not that bad I think, but on that partcilar clip in some scenes there is still some blockiness compared to x264.

Zambelli, can you confirm that Nic's wmencoder is using the correct codec?

Sharktooth

18th May 2006, 14:09

you can compare vc1 to real or xvid...
comparing it with x264 is a bit unfair.

GodofaGap

18th May 2006, 15:21

Ok, I feel I am doing something wrong...

I forgot to tick VBR. :stupid:

Still WMV9 AP is way behind DivX or XviD in PSNR score (about 0.5 dB) and even normal WMV9 is a bit better... I still feel I am doing something wrong.

you can compare vc1 to real or xvid...
comparing it with x264 is a bit unfair.
I don't know why it would be unfair, but for me it is certainly relevant. Encoding and decoding speed is certainly not any faster than with x264/CoreAVC.

dohcmark8

18th May 2006, 16:15

Try enabling 'Compression Optimization Type' to a value of '1' in the registry...

http://www.microsoft.com/windows/windowsmedia/howto/articles/codecadvancedsettings.aspx

edit: Yeah wow... that one registry setting really makes the videos look MUCH better. Credit goes to 'mpgxsvcd' at the AVS Forum for the find.

GodofaGap

18th May 2006, 16:43

Will try. Don't think regedit is a suitable GUI for an encoder though. :S

LotharZ

18th May 2006, 16:57

At Softpedia can be found the new runtimes if you dont want the player.

Im not sure if its a beta or final version, links dont come from MS.

Windows Media Format Runtime 11
http://www.softpedia.com/get/Multimedia/Video/Codec-Packs-Video-Codecs/Windows-Media-Format-Runtime-11.shtml

Latexxx

18th May 2006, 18:13

Actually the new player is pretty decent. It's faster than WMP 10 and 9 and looks better.

bond

18th May 2006, 18:55

is the encoder dmo useable in graphedit? eg with .avs input?

edit:
and did anyone manage to register the encoder/decoder dmos without needing to run the drm infected installer?

trbarry

18th May 2006, 21:14

and did anyone manage to register the encoder/decoder dmos without needing to run the drm infected installer?

What??

What does drm infected mean? Some non-obvious side effects of this?

- Tom

GodofaGap

18th May 2006, 21:39

I believe "DRM infected" is another way of saying they don't like Microsoft. You cannot install DRM. You can install a player that can handles files that are protected with DRM. I wouldn't know what the practical downside of this is though. I must say I'm positively surprised with the interface of WMP11. After WMP6 my favorite of the WMP series.

Try enabling 'Compression Optimization Type' to a value of '1' in the registry...
It hurted PSNR quite badly on the HP2 trailer (~0.4 dB). Also interesting to see that using b-frames causes a decoding delay.

zambelli

18th May 2006, 21:49

What?? What does drm infected mean? Some non-obvious side effects of this?
Bond is just being overly dramatic, as usual. What's "DRM infected" anyway? The player supports playback of DRM protected content, yes. How exactly does that translate to "DRM infected"? Nobody is forcing anybody to use DRM for their own files.

dohcmark8

18th May 2006, 21:58

Ok. So yesterday I encoded a 720p video in one hour... today with compression optimization on, the same video took five hours... that is unacceptable IMO for a 4 minute video.

I really want to use this as my only HD archiving format... but man even the slowest x264 encodes of this vid only took two hours. :(

So Microsoft... Is there a secret hidden setting or something?!?

zambelli

18th May 2006, 22:02

Try enabling 'Compression Optimization Type' to a value of '1' in the registry...

edit: Yeah wow... that one registry setting really makes the videos look MUCH better. Credit goes to 'mpgxsvcd' at the AVS Forum for the find.
As the documentation mentions, that preset controls the following parameters:
Motion Search Level = 1
Motion Match Method = -1
Dquant Option = 2
Motion Search Range = -1
NumBFrames = 1
Motion Vector Cost Method = 1
Force LoopFilter = 1
It also fixes the v_performance level to 60 (if you're familiar with wmcmd.vbs syntax).

You can probably further improve quality by tweaking some of these parameters. For example, Motion Search Level does a better job when set to 2 - but it's a lot slower, too.

siddharthagandhi

18th May 2006, 22:11

Yes, but you can't cut codecs breaks. In the real world, it is survival of the fittest. And if it can't beat the other codecs, then its bye-bye. But for a codec as popular as this one its going to take a little more than a slight decrease in quality for the codec to die out. Darwin's theories man....

bond

18th May 2006, 22:12

the installer (not only the wmp11 one, but also the codec one cruncher pointed to) installs m$' drm system, thats what i call "drm infected"

as i am someone who wants to test the encoder and decoder (and not the drm) i asked if someone has manged to install the encoder and decoder without the other stuff around it?

zambelli

18th May 2006, 22:14

It hurted PSNR quite badly on the HP2 trailer (~0.4 dB). Also interesting to see that using b-frames causes a decoding delay.
Yeah, PSNR is likely to decrease when Dquant optimization and Loopfilter are used. This preset yields better perceived quality because it involves some filtering, but PSNR might not reflect that. Try SSIM instead.

Good parameters to experiment with:
NumBFrames
Dquant Option
Dquant Strength
Force LoopFilter
Motion Search Level
Motion Search Range
Motion Match Method
In my experience, some combination of these parameters will usually give you the best quality for a given video.

If you're using wmcmd.vbs for encoding, varying -v_performance between 60, 80 and 100 will give you different quality results too. However, expect 80 and 100 to be slower than the default 60.

dohcmark8

18th May 2006, 22:16

Retracted (See post #39): Ok... I believe we were taken in by hype.. I really wanted to like this codec.. but I just cant... I havent formed a final opinion yet and and am still experimenting with this...

Using 'Compression Optimization Type' on a video makes the decoder requirement increase thus making the video stutter in WMP11 for me. Keeping in mind I can play a VC1 1080p24 encode at the same cpu usage as a Comp Opt enabled 720p vid but the 1080p vid plays smoothly. It also makes encoding ridiculously slow... even half the speed of a HQ x264 encode.

Same exact video:

Compression Optimization Type = 1 on a video increases decoding requirments on the CPU by 30%. Thus making it right in line with non-CoreAVC AVC decoders.

zambelli

18th May 2006, 22:22

the installer (not only the wmp11 one, but also the codec one cruncher pointed to) installs m$' drm system, thats what i call "drm infected"
Oh no! DRM! What's it going to do, steal your wallet while you sleep?
If you're not purchasing DRM protected music/movies from online stores - which I assume you're not - I really don't see how just having WM DRM binaries on your system affects you in any way.
But to each their own, I guess. Have you thought about investing into one of these: http://zapatopi.net/afdb? :sly:

bond

18th May 2006, 22:37

noone knows what it does or will do somewhen ;)

apart from that
- i dont need it
- i dont want it
- i dont give a damn about it
so why should i install it?

anyways is there a way to register the decoder/encoder without running the installer?

zambelli

18th May 2006, 23:32

anyways is there a way to register the decoder/encoder without running the installer?
Yes, theoretically, but in practice it could get tricky because the DLLs are named differently, have different dependencies and the encoders are grouped differently within them than they were in v10. Doing this manually could be an arduous task. I've never really attempted to do this... like I said, we have a standalone WMV9 AP DMO that fits into the v10 world a lot better and we're trying to release that soon.

dohcmark8

18th May 2006, 23:45

Ok I wanna retract my previous statement. The video stuttering was a problem on my PC not the codec. Sorry MS.

Anyway... now I'm just playing with it, so I can get some good quality but a bit faster speed wise than with the compression optimization option enabled.

Hey Zambelli, what setting in the comp opt decreases the speed the most?

zambelli

19th May 2006, 00:46

Hey Zambelli, what setting in the comp opt decreases the speed the most?
Motion Search Level and Motion Search Range, most likely.

dohcmark8

19th May 2006, 00:56

Motion Search Level and Motion Search Range, most likely.

Thanks.

siddharthagandhi

19th May 2006, 01:55

I did another comparison. There is a pretty big improvement (btw I used Windows Media Encoder 9 with the AP 9 codec) compared to the leaked version, but the tweaking of settings could also account for that. Nonetheless, it does not beat Xvid, x264, or ND AVC, although I haven't compared it to RealVideo. It was just a brief and quick test, so I don't have any shots or PSNR's to tell you. But the visual quality was evident. I tweaked most of the settings, except for the weird ones. I went for maximum quality without regard to the encoding time. And the encode was LOOOOOOOOOOOOOOONNNNNNNNNNNNGGGGGGGGG....

If anyone gets different results post them and say your tweaks and stuff....maybe I'm doing something wrong.

But its not a totally bad codec. My results show that it is definetly is better than VP7 and similar "other" codecs, and probably it is better than Divx 6.2.2, although the contest is close.

zambelli

19th May 2006, 03:21

Zambelli, can you confirm that Nic's wmencoder is using the correct codec?
Yes, just verified that.

dohcmark8

19th May 2006, 05:46

So... any chance the codec will get some optimizations to speed up encoding time.. by say official WMP11 release?

Blight

19th May 2006, 06:57

zambelli:
Do these new codecs improve on the WMF -> DirectShow bridging filters? There are several bugs in the bridging filters that prevent WMV from being properly played under DirectShow.

I'm referring to:
1. Inability to change language on a multi-language WMV file (graph freezes or audio mutes).
2. WM streaming (under DirectShow) just cutting out all by itself, while streaming using the WMF SDK doesn't disconnect.

zambelli

19th May 2006, 07:25

So... any chance the codec will get some optimizations to speed up encoding time.. by say official WMP11 release?
It's possible. If you have specific data, such as comparisons with other encoders - that's always useful.

zambelli

19th May 2006, 07:28

Do these new codecs improve on the WMF -> DirectShow bridging filters? There are several bugs in the bridging filters that prevent WMV from being properly played under DirectShow.
The codecs don't, but the new WMF SDK runtime does update those filters. If you're willing to install WMP11 beta, give it a try. If those two issues still repro, send me a PM and I'll open bugs on the WMF SDK.

scilli

19th May 2006, 10:47

I have never really encoded in WMV.. I recently stoped encoding to Mpeg2 for DVD, and am now just using x264. But since your the only MS insider I've ever seen in a forum, and you don't come off as member of an "evil empire", I'll give it a try Zambelli.:)

GodofaGap

19th May 2006, 11:02

Thanks Zambelli for your help so far.

A new question: does anyone know how to enter a negative DWORD value in the registry? (like -1)

foxyshadis

19th May 2006, 11:26

Subtract it from 2^32. ie, -1 is 4294967295.

LordDethstar

19th May 2006, 12:29

I have done a small test encoding a short stretch of anime at 600 kbps using main and ap (two encodes). I have also encoded the same stretch using XVID, X264, QuEnc (MPEG2), VP7, Theora (libtheora 1.0alpha5), and the Windows Media 9 VCM. I have come to the following conclusions.

WM9AP:
The only encoder that I could actually use without crashing when feeding it my AviSynth script was Nic's Encoder. The results were mixed. The picture was nice and clear, but degraded if there was a lot of motion. The encoder also appeared to drop frames, which is unacceptable (is there a way to turn it of). Encoding speed was not the best, but acceptable.

WM9:
As above, except that the picture quality way worse. Same frame-dropping problem.

XVID:
Better that WM9AP's worse-quality frames, but not as good as its best frames. The quality was consistent. 600 kbps per second is apparently not enough for XVID and anime. Used default setting + VHQ4 (+for b-frames) + GMC + Cartoon Mode + Adaptive Quant + Chrome Optimization.

X264:
Provided the best quality of all, providing a good picture at all times. This comparison is probably not fair, as I did not use all the quality setting available: --ref 2 --mixed-refs --bframes 3 --b-pyramid --weightb --subme 5 --analyse all --8x8dct. Still it beat all other codecs. Encoding speed was the best of all codecs.

QuEnc:
MPEG2 is not a good codec for low bitrates. No comment.

VP7:
The quality is almost as good as X264, except when text fades in or out (blocking). The encoding speed was 2-4 fps on my Athlon 64 x2 3800, which is unacceptable.

Theora:
I am still trying to decide if this really looks better that the MPEG2 version.

WM9VCM:
This one actually did not skip any frames. The quality was close to that of XVID.

At 600 kbps, the only codec I would use is X264. VP7 has a decent quality as well, but encoding is unacceptably slow.

WM9AP did have a decent picture quality (as far as I could see), and the cpu load while playing is back was undetectable on my maching. However, unless the problem with the frame skipping is solved, I won't ever consider it.

dohcmark8

19th May 2006, 13:08

I've heard of people having frame-drop problems with WME9... I personally have never experienced that problem and I've been playing with WM9 for months now.

I'd try feeding WME the video and audio seperately... I always do this and have never had a problem.

It's possible. If you have specific data, such as comparisons with other encoders - that's always useful.

Sure. As soon as I finish this 4-day long (WME ETA) VC-1 720p encode I have going here... :)

Another side effect of comp opt.. is that it gives slightly smaller file sizes. :)

Sagittaire

19th May 2006, 13:33

benwaggoner

19th May 2006, 17:13

I'm a Microsoft guy now too (Program Manager for Video Encoding in the Professional Content Group. We're the folks working on the new Windows Media Encoder Studio Edition, which is coming out soon, and has a GUI for many of these parameters).

Speaking of good quality/speed combos, I'm no Zambelli, but for HD DVD style encodes, we've had good luck with Motion Search Level of -2 (Macroblock adaptive with integer chroma search). At little faster than the default -1, and can actually look a little better in some cases.

1 B-frame is good for high bitrate HD, 2 B-frames as data rates get lower.

The new codec is also 4-way multithreaded, so a dual-dual is the performance sweet spot. Real cores are a lot better than hyperthreading for encoding.

If you're doing 1-pass CBR encoding (which really is only for real-time encoding these days), use the Lookahead parameter, which improves rate control. Maxing it out to 16 can be quite helpful for real-time broadcasting (albeit with increased latency).

And again, I can't emphasize enough that the codec is optimized for real-world video quality, not PSNR. Use your eyes on comparisions to other codecs.

Sagittaire

19th May 2006, 18:08

@ benwaggoner

wellcome on this forum ... ;-)

"And again, I can't emphasize enough that the codec is optimized for real-world video quality, not PSNR. Use your eyes on comparisions to other codecs."

Quality Comparison
VC-1 achieves clearly superior quality to MPEG-2 at comparable bit rates, and has been judged superior to H.264 in several independent studies.

Measuring the quality of a video codec is not easy, because the reconstructed image is not meant to be identical to the original. Ideally, only information that is perceptually irrelevant will be lost in the compression/decompression process, but what counts as "irrelevant" depends on the viewer's subjective response.

One useful objective metric is the peak signal-to-noise ratio (PSNR) plotted against bit rate. PSNR is the ratio between the maximum value of a signal (255 for 8-bit video) and the quantization noise. A higher PSNR indicates a less noisy signal. For any codec, PSNR is expected to increase at higher bit rates, because higher bit rates translate to less aggressive compression. Thus, a graph that plots PSNR against bit rate shows the performance of the codec over a range of compression settings.

In Microsoft's own internal tests, VC-1 performs 2 to 3 times better than MPEG-2. In other words, to achieve a given PSNR, MPEG-2 requires a bit rate up to 3 times higher than VC-1. These results were measured using both low-motion and high-motion video sequences. Microsoft also compared VC-1 with H.264 and found that both codecs have comparable performance when PSNR is plotted against bit rate.

The final arbiter of codec quality is the subjective appearance of the decoded video. In subjective tests, the perceived quality of VC-1 equals or exceeds that of H.264. The DVD Forum conducted tests in the winter of 2002 to select codecs for the next-generation red-laser HD-DVD. Viewers from Hollywood film studios and major consumer electronics companies rated video clips on a scale of 1 to 5 for resolution, noise, and overall impression. Multiple codecs were tested, including MPEG-2, VC-1, H.264, and MPEG-4 Advanced Simple Profile. The baselines against which the codecs were compared were D5 masters and D-VHS (24 Mbps). During the tests, viewers were not told which codec was used to encode each of the clips.

On all three measures (resolution, noise, and overall impression), the quality of VC-1 was judged closest to the original D5 master. By comparison, the H.264 codec was rated as comparable only to MPEG-2 on two of the three measures (resolution and overall impression), and was rated somewhat worse than VC-1 on noise.

VC-1 has performed well in other independent subjective quality tests:
DV Magazine found VC-1 to be superior to both MPEG-2 and MPEG-4.
TANDBERG Television found VC-1 produces significantly better quality than MPEG-2 and comparable quality to H.264. These results were presented at the 2003 International Broadcasting Convention (IBC).
C'T Magazine, Germany's premiere audio-video magazine, compared various codecs, including VC-1, H.264, and MPEG-2, and selected VC-1 as producing the best subjective and objective quality for high-definition (HD) video.
The European Broadcasting Union (EBU) found VC-1 had the most consistent quality in tests that compared VC-1, RealMedia V9, the Envivio MPEG-4 encoder, and the Apple MPEG-4 encoder.

