Log in

View Full Version : Wmv9 Vcm


Pages : [1] 2

ooops!
7th May 2003, 12:41
Iv'e just started conducting some low bit rate 2pass 'video only' tests, using WMV9 VCM.

For those who are not quite up to speed with this particular version of the WinMedia codec (as I must admit I wasn't).

The finished encodes have a .avi file extension and not an .wmv file extension. Which means audio streams, such as .mp3 or aac(.mp4) etc can be added.

And the finished results can be viewed in a variety of players, such as DivX 2.1, RealOne, Zoom etc, instead of just the WM9 player.

Preliminary results are very encouraging. But it's still the slowest codec out there, when it comes to encoding.

CaptainCarrot
7th May 2003, 13:18
Come ON, look at this thread (http://forum.doom9.org/showthread.php?s=&threadid=52308) , or this one (http://forum.doom9.org/showthread.php?s=&threadid=51182) , or how about this thread (http://forum.doom9.org/showthread.php?s=&threadid=49796) ? And there are probably more. Why do you start a new thread for this?

ooops!
7th May 2003, 15:26
Obvious - because there are alot of newbies out there, surfing forums, such as this, who are not as technically minded as you are. They need the information in simple terms.

Sirber
7th May 2003, 16:50
WMV9 is not bitrate-acurate for low-bitrates encodes. I also think WMV9 is based on MPEG4...

ooops!
7th May 2003, 19:04
> Sirber

Hello,
After my brief spell of testing I have already concluded that 'you do' have to set the bitrate higher, to obtain the same file size, when using WMV9 VCM, as opposed to using Std WM9Encoder.

However, this will almost certainly be down to the fact that the WM9Wncoder adds a few 'overhead' bits.

This is yet another reason why, when you test video codec's, you encode the video without the audio.

And you may well be correct about the base code within WMV9. Hell, I have my suspicions that WM9 is really Mpeg4.

Maybe some Microsoft coders can confirm?

Sgt_Strider
7th May 2003, 22:17
Where is this WM9 is base off Mpeg 4 is coming from? I think Amir has mentioned in the past that WM9 is a new technology and that its not base off mpeg 4. Maybe I'm wrong but I'm pretty sure he mentioned that in the past.

Vladdy
8th May 2003, 04:19
I've done about 8 encodes now, using GK to set up my .avs file and bitrate. I make a 2 cd rip with AC3 audio, and all my rips have been perfect file size. I get filesizes of around 1,454,000k which is pretty much spot on for a 2 cd rip. Bitrates vary from 1200-1600k and have yet to re-encode any movie that was oversized.. or even undersized for that matter. For my encodes I do an anamorphic rip and it looks better than my non anamorphic xvid rips. I can't really ask for any more.. Should be getting a DVD burner soon so i can squeeze 3 movies onto 1 dvd-r. No cd swapping at all =)

as you can see, my filesizes have been pretty much spot on.

1,454,145,536 Bicentennial Man.avi
1,455,118,336 Get Shorty.avi
1,454,698,496 Ghost Ship.avi
1,455,288,320 Swordfish.avi
1,454,770,176 The Mummy 2.avi
1,454,743,552 The Sixth Day.avi
1,454,405,632 The Thomas Crown Affair.avi

kilg0r3
8th May 2003, 07:13
@vladdy

how do you do the resizing during playback. is there something like a flag telling the player to resize?

Vladdy
8th May 2003, 07:56
I tell zoomplayer to resize to 1030x436 (for 2.35:1 rips).. which kinda works just nicely for a 1024x768 screen. about 30% more vertical resolution than a 720x304 rip. You just tell zoomplayer to use 'custom' aspect ratio and it stretchs it to fit (in the options you must tell it to use 2.35:1 or 1.85:1) or whatever..

kilg0r3
8th May 2003, 20:07
Can wmv9 be used in ogm?

amirm
10th May 2003, 04:33
WMV9 is not based on any existing technology, MPEG-4 or otherwise. Yes, we are a patent holder in MPEG-4 and helped developed that technology some 6 years ago. But folks have learned a lot about compression since then and you are seeing the results of that learning in WMV9.

As for the question regardind padding, indeed, this can occur on short clips. You can see the effect of it by zipping the file. If it gets smaller, then you know it is zero padded. On longer clips though, this should be a noticeable fraction of the total file size.

Amir

ooops!
10th May 2003, 12:26
Well, Amirm has certainly put the Mpeg4 suspicion to bed on Microsofts behalf. Thanks.

Do you if the encoder will get quicker any time soon.

My poor old electricty meter has to spin round for so long, it's gone into orbit!

Apart from this I really like it. And I've had no complaints from clients when they view their promos at full screen.

ChristianHJW
10th May 2003