... and it's false ... ;)

1) For metric H264 done better result and by far (for all resolution, source and bitrate).

2) for my eyes H264 done better result and by far (for all resolution, source and particulary in low bitrate scenario).

3) Good MPEG2 encoder done very good result too and certainely not with 2-3 ratio bitrate like in high bitrate HD-DVD scenario.

pwimmer

19th May 2006, 18:11

I've encoded a few samples in WVC1 and WMV3 but I can't see any improvement over the old WMV3 codec. Of course I also tried the 'Compression Optimization Type' Registry option.

The only difference i've noticed: The new one takes 2x - 4x as long to encode.

There must be something wrong with the WVC1 codec included in the WMP 11 Beta (at least I hope there is and that a better version will follow).

I expected that the new WVC1 codec will catch up to H.264, since Microsoft always claimed quality of VC-1 and H.264 is approximately the same. But it didn't.

benwaggoner

19th May 2006, 18:33

I've encoded a few samples in WVC1 and WMV3 but I can't see any improvement over the old WMV3 codec. Of course I also tried the 'Compression Optimization Type' Registry option.
Can you share some details of your scenario, or a link to the encoded file? We might be able to recommend some different settings. One thing I really like is to use integer chroma search, which isn't on by default in the above.

When we ship Studio Edition, that'll change a lot of stuff, since it'll be the first mainstream tool for an advanced codec that will offer scene-by-scene reencoding. So you can tweak encoding parameters for particular shots (not a lot of options in the really-soon-now beta, but a lot more coming in 1.0).

There must be something wrong with the WVC1 codec included in the WMP 11 Beta (at least I hope there is and that a better version will follow).

Well, WMP11 is a beta, and clearly the video codec wasn't Job #1 for the release :). Help us find places and scenarios where we aren't doing as well as you'd like or expect, and let's see if we can help tweak settings, or find something we can share with our developers.

While codec standards can be completed, codec implementations never are :).

Nic

19th May 2006, 19:23

I'll be updating my Windows Media Encoder app so you can adjust the Advanced Profile Registry settings from it. I'll also do some testing and see how things compare.

Thanks for the updates Zambelli,

-Nic

GodofaGap

19th May 2006, 20:07

dohcmark8

19th May 2006, 20:16

Can zambelli or benwaggoner provide a list of the default settings in the codec? Sometimes when I add a registry value, I find that it doesn't change anything at all just because I used the default setting. :)

I believe removing everything but the number of threads settings puts the codec back on default settings.

zambelli

19th May 2006, 20:20

Well, WMP11 is a beta, and clearly the video codec wasn't Job #1 for the release :). Help us find places and scenarios where we aren't doing as well as you'd like or expect, and let's see if we can help tweak settings, or find something we can share with our developers.
While codec standards can be completed, codec implementations never are :).
That's worth repeating. We're expecting VC-1 to be around for a long time (just look at MPEG-2 and its DVD legacy) so even though we might call some version later this year "final", development will continue for quite some time. With your help, we can make it better.

And welcome to the forums, Ben! :)

zambelli

19th May 2006, 20:27

This is something I made for myself while testing WMV9AP.
http://microsoftuse.temp.powweb.com/mcored/downloads/dvd2wmvassistant-1.3beta1-bin.zip
I somehow missed your post first time around. :)
That's a cool tool. Make sure you take a look at the updated list of Advanced Profile settings - there's now almost twice as many settings to configure in WVC1.

I was wondering why you call "WMV9 Advanced Profile (WVC1) video codec BETA" in the first post. Isn't WVC1 finalized yet? or are you just calling it beta because WMP11 is beta?
The spec is finalized, but the implementation can continue being improved forever. The "beta" coincides nicely with WMP11 being beta, but also conveys where we currently are in the development process - doing pretty well, but still not 100% confident.

zambelli

19th May 2006, 20:29

I believe removing everything but the number of threads settings puts the codec back on default settings.
Yeah, that's a good method of setting defaults. In most cases "0" is the same as the default value, although there are some exceptions but I think they are noted in that web page.

pwimmer

19th May 2006, 20:31

Can you share some details of your scenario, or a link to the encoded file? We might be able to recommend some different settings. One thing I really like is to use integer chroma search, which isn't on by default in the above.

Here are some sample files.

Nero H264 6 Mbps: http://www.3dtv.at/Temp/Test_1440x576_H264_6Mbps.mpg

WMV3 (using old WME9 codec) 6 Mbps: http://www.3dtv.at/Temp/Test_1440x576_Old_WMV3_6Mbps.wmv

WMV3 (using old WME9 codec) 8 Mbps: http://www.3dtv.at/Temp/Test_1440x576_Old_WMV3_8Mbps.wmv

WVC1 6 Mbps: http://www.3dtv.at/Temp/Test_1440x576_WVC1_6Mbps.wmv

The old WMV3 and WVC1 6 Mbps files basically look the same. The 8 Mbps file is better, but still worse than the H264 file. My guess is that Windows Media requires a 50% higher bitrate compared to H264 at this resolution. I didn't do any tests at low resolutions and bit rates.

Anyway, I'll continue to use Windows Media for two reasons:
I can't play the H264 file on my Pentium D 2 x 3 GHz without dropped frames and I need DRM.

Here's a screenshot of my Registry settings.

http://www.3dtv.at/Temp/WVC1Reg.gif

Sagittaire

19th May 2006, 20:39

Anyway, I'll continue to use Windows Media for two reasons:
I can't play the H264 file on my Pentium D 2 x 3 GHz without dropped frames and I need DRM.

My little sempron 2.1 Ghz can play 1920*1088 H264 sample with coreAVC decoder ... ;-)

pwimmer

19th May 2006, 20:44

My little sempron 2.1 Ghz can play 1920*1088 H264 sample with coreAVC decoder ... ;-)

Unfortunately I use a different one. That's reason 3 to use Windows Media: no decoder hassle :-) It's preinstalled on every Windows system and it's free.

Doom9

19th May 2006, 20:50

That's reason 3 to use Windows Media: no decoder hassle :-) It's preinstalled on every Windows system and it's free.Actually.. it isn't. A WinXP, even with SP2 slipstreamed lacks the WMV9 runtime, let alone the VC-1 one. Of course there's automatic download.. and it's obvious that Microsoft will not add alternative codecs to the autodownload list, but you don't have the codecs right out of the box - and in a corporate environment where software installation is centrally controlled, windows media player won't be able to download and install those codecs.

I'm pretty sure you'll find plenty of evil for AVC (DRM ;).. didn't Nero have something, too? But if you think you need DRM, ask your customers.. they know better..

dohcmark8

19th May 2006, 21:39

Here's a informative thread about VC-1 encoding over at AVS... http://www.avsforum.com/avs-vb/showthread.php?t=678185&highlight=wmp11

GodofaGap

19th May 2006, 22:46

I believe removing everything but the number of threads settings puts the codec back on default settings.
Obviously, but this is not what I asked. :)

If I add "Force B Frame Delta QP" with value 2 to the registry, it will effectively change nothing because 2 is the default value anyway. So I'd like to know what the defaults are, because if I'm trying to use a setting I don't want to run in the defaults each time. It saves me encoding time. :)

For some the defaults are given, but not for others.

dohcmark8

19th May 2006, 22:48

Obviously, but this is not what I asked. :)

If I add "Force B Frame Delta QP" with value 2 to the registry, it will effectively change nothing because 2 is the default value anyway. So I'd like to know what the defaults are, because if I'm trying to use a setting I don't want to run in the defaults each time. It saves me encoding time. :)

For some the defaults are given, but not for others.

Have you taken a look at this? http://www.microsoft.com/windows/windowsmedia/howto/articles/codecadvancedsettings.aspx

GodofaGap

19th May 2006, 22:50

Yes I have, not for every setting the defaults are given there.

benwaggoner

19th May 2006, 22:51

Actually.. it isn't. A WinXP, even with SP2 slipstreamed lacks the WMV9 runtime, let alone the VC-1 one. Of course there's automatic download.. and it's obvious that Microsoft will not add alternative codecs to the autodownload list, but you don't have the codecs right out of the box - and in a corporate environment where software installation is centrally controlled, windows media player won't be able to download and install those codecs.
Actually SP2 includes the upgrade to WMP9. WVC1 is available on the codec download server for WMP9 and WMP10, and of course is included with WMP11.

I'm pretty sure you'll find plenty of evil for AVC (DRM ;).. didn't Nero have something, too? But if you think you need DRM, ask your customers.. they know better..
Any HD DVD or Blu-ray player uses DRM, on top of any codec. Codecs and DRM are orthogonal. And DRM is a value-add in a player - non DRM'ed content works fine on a DRM-capable player, but DRM'ed content will not work on a player without DRM.

zambelli

19th May 2006, 22:54

Actually.. it isn't. A WinXP, even with SP2 slipstreamed lacks the WMV9 runtime, let alone the VC-1 one.
That's not true. SP2 includes WMP10 (or WMP9 - can't remember exactly), so all new XP systems (retail and preinstalled) have WMV9 decoders installed.

You are right about Advanced Profile though: the WVC1 decoder is not preinstalled and automatic codec download will be triggered the first time you play a WVC1 file in WMP.

Of course there's automatic download.. and it's obvious that Microsoft will not add alternative codecs to the autodownload list
To be clear, the reasons behind Microsoft not hosting any 3rd party codecs are: security, support and security. :) We used to host some 3rd party codecs such as ACELP.net and Voxware on the codec server back in the day, but it quickly became evident that we couldn't support code that wasn't ours (and in the case of old codecs such as Voxware and Indeo - wasn't even truly supported by their legal owners) and then during the SP2 security push it became even less desirable to keep pushing untested (potentially exploitable) 3rd party code to customers.

Having said all that... I think the Microsoft codec server system could be corrected and improved with a WHQL-like process in place. Codecs could be treated like drivers - it seems logical. But I'm not in a position to make those type of calls.

Doom9

19th May 2006, 23:19

SP2 includes WMP10 (or WMP9 - can't remember exactly),it's WMP9. It seems I can play WMV9 content without having installed anything extra but I was quite sure of being unable to play WMV stuff on another machine (where I didn't even have admin privileges and was thus unable to install the codec). But it has been a while so it might've been an SP1 machine.

As far as installing codec without the player goes, you can download the player and open the exe with winrar (and I suppose also other zippers), extract wmf11dist.exe, then install just that.

dohcmark8

19th May 2006, 23:31

So whats the difference between the macroblock and luma settings in the 'Motion Search Level' ?

benwaggoner

20th May 2006, 00:48

So whats the difference between the macroblock and luma settings in the 'Motion Search Level' ?
Macroblock adapative is what you want, pretty much anytime you get that option in the codec. It basically means "figure out the right option for each macroblock." Zambelli will need to remind me what this particular macroblock mode does :).

Motion Search Macroblock Adaptive is very cool - it dynamically selects the optimum motion search mode among the bottom three choices (shorter vectors are more efficient with lower motion, since they user fewer bits).

dohcmark8

20th May 2006, 02:41

I found that if you disable compression optimization and manually add the settings yourself... that encoding time will be MUCH MUCH faster... but you will get slightly bigger file sizes.

So that means another question for the Microsoft guys... what else does compression optimization do?

I mean..

VC1 @ QVBR 89
---------------------

1 - 00:00:00:51 - Average 3257.58 Kbps - Comp Type Optimization = 0

2 - 00:00:03:21 - Average 2448.02 Kbps - Comp Type Optimization = 1

3 - 00:00:01:08 - Average 3186.2 Kbps - Comp Type Optimization = 0, 2 bframes, loopfilter on, motion search level -2, motion method -1, duant opt 2, motion search range -1, motion vector cost method = 1, macroblock mode cost method = 1, motion vector coding method = 3

Any suggestions are far as getting the bitrate down with #3?

benwaggoner

20th May 2006, 03:22

3 - 00:00:01:08 - Average 3186.2 Kbps - Comp Type Optimization = 0, 2 bframes, loopfilter on, motion search level -2, motion method -1, duant opt 2, motion search range -1, motion vector cost method = 1, macroblock mode cost method = 1, motion vector coding method = 3

Any suggestions are far as getting the bitrate down with #3?
You could try a higher DQuant strength

If the content is noisy, Force Median can help (it excludes noise when calculating motion vectors)

Force Overlap=1
Motion Match Method=-1
Perceptual Option=1

dohcmark8

20th May 2006, 03:26

You could try a higher DQuant strength

If the content is noisy, Force Median can help (it excludes noise when calculating motion vectors)

Force Overlap=1
Motion Match Method=-1
Perceptual Option=1

Thanks.

When I used DQuant 3 the bitrate actually shot up by about 500 Kbps.

The content is actually a 720p live music performance. So it should be clean other than the blocking artifacts which I deal with in DGDecode via cpu=4.

benwaggoner

20th May 2006, 03:35

Thanks.

When I used DQuant 3 the bitrate actually shot up by about 500 Kbps.

DQuant 3 is the DQuant LEVEL. I mean increasing the STRENGTH (1-3, or 0 for codec chooses). That said, it's something you'd experiment with, not always use.

Also, if Zambelli and I disagree, do what Zambelli says :).

dohcmark8

20th May 2006, 03:57

DQuant 3 is the DQuant LEVEL. I mean increasing the STRENGTH (1-3, or 0 for codec chooses). That said, it's something you'd experiment with, not always use.

Also, if Zambelli and I disagree, do what Zambelli says :).

Thanks. I encoded.. the video encoded in under an hour compared to 5 hours with comp opt on.. :)

It looks awesome. Now to start archiving everything with VC-1. :)

Anyway, I think I found a bug or something... If I encode with default settings and comp opt off and play in WMP11 the video plays back perfectly... but if I encode with my above posted settings or with just comp opt on.. the video will playback in WMP11 but the video will start stuttering like its skipping frames, the audio still playsback however, but if I pause the video and then hit play it will be fine again. My CPU useage during this is 30-50% so I have enough power. I wonder if anyone else has experience this. The troublesome videos however playback perfectly fine in Zoom Player and VLC.

It could just be a hardware problem. I'm gonna try re-installing my video card drivers and WMP11.

zambelli

20th May 2006, 04:00

As far as installing codec without the player goes, you can download the player and open the exe with winrar (and I suppose also other zippers), extract wmf11dist.exe, then install just that.
Sure, but like I already told Cruncher, WMP10 (or 9) running with WMF11 bits is an unsupported and untested scenario. I advise against it, but I can't control what people do. Proceed at your own risk.

zambelli

20th May 2006, 04:07

3 - 00:00:01:08 - Average 3186.2 Kbps - Comp Type Optimization = 0, 2 bframes, loopfilter on, motion search level -2, motion method -1, duant opt 2, motion search range -1, motion vector cost method = 1, macroblock mode cost method = 1, motion vector coding method = 3
If you're setting all these overrides manually, you might as well exclude Comp Type Optimization entirely. The one thing that COT does that you can't override with another registry settings is the encoder complexity. It sets it to 3 (on a scale from 0 to 5). If you want to do the same thing, you have to either use the -v_performance parameter in wmcmd.vbs, or move the performance slider in WME9 Options.

dohcmark8

20th May 2006, 04:54

If you're setting all these overrides manually, you might as well exclude Comp Type Optimization entirely. The one thing that COT does that you can't override with another registry settings is the encoder complexity. It sets it to 3 (on a scale from 0 to 5). If you want to do the same thing, you have to either use the -v_performance parameter in wmcmd.vbs, or move the performance slider in WME9 Options.

Thanks. I'd imagine any codec is slow at inception but slowly speeds up as development continues (XViD, x264, etc).

I hope we get to see VC-1 playback implemented for the Xbox 360 and then I will be all set.

But with my new settings, its more than fast enough for me, and the quality is awesome. And I wanna say good job to the guys at Microsoft for this great codec. :)

Now if we could just get Mr. Waggoner to release the Studio Edition Beta so we could have something else to play with... Can we expect it this month?

benwaggoner

20th May 2006, 08:08

I hope we get to see VC-1 playback implemented for the Xbox 360 and then I will be all set.
Main Profile already works!

But with my new settings, its more than fast enough for me, and the quality is awesome. And I wanna say good job to the guys at Microsoft for this great codec. :)
Gratifying to hear (not that I can personally take any credit for the codec).

Now if we could just get Mr. Waggoner to release the Studio Edition Beta so we could have something else to play with... Can we expect it this month?
We're working hard to get the beta out as soon as possible. The good news is that due to the delays we're going to be able to get some more fixes and tweaks in there, so it won't be just a late release, but a later, better release.

FYI, we've also committed to adding support for Main Profile and VC-1 elementary stream encoding this year post-beta. But the initial release will be Advanced Profile only.

dohcmark8

20th May 2006, 09:23

easyfab

20th May 2006, 10:28

Is it also possible to play WVC1 on smartphone?
I have a svp c600 (WM2005) and try to play a WVC1 file but the fourcc "WVC1" is not recognize.
Is it possible to do an WMP codec update?

bond

20th May 2006, 14:04

Yes, theoretically, but in practice it could get tricky because the DLLs are named differently, have different dependencies and the encoders are grouped differently within them than they were in v10. Doing this manually could be an arduous task. I've never really attempted to do this... like I said, we have a standalone WMV9 AP DMO that fits into the v10 world a lot better and we're trying to release that soon.cool, thanks

will there also be a way to encode with vc-1 into .mpg/.ts as smpte defines it (eg for hddvd/bluray compliance...)

Guest

20th May 2006, 15:08

will there also be a way to encode with vc-1 into .mpg/.ts as smpte defines it (eg for hddvd/bluray compliance...) Of course! See here:

http://neuron2.net/misc/rp227.pdf

I have the other VC-1 specs, too, if you are interested.

bratao

20th May 2006, 15:55

Really Too bad that it dont run in Windows 2003...

I have a *legal* Windows 2003 , using like a workstation, but i cant install because its Windows Xp only.. this is insane..
The win2003 has a 98% complatible kernel with xp and the m$ out restritions

benwaggoner

20th May 2006, 19:12

Sorry if I keep asking so many questions.

Which setting would increase decoding requirements?

[HKEY_CURRENT_USER\Software\Microsoft\Scrunch\WMVideo]
"Force NumThreads"=dword:00000001
"Force LoopFilter"=dword:00000001
"Motion Search Level"=dword:ffffff62
"Motion Match Method"=dword:ffffff61
"Dquant Option"=dword:00000002
"Motion Search Range"=dword:ffffff61
"NumBFrames"=dword:00000002
"Motion Vector Cost Method"=dword:00000001
"Macroblock Mode Cost Method"=dword:00000001
"Motion Vector Coding Method"=dword:00000000

Thanks.
You never have to aplogize for asking how to get better results out of our technology!

The Loop Filter definitley increases decoder load - it'll make the single biggest difference. Using B-frames will increase load a little bit. The others shouldn't have much decode impact.

Wilbert

20th May 2006, 20:16

Sure, but like I already told Cruncher, WMP10 (or 9) running with WMF11 bits is an unsupported and untested scenario. I advise against it, but I can't control what people do. Proceed at your own risk.
I'm a bit confused. I have W2K and use WMP6.4. Can I use those codecs?

Valeron

20th May 2006, 20:49

here's a screen shot from Nic's WMNicEnc 1.0 alpha

benwaggoner

20th May 2006, 22:29

Is it also possible to play WVC1 on smartphone?
I have a svp c600 (WM2005) and try to play a WVC1 file but the fourcc "WVC1" is not recognize.
Is it possible to do an WMP codec update?
You'd need a new version of the smartphone player.

Due to the decoder requirements of devices like phones, "WMV3" will probably remain dominant. In fact, PlaysForSure video is Simple profile only, although some devices will at least try to do Main decode.

Now, most of the encoding improvement are applicable to Main Profile, and some to Simple, so you'll get some gains there.

benwaggoner

20th May 2006, 22:34

I'm a bit confused. I have W2K and use WMP6.4. Can I use those codecs?
If it does work, it's certainly not supported. Decode should at least be possible in WMP9 for Win 2000, though.

Any particular reason you're not on XP yet? Fewer and fewer media apps support 2000 at this point.

LordDethstar

20th May 2006, 23:19

dohcmark8

21st May 2006, 00:07

I have now tried another encode using Nic's new 1.0alpha encoder app (since the WM9 Encoder does not appear to like my AVS scripts). This time I tried to encode without audio.

The quality was actually better than with my previous tests, and it appeared to be a viable codec, except that it still dropped frames during high-motion scenes (600 kbps for 640x480 anime should be enough).

Is there any way to get WM9 not to drop frames during encoding? It is the worst thing a codec can do, and no blocking is even nearly as bad as dropped frames.

You can get AVS scripts working in WME by choosing 'both file and device' in the sources menu.

Are you sure its the encoder dropping frames? I've never experienced the encoder dropping frames in WM9.

Try playing your video in another player and see what happens.

dohcmark8

21st May 2006, 01:24

Hey Microsoft.. I think I can confirm there is a issue with WMP stuttering (dropping frames) on VC-1 playback with WMP11. I tried it on three different systems (Athlon 3200+, P4 2.8, Centrino 1.76).

The videos will just not play back smoothly. Unless you hit pause then playback again then they will work for a few seconds. CPU usage is high (60-70% and sometimes shoots up to 97%) during playback. if I disable 'drop frames to maintain a/v sync' in WMP then the vids will work fine... but will lose sync on longer videos. However, on other players the videos will play fine.

All my videos I've encoded so far are 720p and if these cant be played back smoothly on 2.5ghz and up systems, then the codec is not worth using.

Maybe some decoder optimizations are in order?

I cant find where to file a bug report for the WMP11 beta and the VC-1 decoder. If you guys could look into this, it would be great. I wouldnt mind sending you a sample video if need be.

zambelli

21st May 2006, 04:53

Due to the decoder requirements of devices like phones, "WMV3" will probably remain dominant. In fact, PlaysForSure video is Simple profile only, although some devices will at least try to do Main decode.
Now, most of the encoding improvement are applicable to Main Profile, and some to Simple, so you'll get some gains there.
Right. The advantage of supporting Advanced Profile on a Smartphone or PocketPC is dubious. The best PocketPC displays only go up to 640x480 and the CPUs already have a hard time rendering anything above 1Mbps. Main Profile already handles that very well, so there's really no advantage in adding AP support to those devices at this time. I stress "at this time". WVC1 decoder is likely to eventually be added to WMP Mobile, but it's not like anybody will be playing 1080p video on Smartphones any time soon. :)

zambelli

21st May 2006, 04:59

Hey Microsoft.. I think I can confirm there is a issue with WMP stuttering (dropping frames) on VC-1 playback with WMP11. I tried it on three different systems (Athlon 3200+, P4 2.8, Centrino 1.76).
Are you using WMP11 beta, or did you install the WMF11 runtime on top of WMP10?

All my videos I've encoded so far are 720p and if these cant be played back smoothly on 2.5ghz and up systems, then the codec is not worth using.
I have no trouble playing back 720/24p videos on a 2.8GHz Celeron D, so I doubt it's a decoder problem. I think something else is causing the perf issue here.

If you guys could look into this, it would be great. I wouldnt mind sending you a sample video if need be.
Yeah, if we can't troubleshoot it successfully here, I'll PM you and you can send me a video sample.

zambelli

21st May 2006, 05:03

Here's something I forgot to mention, that you'll find interesting:

If you have WMP11/WMF11 beta installed (as opposed to, say, WME9 x64), most of the advanced registry settings will apply to Main Profile (WMV3) encoding, too. I don't remember which ones exactly, but try enabling something like "Compression Optimization Type" and encoding something in vanilla WMV9 - and see if you notice a change.

dohcmark8

21st May 2006, 09:06

Are you using WMP11 beta, or did you install the WMF11 runtime on top of WMP10?

I have no trouble playing back 720/24p videos on a 2.8GHz Celeron D, so I doubt it's a decoder problem. I think something else is causing the perf issue here.

Yeah, if we can't troubleshoot it successfully here, I'll PM you and you can send me a video sample.

Yup they are all using the WMP11 Beta.

bond

21st May 2006, 11:18

Of course! See here:

http://neuron2.net/misc/rp227.pdf

I have the other VC-1 specs, too, if you are interested.yep i know those (i have them too), i just wondered if microsoft will also support those specs?

dohcmark8

21st May 2006, 23:16

Yup they are all using the WMP11 Beta.

Zambelli - Yeah its not a hardware problem. I tried the videos on a 2.8 Ghz P4 laptop that has hardware acceleration for WMV and the videos still stuttered.

benwaggoner

21st May 2006, 23:20

yep i know those (i have them too), i just wondered if microsoft will also support those specs?
What kind of support are you looking for?

dohcmark8

21st May 2006, 23:56

It seems people at a Flight Simulator forum are having trouble playing back a trailer for FS:X that was released by Microsoft and encoded in VC-1 aswell.. the same stuttering issue I am having. So it's a issue with WMP11 it appears.

I'd imagine bug reports have been submitted.. if not could you guys atleast make sure the WMP11 team is aware of this?

Thanks.

Edit: here is a link to that trailer: http://simflight.com/modules.php?name=News&file=article&sid=8002

If others could download and see if they have the stuttering issue maybe?

That trailer barely plays in WMP11 for me, but in every other media player, it plays 100% smooth as butter with no dropped frames.

obieobieobie

22nd May 2006, 02:54

What kind of support are you looking for?
My guess is some kind of parser/splitter/muxer by Microsoft that can handle VC-1 in .ts and .mpg.

Correct me if I'm wrong. :)

bratao

22nd May 2006, 04:00

Well, i post but get any awnser..

Why not Win2003 support ???

benwaggoner

22nd May 2006, 06:39

Well, i post but get any awnser..

Why not Win2003 support ???
I didn't answer because I'm not part of that team, and don't want to speak for them.

But bear in mind that this is via a beta of a consumer media player app - not exactly the kind of thing that runs on servers normally :).

mpgxsvcd

22nd May 2006, 18:44

I can finally post! Man that 5 day wait was killing me.

I have re-encoded the Harry Potter clip with the WMV9 advanced. It does not look as good as the X264 versions though. Here are the settings I used. Maybe the perceptual option made it a little too soft. I think everything else was good though.

Dquant Option = 2
Force LoopFilter = 1
Force Video Scaling = 0
Lookahead = 16
Motion Match Method = ffffff62
Motion Search Level = ffffff61
Motion Search Range = ffffff62
Motion Vector Cost Method = 1
NumBFrames = 1
Perceptual Option = 1

WMV9 Advanced HP test file
http://www.yousendit.com/transfer.p...2AF6D0334FD2B62

original test files
http://forum.doom9.org/showthread.p...nge%3Cbr%20/%3E

Link to posts with more detail about the settings I used.
http://www.avsforum.com/avs-vb/showthread.php?t=678185&page=3

Any chance WME9 Pro will be out this week?

benwaggoner

22nd May 2006, 19:47

I can finally post! Man that 5 day wait was killing me.
Welcome!

I have re-encoded the Harry Potter clip with the WMV9 advanced. It does not look as good as the X264 versions though. Here are the settings I used. Maybe the perceptual option made it a little too soft. I think everything else was good though.

Dquant Option = 2
Force LoopFilter = 1
Force Video Scaling = 0
Lookahead = 16
Motion Match Method = ffffff62
Motion Search Level = ffffff61
Motion Search Range = ffffff62
Motion Vector Cost Method = 1
NumBFrames = 1
Perceptual Option = 1[/QUOTE]
Is that Motion Match Method=-2? I can't read hex :).
You might also try upping BFrames to 2.
I assume this is a 2-pass VBR peak limited? A longer and higher peak are also helpful.

Any chance WME9 Pro will be out this week?
I don't have a date yet, sorry.

zambelli

22nd May 2006, 20:30

My guess is some kind of parser/splitter/muxer by Microsoft that can handle VC-1 in .ts and .mpg.
Well, I'm not sure how much that SMPTE recommendation actually affects the PC platform. Note that the 2 additional documents are titled "Recommended Practice for Television". Placing VC-1 streams inside MPEG-TS makes sense in the context of broadcast TV and and HD-DVD/BluRay discs. However, on the PC it's not the most suitable format. VC-1, as a standard, is container independent so one can store it any format they desire.

Microsoft chooses ASF as the default container for its Windows Media codecs. Other companies implementing VC-1 might choose other containers and will probably supply the necessary filters for them, I'd imagine.

zambelli

22nd May 2006, 20:36

I didn't answer because I'm not part of that team, and don't want to speak for them.

But bear in mind that this is via a beta of a consumer media player app - not exactly the kind of thing that runs on servers normally :).
In addition... The standalone codec release will be available for XP and Server 2003 too.

zambelli

22nd May 2006, 21:32

It seems people at a Flight Simulator forum are having trouble playing back a trailer for FS:X that was released by Microsoft and encoded in VC-1 aswell.. the same stuttering issue I am having. So it's a issue with WMP11 it appears.

That trailer barely plays in WMP11 for me, but in every other media player, it plays 100% smooth as butter with no dropped frames.
I looked into this. This particular trailer is 720p at 60fps with a 384kbps 5.1 soundtrack. That's truly a difficult video to decode.

I tried it on a 3.0 GHz P4 (w HT) with both WMP10 and MPC. Neither player could play it any faster than 10 fps.
On a 3.2 GHz P4 (w HT) and WMP11, it played at about 40 fps, dropping a lot of frames. MPC yielded similar performance.
I then tried it on a 3.8GHz P4 Xeon with WMP11 and it rendered fairly smoothly but with a few dropped frames here and there. In general it achieved the full framerate of the video.
Finally, I tried it on the same 3.8GHz Xeon with DXVA decoding support enabled (not available publicly yet) and it rendered smoothly with no dropped frames.

Bottom line: For this particular video, it does seem like it's a decoder performance bottleneck. 720p60 is a lot of video data to process. I didn't see any significant difference between WMP10, WMP11 and MPC, so I don't think it's a player issue at all.
What's your hardware config? Have you hard more luck playing 720p 24fps videos?

lazyn00b

22nd May 2006, 23:38

Well, I'm not sure how much that SMPTE recommendation actually affects the PC platform. Note that the 2 additional documents are titled "Recommended Practice for Television". Placing VC-1 streams inside MPEG-TS makes sense in the context of broadcast TV and and HD-DVD/BluRay discs. However, on the PC it's not the most suitable format. VC-1, as a standard, is container independent so one can store it any format they desire.

Microsoft chooses ASF as the default container for its Windows Media codecs. Other companies implementing VC-1 might choose other containers and will probably supply the necessary filters for them, I'd imagine.

I think that one reason why some of us are interested in other containers for VC-1 is because we want to be able to retain AC3 audio without transcoding. Can ASF handle AC3? If so, is there a muxer/splitter available anywhere?

What I really want to be able to do is to keep VC-1 + AC3 movies on my PC in watchable form, and still be able to author them on HD DVD-9 or BD-9 if I need to.

zambelli

23rd May 2006, 00:05

I think that one reason why some of us are interested in other containers for VC-1 is because we want to be able to retain AC3 audio without transcoding. Can ASF handle AC3? If so, is there a muxer/splitter available anywhere?
Sure it can. It already does in DVR-MS files.

You can mux AC3 files directly into ASF by using SolveigMM's ASF muxer: http://www.solveigmm.com/?Downloads and http://forum.doom9.org/showthread.php?t=92136.

You'll need the AVI AC3 source filter in order to play them back. One caveat: playback is only possible in DirectShow based players such as MPC. WM Format SDK requires audio codecs to be either of DMO or ACM type, so any WMF based player (WMP, MCE) will complain about missing audio codec.

dohcmark8

23rd May 2006, 06:06

I looked into this. This particular trailer is 720p at 60fps with a 384kbps 5.1 soundtrack. That's truly a difficult video to decode.

I tried it on a 3.0 GHz P4 (w HT) with both WMP10 and MPC. Neither player could play it any faster than 10 fps.
On a 3.2 GHz P4 (w HT) and WMP11, it played at about 40 fps, dropping a lot of frames. MPC yielded similar performance.
I then tried it on a 3.8GHz P4 Xeon with WMP11 and it rendered fairly smoothly but with a few dropped frames here and there. In general it achieved the full framerate of the video.
Finally, I tried it on the same 3.8GHz Xeon with DXVA decoding support enabled (not available publicly yet) and it rendered smoothly with no dropped frames.

Bottom line: For this particular video, it does seem like it's a decoder performance bottleneck. 720p60 is a lot of video data to process. I didn't see any significant difference between WMP10, WMP11 and MPC, so I don't think it's a player issue at all.
What's your hardware config? Have you hard more luck playing 720p 24fps videos?

That doesnt sound right... I mean I can play the FS trailer with NO dropped frames in Zoom Player and MPC on a Athlon XP 2800+ with a Radeon 9200. I check the player stats.. they report full 60fps and no dropped frames. WMP however barely plays.

My 2.8ghz P4 HT laptop with a 9600 with WMV Acceleration... I can play in WMP11 if I disable 'drop frames' and in the statistics it shows 55fps actual rate and CPU usage around 45%.

Edit: I did a 720p30 video (downscaled 1080i video not film). And with 'drop frames' enabled in WMP11 which is the default setting and if I playback the video and look at the statistics, its not that WMP11 is actually dropping frames, it shows no frame drop and actual rate at the proper 29-ish but the actual video is still playing at what looks to be half-speedbut the audio is fine.

Edit 2: Seems 720p30 video wont playback in MPC either on my 2.8 laptop... Same half-speed playback as in WMP.

zambelli

23rd May 2006, 09:50

That doesnt sound right... I mean I can play the FS trailer with NO dropped frames in Zoom Player and MPC on a Athlon XP 2800+ with a Radeon 9200. I check the player stats.. they report full 60fps and no dropped frames. WMP however barely plays.
That does sound strange. Try this experiment:
1. Load the FlightSim video in GraphEdit.
2. Disconnect the audio pins so that only video is left.
3. Hit Play, then immediately Stop.
4. From the Graph Menu, disable "Use clock".
5. Right click on the Video Renderer and select Properties.
6. Hit Play again and let the entire video play through.

What's the average framerate reported by the renderer at the end of playback?

The point of this is exercise is to determine what's the fastest your computer can render the video without relying on audio for clock sync.

My 2.8ghz P4 HT laptop with a 9600 with WMV Acceleration...
Actually, that WMV acceleration won't work with Advanced Profile - yet. ATI needs to update their drivers first.

I did a 720p30 video (downscaled 1080i video not film). And with 'drop frames' enabled in WMP11 which is the default setting and if I playback the video and look at the statistics, its not that WMP11 is actually dropping frames, it shows no frame drop and actual rate at the proper 29-ish but the actual video is still playing at what looks to be half-speedbut the audio is fine.
Are you sure the video is encoded correctly? If you create an .avs script with the single line: DirectShowSource("my720p30video.wmv", fps=29.97) and load that into VDub, can you verify that all the frames are actually there? What tool did you use to encode the video?

dohcmark8

23rd May 2006, 14:50

What's the average framerate reported by the renderer at the end of playback?

Are you sure the video is encoded correctly? If you create an .avs script with the single line: DirectShowSource("my720p30video.wmv", fps=29.97) and load that into VDub, can you verify that all the frames are actually there? What tool did you use to encode the video?

Here are the results for my Athlon XP 2800+... http://zebsadiq.googlepages.com/fsx_2800axp_graph.jpg

Yup. I use an Avisynth script fed directly into WME. The original AVS script has 6524 frames and the *.wmv file has 6523. I know it encoded properly since it plays fine in MPC and Zoom Player. But WMP has this weird half-speed play thing.

WMP stutters and barely plays on the FlightSim vid. But on my 720p30 vid it plays fine for a few seconds then does the half-speed-ish thing.

I should add that, if I encode the same 720p30 videos with the default VC-1 settings... meaning no compression optimization or anything then they will play perfectly smooth every single time in WMP11.

bkman

23rd May 2006, 16:21

Maybe I'm using it wrong, but WMV9 AP sucks compared to x264 :o

Better luck next time, Microsoft! :D

mpgxsvcd

23rd May 2006, 16:54

Maybe I'm using it wrong, but WMV9 AP sucks compared to x264 :o

Better luck next time, Microsoft! :D

Yea you are probably using it wrong. Why don’t you post how you tried it and we can help you to figure out how to use it properly?

bkman

23rd May 2006, 17:14

I'm using it through WMNicEnc.

Settings: http://thehandofkwll.googlepages.com/vc1opts.png
Multipass mode, VBR off.

With different sources for ~650kbps and ~1250kbps comparisons, MS's VC-1 was clearly worse than x264 (more artifacts, more blurry after motion).

Sharktooth

23rd May 2006, 17:48

Yea you are probably using it wrong. Why don’t you post how you tried it and we can help you to figure out how to use it properly?
VC-1 cant be superior to h.264 - in no way. I already explained why in another thread.
Also the MS decoder is not as fast as CoreAVC even if VC-1 is less complex than h.264...
The encoder is even slower...
VC-1 looks more or less like MPEG-4 ASP with postprocessing enabled...

mpgxsvcd

23rd May 2006, 18:34

VC-1 cant be superior to h.264 - in no way. I already explained why in another thread.
Also the MS decoder is not as fast as CoreAVC even if VC-1 is less complex than h.264...
The encoder is even slower...
VC-1 looks more or less like MPEG-4 ASP with postprocessing enabled...

Wow so you have thoroughly tested VC-1 and you have already found the ideal settings for it? Then please post the optimum settings for low bit rate ~800 kb/sec 480p and Medium bit rate ~1600 kb/sec 480p content. As good as WMNicEnc encoder is there are still some settings missing that are in WME9. Also the developers of the WMV9 advanced codec have already conceded that it was not designed for low bit rate SD content. They only claim that it should be within 10% of H.264 at HD resolutions.

zambelli

23rd May 2006, 23:18

WMP stutters and barely plays on the FlightSim vid. But on my 720p30 vid it plays fine for a few seconds then does the half-speed-ish thing.
I should add that, if I encode the same 720p30 videos with the default VC-1 settings... meaning no compression optimization or anything then they will play perfectly smooth every single time in WMP11.
Hmmm. That doesn't sound good. :( I'll just have to keep an eye open for that. I can't really open a bug until I have a good local repro - but don't worry, I'll keep trying. We still have plenty of time until final WMP11 ships.

zambelli

23rd May 2006, 23:24

I'm using it through WMNicEnc.
Multipass mode, VBR off.
With different sources for ~650kbps and ~1250kbps comparisons, MS's VC-1 was clearly worse than x264 (more artifacts, more blurry after motion).
First of all, you don't have to explicitly disable options that you don't intend to use. You can just leave them not set. Also, if you plan on entirely overriding Compression Optimization Type, what's the point of enabling it in the first place?

I appreciate your feedback, but please take some additional time to play with the encoder options. Just like there's a million ways to customize x264 encodes depending on the content, it might take more than 5 minutes to find a perfect WMV9 encoding combination too.

zambelli

23rd May 2006, 23:29

Also the MS decoder is not as fast as CoreAVC even if VC-1 is less complex than h.264...
Last time I tested CoreAVC (right before they went commercial), they were still lagging behind WMV decoding. Do you have more recent data?

zambelli

23rd May 2006, 23:50

I notice a lot of people are setting NumThreads to 1 on their (presumably) single proc machines. That's not necessary. The NumThreads setting is only really useful in a reverse scenario - when you have a multi-proc machine but you only want to use, for example, 1 of its CPUs/cores. Otherwise you can just leave it not set and the CPU will still crunch away as fast as it possibly can.

In other news... Vista Beta2 has been released! It also includes the latest WMV encoder DMOs. You can download it from here: http://www.microsoft.com/windowsvista. (Not just for the codec, of course. I'm just mentioning it in case somebody's planning on beta testing Vista.)

In anticipation of the inevitable question: Yes, there might be minor differences between the various encoder DMO versions that are now becoming available (WMP11, WME SE, Vista, etc.) but they're largely dependant on when each project was "frozen" for its release. They all share the same common code, but they're built for specific platforms and might have different dependencies - so don't assume they're interchangable. Latest code does typically have less bugs, but there's no guarantee it doesn't have new bugs too. :)

Sharktooth

24th May 2006, 00:31

Wow so you have thoroughly tested VC-1 and you have already found the ideal settings for it? Then please post the optimum settings for low bit rate ~800 kb/sec 480p and Medium bit rate ~1600 kb/sec 480p content. As good as WMNicEnc encoder is there are still some settings missing that are in WME9. Also the developers of the WMV9 advanced codec have already conceded that it was not designed for low bit rate SD content. They only claim that it should be within 10% of H.264 at HD resolutions.
I already explained why VC-1 cant be superior or even on par with h.264 in another thread. Please :search:
Last time I tested CoreAVC (right before they went commercial), they were still lagging behind WMV decoding. Do you have more recent data?
A 1280x720/59.94fps VC-1 encode @ 7340kbit hogs the CPU (100%) and stutter on one of my PCs (athlonxp 3000+) while the same encode made with x264 (with High-Profile, CABAC and other crucial features enabled) using CoreAVC as decoder, plays flawlessly at 70% CPU load and looks definatly better...
However VC-1 playback on PCs is not so relevant as long as it plays ok with Harware decoder chips.
Please note there was no WMV HW decoding acceleration for that test.

Sharktooth

24th May 2006, 01:19

http://forum.doom9.org/showthread.php?p=821233#post821233
the discussion follows in the subsequent pages.
however, dont get me wrong. Im not saying VC-1 = sh!t...
Probably, having MS behind the development, it will be the first complete solution for HD-DVD authoring but it's clearly inferior in terms of compression to h.264.

mpgxsvcd

24th May 2006, 05:32

mpgxsvcd

24th May 2006, 05:32

Ok so here is what I have found with my testing. I am using the Harry potter 720x576 progressive 25 FPS .VOB file as my source. I am cropping it down to 720x304 and trying to get it down to ~750 kb/sec. These are my results.

1. First the single best option for drastically reducing average bit rate is the Force Video Scaling option. Now that does not mean that I recommend using this option. It will also drastically reduce the perceived resolution. However, if you find that you can not get the results you are looking for at your specified “LOW”(800 kb/sec or less) bit rate then you might want to try this setting with a value of 1 or

2. If you use it then you can reduce bit rate by up to 20%. Otherwise just leave it at zero.

3. Increasing Dquant Strength to anything above zero will drastically increase the quality and the required bit rate for a video. I could not get a usable bit rate with any setting other than zero.

4. Increasing Dquant Option from 2 to 3 improves the quality of the video significantly. However, it also increases the bit rate slightly(~5%) which is not always worth it.

5. If you have a single processor then you don’t need to set the Force NumThreads to zero. Just leave it out all together. The processor will go as fast as it can without a value in this setting.

6. Increasing the NumBframes above 1 will only increase your bit rate without a significant increase in quality.

7. A motion search level of 2 gives excellent quality. However it can slow the encode down considerably. A value of -1 has UNBELIEVABLE quality but it becomes PAINFULLY SLOW. Use this value only when you really, really need to achieve the best possible results or if you have a short file. I was able to equal H.264 quality only when I had this option set to -1. A value of -2 has good quality and is faster. -2 is probably the best option for most material.

8. The perceptual option can lower the bit rate. However, it can have a tendency to smooth the video out too much. Only use this option if you really need to reduce bit rate drastically.

That being said here are the settings I used to produce an HP encode that was better(Yes I said better than) any of the H.264 encodes I have seen. It also took all night to encode it. It does 1 frame per second with these settings on a 1.8 Ghz centrino. I will post the HP sample in the morning. It is quite spectacular!

Compression Optimization Type = 0
Dquant Option = 3
Dquant Strength = 0
Force LoopFilter = 1
Force Range Reduction = ffffffff
Force Video Scaling = 0
Lookahead = 10
Minimum/Maximum Complexity = 1
Motion Match Method = ffffffff
Motion Search Level = ffffffff
Motion Search Range = ffffffff
Motion Vector Coding Method = 0
Motion Vector Cost Method = 1
NumBFrames = 1
Perceptual Option = 0

bkman

24th May 2006, 07:52

^^ I did some test with those settings, and they came out a lot closer to x264 this time. But it still reacts badly to high motion, and is prone to ringing and other artifacts, even with inloop blur-o-matic on.

Plus it decodes slower than CoreAVC, as Sharktooth said.

Valeron

24th May 2006, 08:20

Sharktooth

24th May 2006, 12:55

can you please post the x264 settings and what x264 encoder (CLI or VFW) you used?

mpgxsvcd

24th May 2006, 14:06

@mpgxsvcd:

sorry to interrupt you, but i'm so confuse about your comment for the settings as "increasing the required bit rate". Aren't you encoding in 2 pass unconstrained VBR?

from experience with x264 encode, the setting of VC-1 may have very similar connotation. i think you should try similar tweak to the VC-1 encode, such as 2 BFrame for low bit rate and increase the delta quantization for B frame to 3 or 4 etc.

i would like to see everything go to max possible option like Dquant Strength to 3? and leave all the MB motion decision to adaptive.

did you give it a try? and how the result will be?

i'll test it asap, but if you would like to tell me the result, you'll save my PC from such long time encoding......

Actually I was using the quality mode for my initial testing. Basically I would encode the file at a fixed quality level. Then I would recode the file again but I would change one setting at a time. Then I would examine the final average bit rate to see if the change in the setting increased or decreased the bit rate. I went through this many times with the settings I listed above. Then when I had determined what settings gave me the lowest bit rate with the best quality I did the two pass encode at the specified bit rate. I didn’t have enough time to completely explore the settings you mention. But from the little bit of testing I did this morning you are right those settings will help, maybe even significantly. Specifically a delta quantization of 4-6 was working well. I will do some more testing and then post the results and the file. WMV9 advanced is not perfect and it may not be the best at low bit rates. However, it is a whole lot better than the old wmv9 and it is so easy to use that it everyone should at least consider it when choosing a codec. Has VC-1 ever actually been included in one of the Doom9 comparisons? Didn’t it get disqualified last time because they didn’t have the 2 pass VBR ready yet. I am hoping that it will actually get tested properly this time. I think it will hold its own.

I also did not cap the max bit rate for the two pass encoding(2-pass unconstrained). I might try doing a 2-pass encode with a specified maximum bit rate. That might help to increase the quality in the low end of the file. Those are the sections that I saw the most room for improvement.

benwaggoner

24th May 2006, 15:02

I also did not cap the max bit rate for the two pass encoding(2-pass unconstrained). I might try doing a 2-pass encode with a specified maximum bit rate. That might help to increase the quality in the low end of the file. Those are the sections that I saw the most room for improvement.
You should definitely be using Constrained VBR for real-world encoding. Unconstrained doesn't let you know what the peak decode complexity will be, and so it can be hard to predict playback. It's also the mode where you're likely to get the best overall compression efficiency. Results from Quality VBR can be hard to extrapolate from to bitrate limited encodes.

mpgxsvcd

24th May 2006, 15:18

You should definitely be using Constrained VBR for real-world encoding. Unconstrained doesn't let you know what the peak decode complexity will be, and so it can be hard to predict playback. It's also the mode where you're likely to get the best overall compression efficiency. Results from Quality VBR can be hard to extrapolate from to bitrate limited encodes.

The examples that I did with Qualty VBR seemed to be right in line with the 2 pass unconstrained VBR ones that I did. I just had to find a quality setting that gave me the average bit rate that I was looking for. I will definitely try the 2 pass constrained though. I think that is the last component I need to tweak to get the quality where I want it. Thanks for the help.

pwimmer

24th May 2006, 21:28

Here are some sample files.

Nero H264 6 Mbps: http://www.3dtv.at/Temp/Test_1440x576_H264_6Mbps.mpg

WMV3 (using old WME9 codec) 6 Mbps: http://www.3dtv.at/Temp/Test_1440x576_Old_WMV3_6Mbps.wmv

WMV3 (using old WME9 codec) 8 Mbps: http://www.3dtv.at/Temp/Test_1440x576_Old_WMV3_8Mbps.wmv

WVC1 6 Mbps: http://www.3dtv.at/Temp/Test_1440x576_WVC1_6Mbps.wmv

The old WMV3 and WVC1 6 Mbps files basically look the same. The 8 Mbps file is better, but still worse than the H264 file. My guess is that Windows Media requires a 50% higher bitrate compared to H264 at this resolution. I didn't do any tests at low resolutions and bit rates.

Anyway, I'll continue to use Windows Media for two reasons:
I can't play the H264 file on my Pentium D 2 x 3 GHz without dropped frames and I need DRM.

Here's a screenshot of my Registry settings.

http://www.3dtv.at/Temp/WVC1Reg.gif

Could anybody at Microsoft take a look at these files? In case I did something wrong during encoding, I'd really like to know because every MB I can make the files smaller saves money for webhosting ;-)

The files clearly show that there's no visible improvement of the new WVC1 and WMV3 codec over the old WMV3 codec but they are definitely worse than H.264. And no, I did no fine-tune the Nero codec, I encoded the H.264 file only once wheras I did several tests with Windows Media.

I encoded another file this week using the old WMV3 codec and experienced dropped frames at 8 Mbps VBR (15 Mbps peak). The problem did not occur at 8 Mbps (12 Mbps peak), which I finally used.

zambelli

25th May 2006, 03:52

The files clearly show that there's no visible improvement of the new WVC1 and WMV3 codec over the old WMV3 codec but they are definitely worse than H.264. And no, I did no fine-tune the Nero codec, I encoded the H.264 file only once wheras I did several tests with Windows Media.
I reviewed the files and there's definitely a difference. However, WMV3 actually looks sharper than WVC1 at times, but I think I know the reason.
If you're familiar with AviSynth, try this script:
v2 = DirectShowSource("Test_1440x576_Old_WMV3_6Mbps.wmv", fps=25.00).Subtitle("WMV3 6Mbps")
v3 = DirectShowSource("Test_1440x576_WVC1_6Mbps.wmv", fps=25.00).Trim(1,0).Subtitle("WVC1 6Mbps")
v4 = DirectShowSource("Test_1440x576_Old_WMV3_8Mbps.wmv", fps=25.00).Subtitle("WMV3 8Mbps")
Interleave(v2, v3, v4)
Save the .avs file and load it into VirtualDub. Hit Ctrl+G and skip to frame 1000 or so. The frames should now be in sync. If you move between them, you'll see a noticeable difference.

The problem is that you're not comparing WMV3 and WVC1 with equal settings. The WMV3 videos were encoded with default settings, and probably at complexity level 4 or 5 (-v_performance 80 or 100). The WVC1 video, on the other hand, was encoded with "Compression Optimization Type" enabled. Even with LoopFilter disabled, that still uses Dquant perceptual optimization and fixes the encoder complexity to level 3. Try encoding without COT, and instead set these:
"Motion Match Method" = 1
"Motion Search Level" = 2 or -1
"Motion Search Range" = -1
"NumBFrames" = 1

Encoding will take a lot longer than with WMV3, but it ought to produce a noticeably sharper image. If you encounter macroblocking and such artifacts, it means you'll probably have to apply some sort of filter or perceptual optimization option in order to make the content compress better at 6Mbps.

zambelli

25th May 2006, 04:11

When doing any kind of WMV comparisons, it's a good idea to disable post-processing in the decoder or set it to a very low level. To do this, set these DWORD registry values

HKEY_CURRENT_USER\Software\Microsoft\Scrunch\Current Post Process Mode
HKEY_CURRENT_USER\Software\Microsoft\Scrunch\Force Post Process Mode
HKEY_CURRENT_USER\Software\Microsoft\Scrunch\Post Process Mode

to a number between 0 and 4.
0 disables PP entirely. 1 is a good low value (weak deblocking).

mpgxsvcd

25th May 2006, 04:33

Ok tell me if I am way off base here but it does not look like the H.264 samples on this page are using YV12 output?

http://forum.doom9.org/showthread.php?t=90784&highlight=quality+challenge%3Cbr%20/%3E

I checked VirtualDub and it looks like the default is to convert YV12 to 24 BIT RGB(888). What does that mean exactly? When I test the different color depths it looks like the high bit RGB ones provide a much sharper picture with much better black levels. Is this true? If this is the case then that explains the "soft" affect that wmv always has since its default is always YV12 or IYUV. When I tried setting the color depth to RGB 32 in WME9 the file suddenly looked much more like the H.264 examples. Is there a way to find out what color depth my source was recorded at? What are most Film based DVDs done at?

Valeron

25th May 2006, 07:02

@mpgxsvcd:
VC-1 natively accept YV12 as the only color space format like the other pop video coding techs
therefore, there's no way to make a real RGB-24/32 VC-1 encode(even AVC High Profile can't do that, but it support up to 4:4:4 YUV encode)
but you can easily decode the YV12 to any color space you'd like with ffdshow raw trasform filter.(to manipulate it, you should take a glance at the ffdshow raw decoder part)

i've never try encode wmv with RGB option enabled in WME. but according to your description, that maybe a bitstream/asf container tweak to make it "work":tell the vc-1 decoder to do a RGB transform after decoding.

bye

Sagittaire

25th May 2006, 12:44

VC-1 cant be superior to h.264 - in no way. I already explained why in another thread.
Also the MS decoder is not as fast as CoreAVC even if VC-1 is less complex than h.264...
The encoder is even slower...
VC-1 looks more or less like MPEG-4 ASP with postprocessing enabled...

well be carefull I think that

1) H264* is faster than VC-1* with post process
2) H264* is slower than VC-1* without post process

* H264 with CoreAVC decoder and maximum complexity setting for encoding, VC-1 with DMO decoder and maximum complexity setting for encoding, Software decoding only for H264 and VC-1.

Sharktooth

25th May 2006, 13:55

well be carefull I think that

1) H264* is faster than VC-1* with post process
2) H264* is slower than VC-1* without post process

* H264 with CoreAVC decoder and maximum complexity setting for encoding, VC-1 with DMO decoder and maximum complexity setting for encoding, Software decoding only for H264 and VC-1.
I did my tests with PP disabled and it's still slower than CoreAVC...

bond

25th May 2006, 15:08

well mpeg-2 decoding is also faster than avc decoding? so what? you cant compare apples with pears

what would be important is the speed vs. quality relationship. speed alone doesnt mean much

pwimmer

25th May 2006, 15:21

The problem is that you're not comparing WMV3 and WVC1 with equal settings. The WMV3 videos were encoded with default settings, and probably at complexity level 4 or 5 (-v_performance 80 or 100). The WVC1 video, on the other hand, was encoded with "Compression Optimization Type" enabled. Even with LoopFilter disabled, that still uses Dquant perceptual optimization and fixes the encoder complexity to level 3. Try encoding without COT, and instead set these:
"Motion Match Method" = 1
"Motion Search Level" = 2 or -1
"Motion Search Range" = -1
"NumBFrames" = 1

Encoding will take a lot longer than with WMV3, but it ought to produce a noticeably sharper image. If you encounter macroblocking and such artifacts, it means you'll probably have to apply some sort of filter or perceptual optimization option in order to make the content compress better at 6Mbps.

Thanks. I'll try the suggested settings and tell you about the results.

Another question: I'm running the encoder on a dual core CPU. Will quality really be better when using a single thread only? Could you please explain why the quality is lower when using two threads?

Does one thread process the left and the second thread the right image? Do you think it makes a difference if I encode left and right view of my stereoscopic files in side-by-side or over/under format?

Sharktooth

25th May 2006, 15:30

well mpeg-2 decoding is also faster than avc decoding? so what? you cant compare apples with pears

what would be important is the speed vs. quality relationship. speed alone doesnt mean much
i said, being VC-1 less complex than AVC, it should decode faster but that's not actually true.
also quality is lower than AVC...
so, at the end VC-1 is slower (in both encoding and decoding) and provides lower quality than AVC.

Inventive Software

25th May 2006, 16:49

Either that, or somebody needs to implement a decent decoder that's on a par with CoreAVC.

zambelli

25th May 2006, 19:45

Another question: I'm running the encoder on a dual core CPU. Will quality really be better when using a single thread only? Could you please explain why the quality is lower when using two threads?
Yes, although it's probably negligible in most cases. Multithreaded encoding is done by splitting up the image into splices and then encoding each splice in a separate thread. This can affect the estimations because each thread only has access to a part of the image.

The WVC1 encoder automatically detects the number of CPUs present so in your case it's encoding with 2 threads by default. You can try forcing it to single thread - but I doubt you'll notice a difference in quality. We're probably talking about hundredths of a dB point in PSNR.

Update: I just encoded a short video on a 3.2 GHz P4 with HyperThreading. Not true dual-proc, but close enough for the experiment. With NumThreads set to 2, encoding was 15% faster than with NumThreads set to 1. The quality - and this was very surprising - was actually higher (!) in the 2-thread video - but only by 0.0261dB PSNR and 0.1 SSIM. I'm guessing this is just a coincidence based on the nature of the content - for some other content it could've been the other way around. My point is still the same: the quality difference is pretty negligible, whereas the perf advantage is significant. 15% on a HT processor is pretty good - I'm guessing that with true dual-cores the advantage could be above 30%.

Inventive Software

26th May 2006, 14:26

So this is even more reason for me to upgrade to dual core then! When I do, it's time for a new codec test!

mpgxsvcd

26th May 2006, 14:52

I have done some exhaustive testing of the new WMV9 codec and I have come to the conclusion that it can get pretty close to H.264 but it still has one huge flaw. It seems that every file I encode no matter how high I set the bit rate it always reduces the color range. Every video looks soft because it never shows true “black”. It looks like all of the blacks are actually a shade of grey. When I use H.264 it always keeps black as black. Now this might have something to do with the color space that I use with H.264(It converts it to RGB by default). However, no matter what I do the H.264 looks better because it is showing true black. I set the registry values below when I encoded it. I always set Force Range Reduction = 0. Will this eliminate the color range reduction? Should I set this to another value or should I leave the registry entry out all together? Any help on this would be appreciated. I really want to use this codec but the color reduction is just not acceptable.

Compression Optimization Type = 0
Dquant Option = 3
Dquant Strength = 0
Force B Frame Delta QP = 4
Force LoopFilter = 1
Force Range Reduction = 0
Lookahead = 10
Motion Match Method = ffffffff
Motion Search Level = ffffffff
Motion Search Range = ffffffff
Motion Vector Cost Method = 1
NumBFrames = 1
Perceptual Option = 1

Here is sample HD 720pWMV9 Advanced profile video clip. It is from American Idol. It is of one of Katherine Mcphee’s better performances. The first one is at a really high bit rate and the second one at a lower bit rate. The first file also has the 5.1 audio but I don’t think it really decoded the DD correctly to each channel. Anyway enjoy the samples.

High Quality file
http://www.yousendit.com/transfer.php?action=download&ufid=5BCE19E1560A3B0C

Low Quality file
http://www.yousendit.com/transfer.php?action=download&ufid=27E3A79126D39F4A

Sharktooth

26th May 2006, 15:17

mpgxsvcd

26th May 2006, 16:16

same problems as i described few posts above...
the decoder is slooooow... the videos are unwatchable on an athlon xp 3000+ (even with overclocking).
however h.264 (x264) can do almost the same at half or 2/3 the bitrate (your LQ video is encoded @3790kbps and the HQ @4709... and both show blocks and other artifacts)

I watched the HQ one this morning on a 55” LCD TV at 1080i and there were absolutely no compression artifacts what so ever. That was also running on an old Athlon 3200 with no over clocking and no dropped frames. I am not sure what you are doing wrong but something is not right with your setup if you saw blocks in the HQ one. Did you turn off your decoding processing for wmv9? That may be the problem. I left mine on and there were no compression artifacts.

What bit rate do you think you can get this file down to? I have a 2.5 mb/sec version that I have not posted yet. It still retains most of the original quality. The @3790kbps LQ one was done without any of the advanced parameters set so it should not look as good.

Could you recode this file to X.264 if I give you the original file? I need to find a way to cut an MPG2 file losslessly though. Any ideas on how to do that?

mpgxsvcd

26th May 2006, 16:21

same problems as i described few posts above...
the decoder is slooooow... the videos are unwatchable on an athlon xp 3000+ (even with overclocking).

Remember, this is a 30 FPS file instead of the original 60FPS that it was broadcast at. You will not see as smooth motion with this file as the original because it is at half the frame rate. You might have mistaken the 30 FPS for dropped frames. My 2.2 Ghz P4 with 512 megs of ram and only 533 FSB played the HQ version flawlessly.

Sagittaire

26th May 2006, 16:48

I compare VC1 vs H264 vs WMV9
http://multimediacom.free.fr/Video/

First visual Impression
- VC1 is better than WMV9 VCM
- H264 is really better than VC1

Detailled Visual Impression
In fact first impression is not good. Quality for high motion and complexe texture are very bad for VC1. IMO VC1 use very high quant (too high?) for complexe sequencies and very low quant (too low?) for simple sequencies : Visual quality is not constant.

I make sample for high motion part
http://multimediacom.free.fr/Video/sample-wmv9.avi
http://multimediacom.free.fr/Video/sample-vc1.wmv
http://multimediacom.free.fr/Video/sample-x264.mp4
http://multimediacom.free.fr/Video/sample-ateme.mp4
http://multimediacom.free.fr/Video/sample-elecard.mp4

As you can see WMV9 and H264 produce better visual result in these particular part. But why ... ???

Here the size for these part:
VC-1 done 1503 Ko
WMV9 VCM done 1707 Ko
Ateme done 1808 Ko
Elecard done 1657 Ko
x264 done 1735 Ko

Simply because all the codec use higher size for these particular part.

Metric Comparison
- H264 done very better overall result than VC1
- WMV9 VCM done better overall result than VC1 ... ???

H264 vs VC1 450 Kbps
http://multimediacom.free.fr/Video/H264vsVC1-450.PNG

H264 vs VCM 450 Kbps
http://multimediacom.free.fr/Video/H264vsVCM-900.PNG

H264 vs VC1 900 Kbps
http://multimediacom.free.fr/Video/H264vsVC1-900.PNG

Minimum local peak bitrate are generaly complexe texture or high motion part.
Maximum local peak bitrate are generaly simple texture or low motion part.

As you can see delta for maximum/minimum local peaK PSNR are very higher for VC1. very high quality for simple part and very low quality for complexe part. IMO quantizer variability is too high for VC1 : Rate Control is certainely really near than CBR mode with low variability for bitrate.

As you can see WMV9 VCM and H264 done very similar graph for PSNR. H264 is better than VCM (compression efficiency is better) but Rate Control are in practice exactly the same.

Conclusion: Your Rate Control is bad for Overall PSNR and IMO Overall PSNR can check good Rate Control. I can't say if your codec is good but your Rate Control is certainely bad: Quality in high motion / complexe texture are really bad. Actually VC-1 AP done very better visual result for low bitrate than MPEG4 ASP.

pwimmer

26th May 2006, 19:29

It seems that every file I encode no matter how high I set the bit rate it always reduces the color range. Every video looks soft because it never shows true “black”. It looks like all of the blacks are actually a shade of grey.

I didn't had this problem during my tests. It is probably caused by the graphics hardware. Virtually all video cards are faulty and do YUV->RGB conversion wrong, either in DirectX mode (used by VMR-9) or in overlay mode (used by VMR-7 and old VR).

Microsoft HW quality lab shouldn't certify drivers anymore which are not able to convert YUV2 and UYVY textures to RGB properly or cannot display a proper hardware overlay.

For codec comparisions, I always turn hardware acceleration off in the display properties. Performance will be horrible, but colors will be accurate.

pwimmer

26th May 2006, 19:40

Thanks. I'll try the suggested settings and tell you about the results.

I tried the settings, but the result was basically the same.

WMV/VC-1 is a good codec, definitely better than MPEG-4 ASP, but it cannot catch up to H.264. My guess is that Microsoft will have to revise their VC-1 strategy sooner or later.

For HD content on Bluray or HD DVD, it's not necessary to get the best compression efficiency possible. But for content hosted on web servers it is, because each additional MB traffic costs money.

I wonder if Microsoft will integrate H.264 in the Windows Media platform or develop a new codec, e.g. VC-2...

bond

26th May 2006, 20:07

mpeg-4 asp/avc, vc-1 aso are formats, not codecs

it doesnt make sense to make "h.264 vs. vc-1" encode comparisons. if you are comparing two formats you should always also mention what actual codec (eg xvid, x264 or whatever) and codec settings you used, cause this can make a very big difference

Sharktooth

27th May 2006, 00:36

I watched the HQ one this morning on a 55” LCD TV at 1080i and there were absolutely no compression artifacts what so ever. That was also running on an old Athlon 3200 with no over clocking and no dropped frames. I am not sure what you are doing wrong but something is not right with your setup if you saw blocks in the HQ one. Did you turn off your decoding processing for wmv9? That may be the problem. I left mine on and there were no compression artifacts.
I turned off HW acceleration to compare VC-1 with coreavc (software decoding only) and your samples are both unplayable on the XP 3000+ (video lagged behind audio and several frames where dropped). That means the VC-1 decoder is not properly optimized.
also i compared the results with PP turned off and there are clearly visible artifacts even in the HQ sample.
here are some sample images taken from the HQ encode (it's a mix of frame types randomly taken, there are also I-Frames taken at scene changes):
http://www.webalice.it/f.corriga/temp/vc-1.7z
watch them in full screen...
there are artifacts everywhere (blocks, rings, textures discolorations/blurring, moving textures, etc...). i also tried to not take screenshots on hi-motion frames coz there were even more artifacts (read as: those in the archive are not even the worst frames you can find in the encode!)
What bit rate do you think you can get this file down to? I have a 2.5 mb/sec version that I have not posted yet. It still retains most of the original quality. The @3790kbps LQ one was done without any of the advanced parameters set so it should not look as good.

Could you recode this file to X.264 if I give you the original file? I need to find a way to cut an MPG2 file losslessly though. Any ideas on how to do that?
Im not sure about the bitrate but i can recode it at the same bitrate and at lower ones.

zambelli

27th May 2006, 00:50

I turned off HW acceleration to compare VC-1 with coreavc (software decoding only)
HW acceleration? Umm, WVC1 is software decoding only. No hardware vendor has released video drivers yet that support DXVA decoding of WVC1.

Sharktooth

27th May 2006, 01:16

For safety i turned off WMV acceleration.
i edited my previous post. please check the comments and get screeshots archive.
(link for the lazy ppl: http://www.webalice.it/f.corriga/temp/vc-1.7z)

dohcmark8

27th May 2006, 05:23

Yeah the decoder has issues. Needs optimization badly.:(

Sagittaire

27th May 2006, 09:02

benwaggoner

27th May 2006, 21:13

The offical beta release of the new codec has been posted. This means you don't have to install WMP11 if you want to play with the codec.

http://www.microsoft.com/windows/windowsmedia/forpros/encoder/default.mspx

zambelli

28th May 2006, 04:52

Thanks, Ben - I updated the first post in the thread with WVC1 codec release news, but forgot to post something at the end of it. :)

Everyone should take a look at http://forum.doom9.org/showthread.php?t=111633 for more details about the standalone WVC1 codec release.

mpgxsvcd

30th May 2006, 19:34

Ok I have finally settled on some settings for WMV9. I have two different cases that I setup for WMV9. They are both listed below. I was able to get about 25% better compression than I was getting before by raising the Force B Frame Delta QP and NumBFrames values.

For high quality encodes (3-5 Mb/Sec for 720p 2-3 Mb/Sec for 480p)

Compression Optimization Type = 0
Dquant Option = 3
Dquant Strength = 0
Force B Frame Delta QP = 4-8 (Set this to 4 if you are not that concerned about compression you just want to optimize quality. Set it to 8 if you want decent quality but you want to optimize compression. Try values between 4 and 8 if you want both good quality and medium compression.)
Force LoopFilter = 1
Force Range Reduction = 0
Force Video Scaling = 0
Lookahead = 10
Motion Match Method = ffffffff
Motion Search Level = ffffffff
Motion Search Range = ffffffff
Motion Vector Cost Method = 1
NumBFrames = 2-3(Use 2 for optimal quality. Use 3 if you are more concerned with compression)
Perceptual Option = 1

For Medium quality encodes (2-3 Mb/Sec for 720p .6-2.0 Mb/Sec for 480p)

Compression Optimization Type = 0
Dquant Option = 2
Dquant Strength = 0
Force B Frame Delta QP = 7-9 (Set this to 9 only if you need to get down to about .6 Mb/sec for 480p content. Quality will be sharply reduced if you go much over 8 for this value.)
Force LoopFilter = 1
Force Range Reduction = 0
Force Video Scaling = 1 (I have found that a value of 1 for this setting doesn’t really reduce resolution much but you can get about 5% more compression. If you need to get down to really low bit rates then you can use a setting of 2 but this will reduce the sharpness considerably.
Lookahead = 10
Motion Match Method = ffffffff
Motion Search Level = ffffffff
Motion Search Range = ffffffff
Motion Vector Cost Method = 1
NumBFrames = 3-4 (3 is a good setting here. Use 4 if you need extreme compression)
Perceptual Option = 1

zambelli

30th May 2006, 20:23

Compression Optimization Type = 0
Dquant Strength = 0
Force Range Reduction = 0
Force Video Scaling = 0

You probably don't need to explicitly set these to 0. It ultimately doesn't make a difference, but it'll keep your list of registry settings shorter. :)

mpgxsvcd

30th May 2006, 21:32

You probably don't need to explicitly set these to 0. It ultimately doesn't make a difference, but it'll keep your list of registry settings shorter. :)

Yes I know that you don’t have to explicitly state them. I just wanted to be perfectly clear to everyone about what I was using. For example in one of them I recommend using a Force Video Scaling value of 1. In the other I don’t recommend using that setting. Also I was under the impression that the default Force Range Reduction value was “ffffffff’? Doesn’t that mean that the codec chooses when to use the range reduction? I don’t recommend using that setting at all. That is why I stated the value of zero for that one. Finally, I listed the Dquant Strength value of zero because I got very large files if I used any other value besides that. You can leave the setting out all together and it will do the same thing but I wanted everyone to know that a value of zero is the only one that gave me good results.:thanks:

lazyn00b

31st May 2006, 08:02

Sure it can. It already does in DVR-MS files.

You can mux AC3 files directly into ASF by using SolveigMM's ASF muxer: http://www.solveigmm.com/?Downloads and http://forum.doom9.org/showthread.php?t=92136.

You'll need the AVI AC3 source filter in order to play them back. One caveat: playback is only possible in DirectShow based players such as MPC. WM Format SDK requires audio codecs to be either of DMO or ACM type, so any WMF based player (WMP, MCE) will complain about missing audio codec.

Sorry it took me so long to respond, but I did try SolveigMM's stuff without much satisfaction.

How do I create a DVR-MS from a WMV and an AC3? DVR-MS would be great because then I could play it from MCE 2005 interface without any hassle. What we really need is a simple little tool, even a CLI, that can mux (and demux!) tracks from a DVR-MS.

zambelli

31st May 2006, 10:29

Sorry it took me so long to respond, but I did try SolveigMM's stuff without much satisfaction.
What about Solveig's tool didn't work for you? I remember muxing WMV and AC3 quite successfully with it.

Sharktooth

31st May 2006, 10:38

@Zambelli, are there any chances to see a faster decoder in the near future?
i have the impression the actual one is optimized for Intel CPUs and it performs poorly on Athlons (expecially on non-SSE2 enabled CPUs).

mpgxsvcd

31st May 2006, 14:12

Ok forcing range reduction is a little confusing. You actually have to force the value in the registry. If you just try to type the value in then it will only let you do 7 characters. You need to enter 8 characters or it will crash WME9. Try copying the value and pasting into the dword value in the registry.

Try copying 1010101 as the dword value. That should be the minimum reduction besides 00000000 which is no reduction. You could also try 08080808 which is the maximum.

I had never actually tried this value until this morning. It really appears to work. It looks like it introduces black crush. Most of the shadow details get reduced considerably. However, the bit rate is reduced considerably also. I might start using this setting now. I had just used 00000000 before.

Dayvon

31st May 2006, 15:07

Sharktooth

31st May 2006, 15:46

Dayvon

31st May 2006, 16:26

@ mpgxsvcd

Motion Match Method = ffffffff
Motion Search Level = ffffffff
Motion Search Range = ffffffff

What are these values equal to when using NicWMVEncoder? Values available are -1,0,1.

Basically, I'd like to replicate your *best* encoding settings as a place to begin testing WM9VC-1.

Thx!

mpgxsvcd

31st May 2006, 18:35

@ mpgxsvcd

What are these values equal to when using NicWMVEncoder? Values available are -1,0,1.

Basically, I'd like to replicate your *best* encoding settings as a place to begin testing WM9VC-1.

Thx!

Those are all -1.

mpgxsvcd

31st May 2006, 18:45

@ Sharktooth...

Maybe it is just me, but these pictures look like most HD-quality content that I've see on cable-sat. Basically that the video that they send out is a bit too compressed for the quality. I remember seeing VERY similar artifacting at a friends house (60" HDTV) watching American Idol a few months ago (the background red's look like mush). To me that looks very much like what I've seen from on-air HD.

@ mpgxsvcd

How did you get your source file? Did you record it off the air? If so, how did you record it initially? And lastly, if you didn't record it off air, how did you get the file?

My source file was from American Idol as it was broadcast over-the-air at 720p 60 FPS. I use an ATI HDTV wonder that allows me to capture the full unaltered 18 Mbps HD stream from any of the over the air HD stations. The original HD MPG2 stream from American Idol has absolutely no compression artifacts. If you see compression artifacts with American Idol HD then you are getting a compressed signal. The over the air signal I am getting from our local affiliate is artifact free. If you use satellite then you are not actually getting HD. It is actually HD-LITE which is a down-rezzed and bit starved version of HD. If you use over the air and you see the artifacts then your local station must be doing the same thing. I can upload the source file if someone can let me know how to remove a segment of an MPG2 video losslessly? I guess I could use video vegas and the new lossless WMV video and audio codecs? Has anyone messed with those yet? Do you want to see a sample?:D

Dayvon

31st May 2006, 19:22

@ mpgxsvcd

I don't need to see your sample thx :) . But I do have problems with HD over-the-air and their compressions. I have seen HD over-the-air (ALIAS episode specifically), this years Super Bowl (either cable or OTA), and American Idol (either cable or OTA). To say HD is cool and great is true. To say it is without compression artifacts I really believe is being well, untrue. At some point along the chain of camera -> file/stream -> broadcast, the source is MPEG-2 encoded and I really haven't seen any HD OTA or cable, without this compression being easily seen. I have seen true HD with HD trailers and I have seen some of those are truly artifact free.

Now I don't demand to be the authority on all things HD by any means, but I do know that I have seen OTA HD and it was impressive to be sure. But the one thing that didn't impress me was compression. And the files that I uploaded from you did have that "source compression" to my eyes.

@ Everyone

I did a quick comparison between VC-1, x264, and Xvid. AVScript below.

SetMemoryMax(512)

LoadPlugin("C:\Program Files\DGIndex\DGDecode.dll")
video=mpeg2source("D:\TEMP Rev\Pride&Prejudice\VIDEO_TS\VTS_01_1.d2v",cpu=4)
ColorMatrix()
Trim(98083,104506)
crop(2,56,-2,-62)
LanczosResize(864,368)

Multi-pass @ 1000kbps - Encode time was approximately the same (not noticably longer for any of them)

1. P&P-VC1.wmv 32MB (http://church.crossingatwoodland.com/Temp/download/P&P-VC1.wmv)
This file was made with Nic's WMVEncoder using WMV9 Advanced Profile. Settings were nearly identical to High Quality settings posted in this thread on Page 9 by mpgxsvcd (thx BTW).

2. P&P-xvid.avi 30.8MB (http://church.crossingatwoodland.com/Temp/download/P&P-xvid.avi)
Encoded using VirtualDubMod and Xvid 1.1.0 ( w/ QPEL, GMC, B-VOP's, Mode-6, VHQ-1, /w chroma motion)

3. P&P-x264.mp4 32MB (http://church.crossingatwoodland.com/Temp/download/P&P-x264.mp4)
Encoded using MeGUI and x264 r504 (x264.exe --pass 2 --bitrate 1000 --stats ".stats" --ref 3 --mixed-refs --bframes 2 --b-pyramid --bime --weightb --filter -3,-3 --analyse all --8x8dct --threads 2 --thread-input --progress --no-psnr --output)

(NOTE: I might try the WMV9 again with some setting changes (per mpgxsvcd post moments ago). If I do, I will substitute the file in place of the one here, and note the edit.)

Perceptions: I think that this VC-1 really isn't a bad codec. I don't think it is as good as AVC High-Profile, but really that would be expecting too much. I will probably use it for compatability's sake for public stuff, but for my personal use, x264 is really the ticket. I thought that VC-1 did a good job with facial details when the rain wasn't an issue. Overall VC-1 looked smeared, Xvid looked harsh/artifical, and x264 looked pretty much right.

Overall, I'm very happy with VC-1's overall performance, and aside from rain (which gives every codec hell) I think it is a very usable and quality codec.

mpgxsvcd

31st May 2006, 21:03

After testing the Force Range Reduction setting I have come to the conclusion that this is what causes the black crush affect that I was seeing in the H.264 and X.264 samples that I had seen. They all appeared to have less detail in the shadows but they all had much deeper blacks. This gave the videos the appearance of being sharper. At first I thought it was a color space or Pixel Format setting. However, now I realize that it must be some form of color range reduction. Before I could get very good samples out of wmv9 advanced but they never quite got to the level of X.264. With the use of the range reduction I can now meet and even exceed the results of X.264. There are some tricks to this setting though. First off you have to manually force in 8 digits instead of the 7 that it defaults to. If you don’t do this WME9 will crash. It also performs differently with 1 pass Quality VBR than it does with 2-pass unconstrained VBR. With Quality VBR you can set it to a value of 06060606 and you won’t see that much of a range reduction in the file. However, you will see a gigantic file size difference(~25%). If you use a setting of 06060606 with 2 pass VBR then the file will be unwatchable because the colors have been reduced down to basically black. The screen will be so dark that you can not recognize anything. The highest setting I could use with 2-pass is 01010101. So if you can use Quality VBR then give the range reduction a try. It made all of the difference for me.

Dayvon

31st May 2006, 21:14

Before I could get very good samples out of wmv9 advanced but they never quite got to the level of X.264. With the use of the range reduction I can now meet and even exceed the results of X.264.
I don't understand how the color range can change the results of a test like mine, using the same settings as those you posted previously... Maybe that's just me.

Did you look at the files I posted?

mpgxsvcd

31st May 2006, 22:14

Dayvon

31st May 2006, 22:44

As I have limited hosting space no, I cant host my source file. :(

As to artifact free, do you see the horrible blurry mess that is the red background behind Kathryn? I mean to me it is kind of obvious that something is wrong.

I'm on a 1360x768 32" HDTV LCD from 4-5 ft away BTW. And I couldnt get WME9 to work with my AVSynth file which is why I used Nic'sWMVEncoder.

zambelli

31st May 2006, 23:23

There are some tricks to this setting though. First off you have to manually force in 8 digits instead of the 7 that it defaults to. If you don’t do this WME9 will crash. It also performs differently with 1 pass Quality VBR than it does with 2-pass unconstrained VBR. With Quality VBR you can set it to a value of 06060606 and you won’t see that much of a range reduction in the file. However, you will see a gigantic file size difference(~25%). If you use a setting of 06060606 with 2 pass VBR then the file will be unwatchable because the colors have been reduced down to basically black. The screen will be so dark that you can not recognize anything. The highest setting I could use with 2-pass is 01010101. So if you can use Quality VBR then give the range reduction a try. It made all of the difference for me.
I get the impression you're not using this setting correctly. The statement "you have to manually force in 8 digits instead of the 7 that it defaults to" leads me to believe you are entering the value as a decimal number. If you enter 06060606 as a hexadecimal number (as indicated by the 0x format in the documentation) in Regedit, it will retain the leading 0.

Changing this setting can improve compressability and reduce blurring and macroblocking (more bitrate will be used to compress less information), but will reduce the luma and color accuracy. 4:2:0 is already a chroma-starved format. I personally would not reduce its range by a factor of 6, but if you're not seeing the sideeffects - then go for it.

zambelli

31st May 2006, 23:36

And I couldnt get WME9 to work with my AVSynth file which is why I used Nic'sWMVEncoder.
Have you tried the new wmcmd.vbs from http://forum.doom9.org/showthread.php?t=111633?

Dayvon

31st May 2006, 23:53

Have you tried the new wmcmd.vbs from http://forum.doom9.org/showthread.php?t=111633?
Most definitely. Here's my simple AVScript.
SetMemoryMax(512)

LoadPlugin("C:\Program Files\DGIndex\DGDecode.dll")
mpeg2source("D:\TEMP Rev\Pride&Prejudice\VIDEO_TS\VTS_01_1.d2v",cpu=4)
ColorMatrix()
Trim(98083,104506)
crop(2,56,-2,-62)
LanczosResize(864,368)
Here's the error messages I get from WME9.

The requested video codec is not installed on your system. (0xC00D0BC4)
I import the AVS file as a script. Then this happens whenever I select the compression tab, selecting the video to adjust bitrate/settings. I did however, install the download that you linked above.
Also, if I try to source the AVSfile as a video it says that "No such interface is supported".

Sharktooth

1st June 2006, 00:04

3. P&P-x264.mp4 32MB (http://church.crossingatwoodland.com/Temp/download/P&P-x264.mp4)
Encoded using MeGUI and x264 r504 (x264.exe --pass 2 --bitrate 1000 --stats ".stats" --ref 3 --mixed-refs --bframes 2 --b-pyramid --bime --weightb --filter -3,-3 --analyse all --8x8dct --threads 2 --thread-input --progress --no-psnr --output)
Why dont you use r530 and some more "quality" settings for x264?
Your settings doesnt even use RDO...
try those ones: x264.exe --pass 2 --bitrate 1000 --stats ".stats" --ref 5 --mixed-refs --bframes 3 --b-pyramid --b-rdo --bime --weightb --direct auto --filter -3,-3 --subme 6 --trellis 1 --analyse all --8x8dct --me umh --threads 2 --thread-input
(adjust the inloop deblocking for your needs)
and look again at the difference between VC-1 and x264...

zambelli

1st June 2006, 00:07

Most definitely.
Do you get the same error from WMCmd.vbs that you get from WME9 GUI?

Here's the error messages I get from WME9.
The requested video codec is not installed on your system. (0xC00D0BC4)
Can you play the .avs in a DirectShow player such as WMP?

I import the AVS file as a script. Then this happens whenever I select the compression tab, selecting the video to adjust bitrate/settings. I did however, install the download that you linked above.
As script? Oh. :D
No, that "Script" in WME9 does not refer to Avisynth. It refers to script commands that gets inserted into ASF files for subsequent processing during playback.

Also, if I try to source the AVSfile as a video it says that "No such interface is supported".
The way to use .avs in WME9 is to specify "Both device and file" as the source, then point Video and/or Audio to the .avs file. A little convoluted, I know, but WME9 wasn't really designed with Avisynth in mind.

Dayvon

1st June 2006, 00:34

Why dont you use r530 and some more "quality" settings for x264?
Your settings doesnt even use RDO...
try those ones: x264.exe --pass 2 --bitrate 1000 --stats ".stats" --ref 5 --mixed-refs --bframes 3 --b-pyramid --b-rdo --bime --weightb --direct auto --filter -3,-3 --subme 6 --trellis 1 --analyse all --8x8dct --me umh --threads 2 --thread-input
(adjust the inloop deblocking for your needs)
and look again at the difference between VC-1 and x264...
I updated my x264 I swear like a month or two ago and we are already to r530.... sheesh.

You've been making up time since being back (congrats on the recovery BTW)!

As to RDO, in my current version of x264 RDO is a tad to slow for my taste.

Do you get the same error from WMCmd.vbs that you get from WME9 GUI?
I have not tried a cmdline anything and wont if that's what you mean. WME9 just isn't responding. I even tried using the "Both device and file", I also tried that thing you posted in the knowledge base about altering the vbs file to no look for time (or something like that). But alas, WME9 doesn't like me, and I don't like it :sly:

zambelli

1st June 2006, 00:43

I have not tried a cmdline anything and wont if that's what you mean. WME9 just isn't responding. I even tried using the "Both device and file", I also tried that thing you posted in the knowledge base about altering the vbs file to no look for time (or something like that). But alas, WME9 doesn't like me, and I don't like it :sly:
If you're not interested in command-line encoding, why did you try altering the VBS? That's somewhat of a non-sequitur.
I'm not sure what the problem is because I don't have enough information. If you want me to troubleshoot it, you're going to have to give me your exact steps.

Dayvon

1st June 2006, 00:59

If you're not interested in command-line encoding, why did you try altering the VBS? That's somewhat of a non-sequitur.
I'm not sure what the problem is because I don't have enough information. If you want me to troubleshoot it, you're going to have to give me your exact steps.
No troubleshooting necessary. That's why I used Nic'sEncoder and everything seemed to work fine. I was able to select WM9 advancded profile, and change the settings. It encoded the above file fine, so what's to fix?

Thanks for your help and concern though ;) .

Valeron

1st June 2006, 04:14

RDO is way too slow, not standard required, and VC-1 can do that too(nerver be an option for me because it's like a hell).

i've tried my 1st VC-1 encode at 700k with switching mostly MB setting to adpative. and the result is.... an MPEG-1?

ugly.....

lazyn00b

1st June 2006, 04:19

What about Solveig's tool didn't work for you? I remember muxing WMV and AC3 quite successfully with it.

Basically the main problem is that the resulting file can not be played back via the MCE interface.

zambelli

1st June 2006, 05:31

Basically the main problem is that the resulting file can not be played back via the MCE interface.
Ah, yeah. It's an unfortunate issue. The problem is that MCE uses WMF SDK to read ASF files. By design, WMF SDK only supports DMO and ACM audio codecs. Most AC3 decoders are DShow codecs, and thus we have a problem. A DirectShow player such as MPC will read ASF+AC3 without a problem through DirectShow's QASF filter, but WMF native player such as WMP or MCE will choke on the audio stream because it can't use a DShow audio decoder. :(
So to be clear - the problem is not in the file format, but in the current WMF SDK implementation.

lazyn00b

1st June 2006, 08:00

Ah, yeah. It's an unfortunate issue. The problem is that MCE uses WMF SDK to read ASF files. By design, WMF SDK only supports DMO and ACM audio codecs. Most AC3 decoders are DShow codecs, and thus we have a problem. A DirectShow player such as MPC will read ASF+AC3 without a problem through DirectShow's QASF filter, but WMF native player such as WMP or MCE will choke on the audio stream because it can't use a DShow audio decoder. :(
So to be clear - the problem is not in the file format, but in the current WMF SDK implementation.

So how about DVR-MS, then - is there any way to "make" one?

zambelli

1st June 2006, 09:24

So how about DVR-MS, then - is there any way to "make" one?
Possibly through the Stream Buffer Engine.
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/directshow/htm/usingthestreambufferengine.asp

mpgxsvcd

1st June 2006, 13:55

I get the impression you're not using this setting correctly. The statement "you have to manually force in 8 digits instead of the 7 that it defaults to" leads me to believe you are entering the value as a decimal number. If you enter 06060606 as a hexadecimal number (as indicated by the 0x format in the documentation) in Regedit, it will retain the leading 0.

Changing this setting can improve compressability and reduce blurring and macroblocking (more bitrate will be used to compress less information), but will reduce the luma and color accuracy. 4:2:0 is already a chroma-starved format. I personally would not reduce its range by a factor of 6, but if you're not seeing the sideeffects - then go for it.

Nope I am entering 06060606 in hexadecimal format. If I open the value back up after I have saved it then the leading zero is gone and it will crash WME9 every time. I think there are some serious bugs around this setting that go all the way back to the original WME9 Advanced profile. Could you try a 1 pass Quality with a range reduction of 06060606? When I try that the file is smaller but it does not really appear to reduce the color range. When I do a two pass unconstrained then the color reduction is obvious and unacceptable. That is why I said that I was getting good results even using a high reduction value with a Quality encode. Any help on this setting is appreciated. It is still some what of a mystery as to what it actually is supposed to do?

Wilbert

1st June 2006, 14:49

Ah, yeah. It's an unfortunate issue. The problem is that MCE uses WMF SDK to read ASF files. By design, WMF SDK only supports DMO and ACM audio codecs. Most AC3 decoders are DShow codecs, and thus we have a problem. A DirectShow player such as MPC will read ASF+AC3 without a problem through DirectShow's QASF filter, but WMF native player such as WMP or MCE will choke on the audio stream because it can't use a DShow audio decoder.
Can't you simply install an AC3-ACM decoder (like the one from fccHandler)?

Dayvon

1st June 2006, 15:00

@ mpgxsvcd

These are the settings I used.

Compression Optimization type = 0
Denoise option = default
Dquant Option = 3
Dquant Strength = 0
Force Frame Delta QP = 4
Force LoopFilter = 1
Force Median = default
ForceNoiseEdgeRemoval = default
Force Num Threads = 1
Force Overlap = default
Force video scaling = 0
Lookahead = 10
Macroblock Mode Cost Method = default
Motion Match Method = -1
Motion search Level = -1
Motion search range = -1
Motion Vector Coding method = 1
NumBFrames = 2
Perceptual Option = 1

Now do realise that I set the encode to 1000kbps... which for that resolution is a bit low, so I would expect the codecs not to appear perfectly. Secondly, rain is hell for any codec. Even the x264 encode had quite a bit of trouble with the source. So don't worry about the results, this just shows that when extremely stretching the codec x264 behaves better than VC-1. That doesn't mean VC-1 won't or isnt great at anything. I'm sure with a tad more bitrate, or with a less complicated scene, VC-1 would do just fine.

I remember just a year ago when 1500kbps /w Xvid was my bare minimum for an encode, and x264 is just such a low bitrate codec... It's fairly unbelievable. If you render x264 at 600kbps it is passable though smeared. I wouldn't expect the same results with VC-1. Though at 2000kbps Id have a hard time convincing you to use x264 over xvid (BTW, I tend to prefer xvid at 2000kbps+) I'd have to run another set of tests for high bitrate to see how VC-1 stands up there. I bet it would do well though.

HookedOnTV

1st June 2006, 17:20

Basically the main problem is that the resulting file can not be played back via the MCE interface.

I put mine in mkv. Not the smoothest process but it works. Use graphedit to mux the wmv to mkv then you can mux in ac3.

crypto

1st June 2006, 19:36

So how about DVR-MS, then - is there any way to "make" one?

TSConverter (http://www.dvbportal.de/projects/tsconverter) can convert TS and MPG to DVR-MS and vice versa. DVR-MS can hold MPEG-1, MPEG-2, DV1/2 video and LPCM, MPEG-1, MPEG-2, AC3 audio. It is basicly an extended ASF container

zambelli

1st June 2006, 20:22

Nope I am entering 06060606 in hexadecimal format. If I open the value back up after I have saved it then the leading zero is gone and it will crash WME9 every time.
I can't repro this. When I enter a hexadecimal DWORD value of 06060606, it correctly shows up as 0x06060606 (101058054). No crashes either. Can you export that reg key to a .reg file and post its contents here?

I think there are some serious bugs around this setting that go all the way back to the original WME9 Advanced profile. Could you try a 1 pass Quality with a range reduction of 06060606? When I try that the file is smaller but it does not really appear to reduce the color range. When I do a two pass unconstrained then the color reduction is obvious and unacceptable. That is why I said that I was getting good results even using a high reduction value with a Quality encode.
Well I'll be... Nice catch, mpgxsvcd! Yeah, there's something fishy going on here. I wonder if the range expansion factor is somehow getting dropped on the 2nd pass. I'll open a bug. Thanks for catching this!

Any help on this setting is appreciated. It is still some what of a mystery as to what it actually is supposed to do?
It reduces the luma and chroma accuracy. For example, instead of 256 levels, luma could be reduced to only 128 levels. If the expansion factor is the same as the compression factor, then those 128 levels will get expanded back to 256 levels on decode, but there will still only be 128 unique luma levels in the image.
You can see that in 1pVBR the file size gets reduced by a lot if you set the value to 0x08080808. This reflects the fact that there's less luma and chroma information in the picture.

zambelli

1st June 2006, 20:57

Here's an example of range reduction:

Save the attached gradient.png image. Create an Avisynth .avs file with the following line:
ImageSource("gradient.png", end=120, fps=24)
ConvertToYV12()

Using the wmcmd.vbs from the WVC1 package, execute these 2 encodes:

cscript WMCmd.vbs -input gradient.avs -output gradient_full.wmv -v_codec WVC1 -v_mode 2

cscript WMCmd.vbs -input gradient.avs -output gradient_rr8888.wmv -v_codec WVC1 -v_mode 2 -v_rangeredux 8 8 8 8

Then queue these two WMV files in a playlist and play them back to back. You'll notice that the video with the range reduction lacks the fine gradient definition of the other one. The steps are more pronounced, the lines between them appear darker.

In this example the file sizes won't be different, but if you try to apply the same example to a proper movie - you'd see the range reduced file sizes decrease significantly.

pwimmer

1st June 2006, 21:32

How do I play WVC1 content on Windows 2000?

mpgxsvcd

1st June 2006, 21:57

I can't repro this. When I enter a hexadecimal DWORD value of 06060606, it correctly shows up as 0x06060606 (101058054). No crashes either. Can you export that reg key to a .reg file and post its contents here?

Well I'll be... Nice catch, mpgxsvcd! Yeah, there's something fishy going on here. I wonder if the range expansion factor is somehow getting dropped on the 2nd pass. I'll open a bug. Thanks for catching this!

It reduces the luma and chroma accuracy. For example, instead of 256 levels, luma could be reduced to only 128 levels. If the expansion factor is the same as the compression factor, then those 128 levels will get expanded back to 256 levels on decode, but there will still only be 128 unique luma levels in the image.
You can see that in 1pVBR the file size gets reduced by a lot if you set the value to 0x08080808. This reflects the fact that there's less luma and chroma information in the picture.

Thanks for the info Zambelli! I will do some more testing with this feature. I am convinced that this is part of the reason that X.264 looks so good at stupid low bit rates. I can see the black crush in every single X.264 and H.264 example I have. Sometimes it is actually desirable because it can make the image appear sharper without that much loss of detail. Almost every single rear projection microdisplay TV(DLP, LCD, LCOS) are all set to a vivid or “Torch” mode by default. This almost always introduces black crush and yet everyone claims these TV’s look great. And I guarantee every single novice user will claim that the pro mode(which has the correct color range) looks much worse than the vivid mode because the vivid mode appears sharper!

Yes the dword shows the correct value of 0x06060606 (101058054) when I enter it and hit ok. However, the problem comes up when I open that value back up in the registry. The leading zero is gone. If I hit "ok" without manually typing the zero back in there then WME9 crashes everytime.

bond

1st June 2006, 22:28

How do I play WVC1 content on Windows 2000?
for using the codec you need to install the windows media sdk version 9.5, which is only available for windows xp

you can also try to install the old version 9.0, but well, its not really tested and might not work correctly...

and yeah, it also seems to install all the drm stuff

foxyshadis

1st June 2006, 22:35

Thanks for the info Zambelli! I will do some more testing with this feature. I am convinced that this is part of the reason that X.264 looks so good at stupid low bit rates. I can see the black crush in every single X.264 and H.264 example I have.
That is a display/video card/driver issue, not an AVC feature. AVC encoders will not and cannot destroy the sample bit depth that way, but depending on your decoder, renderer, card, and driver, you may get a TV->PC transform or PC->PC when converting back to RGB for output. What you're seeing is input PC levels put through a TV->PC transform, which is almost as bad as the washed out grey look of TV through PC->PC. The fact that it cannot be signaled in the bitstream or controlled in the driver (other than randomly changing rendering modes or output colorspaces) turns it into a huge mess. Many threads exist on this confusion and ways to correct it.

That said, if you get proper output without having to configure anything this way, great.

zambelli

1st June 2006, 23:41

How do I play WVC1 content on Windows 2000?
If you have WMP9 installed with automatic codec download option enabled (default), playing back a WVC1 video will trigger the decoder download.

pwimmer

3rd June 2006, 11:40

If you have WMP9 installed with automatic codec download option enabled (default), playing back a WVC1 video will trigger the decoder download.

Automatic downloads were disabled...

But I still can't play a file protected by DRM 10. Is there a way to make it work (on Windows 2000)?

zambelli

3rd June 2006, 23:00

But I still can't play a file protected by DRM 10. Is there a way to make it work (on Windows 2000)?
I don't think there is at this time. According to http://www.microsoft.com/windows/windowsmedia/forpros/drm/sdksandversions.aspx, only XP and Win Mobile 2005 devices support DRM v10. I'm pretty sure Server 2003 SP1 supports it too.

zambelli

6th June 2006, 02:46

Well I'll be... Nice catch, mpgxsvcd! Yeah, there's something fishy going on here. I wonder if the range expansion factor is somehow getting dropped on the 2nd pass. I'll open a bug. Thanks for catching this!
@mpgxsvcd:
Good news. I checked this with the latest encoder build and the problem appears to have been already fixed since the beta.

mpgxsvcd

6th June 2006, 04:19

@mpgxsvcd:
Good news. I checked this with the latest encoder build and the problem appears to have been already fixed since the beta.

Awesome! Can't wait for the new encoder. I know the range reduction setting is not really the ideal way to get the bit rate down. However, I am trying to get to some very low bit rates and the range reduction was still yeilding some acceptable results at those low rates.

mpgxsvcd

8th June 2006, 16:24

Here is an awesome site on how to use Movie maker!

http://www.papajohn.org/
Just go to www.papajohn.org > Saving Movie > Custom WMV Profiles and it explains how to setup a custom profile.

Here is a WMV9 Package.zip file that contains 4 custom .prx profiles and 3 custom Advanced codec settings. There is also a wmv9ase-1.0.2.5-bin.zip file that contains the program that allows you to easily edit the WMV9 Advanced codec. Just unzip the files in the wmv9ase-1.0.2.5-bin.zip file. Then open up the wmv9ase.exe file. You can import any of the .xml WMV9 settings files from the WMV9 Package.zip file and then hit apply. Then just place the .prx files from the WMV9 Package.zip in the C:\Program Files\Movie Maker\Shared\Profiles directory and you should be able to access them through Movie Maker. Let me know how your results turn out?

WMV9 Package.zip
http://www.avsforum.com/avs-vb/attachment.php?attachmentid=58758

wmv9ase-1.0.2.5-bin.zip file
http://microsoftuse.temp.powweb.com/mcored/downloads/wmv9ase/wmv9ase-1.0.2.5-bin.zip

crypto

8th June 2006, 19:11

Thanks, sounds interesting.

zambelli

8th June 2006, 21:50

Let me know how your results turn out?
It might be a matter of personal preference, but I find your settings to be too aggressive in filtering. That works well with low bitrates because it improves compressability and reduces the chance of macroblocking. However, for those who are interested in preserving detail and have maybe a little more bitrate to spare, they should try these settings:

"Motion Search Level" = 2
"Motion Search Range" = -1
"Motion Match Method" = 1

Somewhat faster settings would be:
"Motion Search Level" = -1
"Motion Search Range" = -1
"Motion Match Method" = -1

On top of that, one should experiment with "Dquant Option" and "Dquant Strength", and also "Force B Frame Delta QP" if they find that quality suffers too much in B-frames. "Force LoopFilter" can be helpful, depending on bitrate available, but it generally will somewhat reduce detail.

Here's an idea: if anyone has a royalty free video clip (high quality - no DVD sources), we could start a "competition" of sorts in a new thread and see who can get the best quality out of WVC1 with the particular video. Quality could be judged by voting (subjective) and with PSNR/SSIM metrics (objective).

Sagittaire

8th June 2006, 23:32

.

Here's an idea: if anyone has a royalty free video clip (high quality - no DVD sources), we could start a "competition" of sorts in a new thread and see who can get the best quality out of WVC1 with the particular video. Quality could be judged by voting (subjective) and with PSNR/SSIM metrics (objective).

Something like Open Source Movie ... ???
http://orange.blender.org/

Here little 720p multichanel sample
http://multimediacom.free.fr/Video/Sample_720p.mp4

If you have high BP serveur you can perhaps make 1080p source master with lossless video codec ... why not.

benwaggoner

9th June 2006, 01:07

Something like Open Source Movie ... ???
http://orange.blender.org/

Here little 720p multichanel sample
http://multimediacom.free.fr/Video/Sample_720p.mp4

If you have high BP serveur you can perhaps make 1080p source master with lossless video codec ... why not.
The Blender guys have offered us the OpenDXF suource for their Elephant's Dream movie, if we can find somewhere to host 100 GB...

zambelli

9th June 2006, 01:54

The Blender guys have offered us the OpenDXF suource for their Elephant's Dream movie, if we can find somewhere to host 100 GB...
How long is the movie anyway? If we picked a 3-5 minute long scene, that'd probably be sufficient. 1080p 24fps, compressed with HuffYUV... It'd probably be about 5GB. We could probably just share a torrent instead.

benwaggoner

9th June 2006, 02:13

How long is the movie anyway? If we picked a 3-5 minute long scene, that'd probably be sufficient. 1080p 24fps, compressed with HuffYUV... It'd probably be about 5GB. We could probably just share a torrent instead.
They've offered source in other forms as well. Probably should make an agreed-upon optimal 8-bit 4:2:0 from it, so we don't spent months arguing about dithering :).

zambelli

9th June 2006, 09:23

They've offered source in other forms as well. Probably should make an agreed-upon optimal 8-bit 4:2:0 from it, so we don't spent months arguing about dithering :).
Do you know how to render the video from their source project?

Sagittaire

9th June 2006, 10:08

Do you know how to render the video from their source project?

They make movie with Open Source Blender Software
http://www.blender.org/cms/Home.2.0.html

All Production files are done with DVD or here (be carefull more than 5 Go for complete download)
http://video.blendertestbuilds.de/index.php?dir=download.blender.org/ED/DVD1/production/
http://video.blendertestbuilds.de/index.php?dir=download.blender.org/ED/DVD2/production/

Complete movie duration is 10 min 30 sec in Pal version.
5.1 audio track, 5.1 audio commentary, more than 40 languages subtitle are available.
Something like 6 Go for 1080p YV12 4:2:0 complete movie with good lossless video codec.

mpgxsvcd

9th June 2006, 13:48

Absolutely, I agree. My settings are extremely aggressive and yes I am trying to get some really low bit rates for 480p. That is what most people are trying to achieve right now. Pretty much all of the comparisons are done with DVD sources. We need to be able to stream medium quality(480p 400-800 Kbps) video over the net. I know that WMV9 is designed for 6-8 mb/sec HD but I have actually got some acceptable results at streamable bit rates. Basically I am trying to host my videos from my digital camera. They are native 480p 4:3 at 30 FPS. The original files are at 16 mb/sec and I am trying to compress them down to 1/20-1/40 their original size. I have actually gone back and re-encoded all of my original files at the same bit rate that I originally tried with regular wmv9. The difference in quality is amazing! The old ones are worthless compared to the results I get now. Anyway, thanks for your help. We can always use pointers when we are going awry.

It might be a matter of personal preference, but I find your settings to be too aggressive in filtering. That works well with low bitrates because it improves compressability and reduces the chance of macroblocking. However, for those who are interested in preserving detail and have maybe a little more bitrate to spare, they should try these settings:

"Motion Search Level" = 2
"Motion Search Range" = -1
"Motion Match Method" = 1

Somewhat faster settings would be:
"Motion Search Level" = -1
"Motion Search Range" = -1
"Motion Match Method" = -1

On top of that, one should experiment with "Dquant Option" and "Dquant Strength", and also "Force B Frame Delta QP" if they find that quality suffers too much in B-frames. "Force LoopFilter" can be helpful, depending on bitrate available, but it generally will somewhat reduce detail.

Here's an idea: if anyone has a royalty free video clip (high quality - no DVD sources), we could start a "competition" of sorts in a new thread and see who can get the best quality out of WVC1 with the particular video. Quality could be judged by voting (subjective) and with PSNR/SSIM metrics (objective).

mpgxsvcd

9th June 2006, 14:00

Here is a thought. Maybe we could get someone with about 2 minutes of HD content from one of the Sony HDV camcorders to let us use that as the source? Their high end camcorders actually produce a broadcast quality 1080i HD image. I think HBO uses them for some of their HD shows. Here is a forum that has some original quality source files. This might be a good test for the interlacing also, since that is one of the biggest features of the new codec.

http://www.sonyhdvinfo.com/forumdisplay.php?f=26

benwaggoner

9th June 2006, 19:59

GodofaGap

9th June 2006, 22:57

http://www.highdefforum.com/showthread.php?t=6537

Here is a whole lot of HD sources. There are also some raw YUV 1080p and 720p samples which are all not copyrighted. Perhaps a few could be pasted together (and maybe resized with a simple avs script to normal proportions for low-end users).?

makoto916

10th June 2006, 06:07

I just happen to be editing an HD project that was shot on a Sony HDV-HC1. I exported a 2minute Quictime DV file from Avid and compressed it in 7Z to save my server some bandwidth.

The resolution is 1920x1080 but this is not the most exciting footage so you likely not find it the worlds best test candidate. But it is HD and it was shot from a Sony camera.

I'm only going to leave this file on my server for a couple of days as I don't want to leave a 250 meg file up for long. But I'll gladly FTP it to someone if you need it after that time. Just PM me here or IM me.

[LINK REMOVED - PM me if you'd like it]

diogen

11th June 2006, 20:37

I'm having some really strange result using the latest AP VBR. And they are not pleasing.

All sources: ~11.75 Mbps MPEG-2 1280x720/60p (satellite hidef) with 384 kbps DD 5.1
Compressing: AP (beta) 2-pass peak constrained VBR + 256kbps CBR WMA Pro (using the DMO only, no WMP 11).
Options: mostly as mpgxsvcd's above (encoding speed is very close to 1/10th real time, AMD64 4000+).

It worked fine on Meet Joe Black (3h movie, no action).
When I tried next Stealth (2h), I noticed something strange: on the second pass up to about 55-60% the bitrate was TWICE the defined average,
i.e. about the peak bitrate allowed. But the final result (file size) was as expected, i.e. the second half was on a real bitrate diet.

I didn't have time watch the final WMV and run the same encoding parameters with Revenge of the Sith (2h20min).
On both jobs, fitting into SL DVD (V: aver/peak 4100/9000 kbps; A: 256kbps) and DL DVD (V: aver/peak 8300/17000 kbps; A: 256 kbps) the same happened:
high bitrate during first half, very low bitrate on second half. The final fight looks plain awfull, macroblocking everywhere.
Divx3 encoded DVDs 2h per CD look better.

Going back and watching Stealth I saw exactly the same.

Am I doing something wrong?
Ben, zambelli - any comments?

Diogen.

EDIT: Installed the WMP 11 and will try it one more time.

diogen

13th June 2006, 17:51

Using the WMP 11 version of the DMOs, everything works the way it suppose to.
There must be some differences in the way old WME works with the standalone DMOs and ones that come in WMP11.

Diogen.

zambelli

13th June 2006, 19:31

There must be some differences in the way old WME works with the standalone DMOs and ones that come in WMP11.
I'll look into it, but it's more likely that the bug afflicting the standalone DMO got fixed in the v11 DMO. Thanks for the bug report - I'll check it out.

easy2Bcheesy

19th June 2006, 09:04

I did my first Advanced Profile encodings yesterday on my Xbox 360 720p/60 footage and I was very pleasantly surprised by how good it was. Even on demanding scenes, 11mbps provided a superb level of playback. Up against my 8mbps encodes at half the frame rate (for the Live Marketplace), the difference was extraordinary. Even up against a 30mbps MPEG2 transport stream, the AP video looked so much better.

The fact that the codec autodownloads when the first stream is played means a LOT to my clients who are not techno-literate and want to double click on a video and watch it without having to install AVC decoders or whatever.

However, what did surprise me was how poor MP10 was with dropped frames and stuttering. Run the same streams in Media Player Classic and they play flawlessly.

I observed the same gulf in performance on both my systems:

Athlon X2 4400 desktop
2.0GHz Merom laptop

I was disappointed to read in the Studio Edition thread that .vc1 elementary streams will not be supported. Speaking as someone about to take the leap from SD to HD disk production, I want to use this technology, so why not now?

At the very least, why not release an AP demuxer if Media Encoder produces VC-1 compliant video?

Although this technology will work with both Blu-Ray and HD-DVD, Microsoft is clearly aligned with HD-DVD and could do the format a hell of a lot of favours by getting tools out there. Right now we have the situation where the Toshiba HD-A1 plays non-professional HD content from DVD, but Blu-Ray doesn't. This is a massive opportunity IMO.

mpgxsvcd

19th June 2006, 13:55

Microsoft is clearly aligned with HD-DVD and could do the format a hell of a lot of favours by getting tools out there.

That’s funny I thought Microsoft was clearly undecided? Aren’t they positioning themselves so that either format could use their codec? Unlike Toshiba which has bet the bank on HD-DVD and Sony which is going to be practically giving away their BLU-Ray PS3 consoles. To me it looks like M$ is just waiting till HD-DVD fails(Yes it will fail. NO Content!) and then they will say it wasn’t our fault they just didn’t have enough titles. Then shortly after that they will back BLU-Ray.

easy2Bcheesy

19th June 2006, 15:58

mpgxsvcd

19th June 2006, 16:42

Microsoft clearly support HD-DVD enough to provide support for it on Xbox 360. Why not Blu-Ray too if they're hedging their bets?

With regards the failure of HD-DVD, well right now they have a machine that provides better HD playback vs Blu-Ray for half the price (according to 'real' customers on the AVSforum who own both)... my personal view is that when the Chinese can produce dirt-cheap HD machines ($199 and lower), then the contest will really begin.

Content, Content, Content. That is why HD-DVD will fail. And you are right Microsoft is backing HD-DVD for now. Just wait till HD-DVD has gone the way of DIVX and BETA. They will be right there trying to jump on the Blu-Ray boat.

There are not any Blu-Ray players available in the US right now, correct? I thought the first Samsung player will not be out for a few weeks?

Farhad

19th June 2006, 18:00

Zambelli,
I was wondering if you could help me to solve a problem I am
having in playing back Xvid (MPEG-4) video in .avi file format
with Windows Media Player 10 (WMP10) using ffdshow plugin.
I have downloaded , built and installed the latest version of ffdshow from SourceForge.net. Unfortunately only the audio
is played back and the video is all garbage on WMP10 display.
Just before audio playback, there was a momentary error
message saying: "Connecting......Error Downloading Codec"
Could you please suggest what I could do to solve this
problem. I have already looked into
http://msmvps.com/blogs/chrisl/archive/2004/08/05/11318
.aspx in this regard but could not formulate a specific
solution. The installation steps that I used were: (1) I
registered ffdshow.ax in windows registry and (2) placed
all generated dlls in windows path. Am I missing some
installation step? Do I need to install the libavcodecs by
following some additional installation
steps? I am very confused and would appreciate any help or
feedback.

Thanks,

Farhad

zambelli

19th June 2006, 19:44

I'm gonna fill in for the moderator for a second and point out that everyone's way off topic. :)
(cracks whip)

Sharktooth

20th June 2006, 01:18

for the first time i completely agree with zambelli:D
since there were already too much hijackings (sometimes by me too - sorry) in Windows Media and VC-1 related threads, please stay on topic. the discussion in this thread is about Windows Media Video 9 Advanced Profile. If you have problems not related to the topic please :search: to find the appropriate place to post them and if you cant find anything related create a new thread.

bond

20th June 2006, 17:47

split

stickies should better not be used for discussions but be for-info-only :)

diogen

20th June 2006, 20:29

Content, Content, Content... I do believe that content will be the ultimate judge on who wins this war/race. But I disagree that ...That is why HD-DVD will fail.What prevents BD/HD go the DVD-A/SACD route: very nice sound and nobody gives a sh*t ("nobody" as in 1-2%)?
How long will it take
- for the critical mass of households to get HDTV sets?
- to learn that a composite cable doesn't carry hidef?
- to really notice the difference between hidef and upconverted DVDs?
- to try existing hidef (cable/Sat/OTA) and crave for something even better?

I think it will take years and BD/HD have a very narrow time window for this to happen _before_ something better shows up.
And I think studious will jump ship in a heartbeat as soon as one of the formats shows more promise. Today's preferences really don't mean much.
Yes, content will define the winner. But content will join that format that shows the winner's attitude first (as in being profitable).

Diogen.

sisman2000

21st June 2006, 16:05

Hi, I'm new to the forum and i have a few questiona about all this HD/VC1 stuff, if any willing soul could address some of my questions...

1. I just got a Sony HC3 HDV camera and recorded high def 1080i footage on it and would like to store it somewhere for future use on HD-DVD/Blue Ray. Now my idea was to convert every hour of my 1080i footage into 4.7GB VC1 files in 1080i to fit on good old DVD-R so i can later use that footage when HD-dvd burners are ready for me to master those disks. I was hoping the Studio Edition with VC1 would do the trick but i see a lot of comments on how VC1 is not HD-DVD compatible? what does that mean?
2. Will i not be able to put a DVD-R (4.7gb) with one VC1 1080i file on it and play it in HD-DVD players?
3. what is the "elementary stream" and how does that prevent me from making HD-DVD player compatible files with VC1
4. Does anyone have the Toshiba HD-DVD players and will they recognize a single VC1 stream file on an old DVD-R disk (sort of like divx players can read any files burned on an DVD disk, no menu needed...)?
5. anyone have better ideas of how i should store my HD footage to be HD-DVD compatible in the future? should i keep it on my DV tapes for a year until HDDVD burners and mastering tools are out there, and be satisfied with downconverted videos to Dvd Mpegs for viewing...?

Thanks guys!

acidsex

22nd June 2006, 00:26

I think it would be great if Zambelli or Ben could suggest what users can do right now to prepare their content for future authoring on HD DVD. Its understood that we cant do any auhtoring right now but why wait for the long encodes later on when we can start encoding now and then only have to mux it later on?

Can either of you guys recommend settings for us to use now to prepare us for authoring in the future when that does become available?

makoto916

22nd June 2006, 01:28

acidsex, I posed this very question in this thread (http://forum.doom9.org/showthread.php?t=112173) but did not receive any responses. I have a whole ton of encoding that I want to do with the potential of making discs down the road.

I imagine that there isn't a whole lot of solid information on this yet. However we should take this discussion to that thread (or some other) so this topic can go back to just be codec related information.

crypto

22nd June 2006, 07:00

I checked the samples in the HDDVD starter kit and they seem to use plain wmv. They play on the emulator, so my guess is: WMV as a container is supported.

vsv

22nd June 2006, 10:35

I checked the samples in the HDDVD starter kit and they seem to use plain wmv. They play on the emulator, so my guess is: WMV as a container is supported.

It is possible use avc inside wmv container for HDDVD authoring?
If i've understand correctly for HDVD no need mux audio and video
streams? They can be separately? Excuse me for offtopic.
May be to start another tread dedicated iHD authoring
or we can talk there http://forum.doom9.org/showthread.php?t=112617

bond

22nd June 2006, 18:48

It is possible use avc inside wmv container for HDDVD authoring?
If i've understand correctly for HDVD no need mux audio and video
streams? They can be separately? Excuse me for offtopic.
May be to start another tread dedicated iHD authoring
or we can talk there http://forum.doom9.org/showthread.php?t=112617hddvd has nothing to do with wmv container

crypto

23rd June 2006, 00:21

Sample from the valdator (just the beginning and it plays on the emulator):

HD DVD Validator v2.3.0.0
Copyright (c) 2005-2006 Microsoft Corporation. All rights reserved.

Configuration:
Verbosity Level: 3
Warning Level: 3
ECMAScript Support: ECMA-262, ECMA-327 (Compact Profile)

[00:00:00.046] Loading schemas.
[00:00:00.093] | Loading resource://xml.xsd
[00:00:00.093] | | Target namespace: http://www.w3.org/XML/1998/namespace
[00:00:00.093] | Loading resource://Playlist.xsd
[00:00:00.093] | | Target namespace: http://www.dvdforum.org/2005/HDDVDVideo/Playlist
[00:00:00.093] | Loading resource://Manifest.xsd
[00:00:00.093] | | Target namespace: http://www.dvdforum.org/2005/HDDVDVideo/Manifest
[00:00:00.093] | Loading resource://iHD.xsd
[00:00:00.093] | | Target namespace: http://www.dvdforum.org/2005/ihd
[00:00:00.109] | Loading resource://iHDstate.xsd
[00:00:00.109] | | Target namespace: http://www.dvdforum.org/2005/ihd#state
[00:00:00.109] | Loading resource://iHDstyle.xsd
[00:00:00.109] | | Target namespace: http://www.dvdforum.org/2005/ihd#style
[00:00:00.109] | Loading resource://iHDSubtitle.xsd
[00:00:00.109] | | Target namespace: http://www.dvdforum.org/2005/ihdsubtitle
[00:00:00.109] Loading complete.

[00:00:00.124] Reading content.
[00:00:00.124] | Reading C:\pcg\multimedia\pcg\ihd\content\ContentGates\ContentGate2A\HDDVD_DISC\ADV_OBJ\VPLST000.XPL
[00:00:00.140] | | Resource is encoded with Unicode (UTF-8).
[00:00:00.140] | | Validating xml document well-formedness.
[00:00:00.140] | | Parsing xml document.
[00:00:00.171] | | Reading file:///dvddisc/HVDVD_TS/mainmovie.wmv
[00:00:00.171] | | | Validating file magic.
[00:00:00.171] | | Reading file:///dvddisc/ADV_OBJ/trivia/trivia.xmf
[00:00:00.187] | | | Resource is encoded with Unicode (UTF-8).
[00:00:00.187] | | | Validating xml document well-formedness.
[00:00:00.187] | | | Parsing xml document.
...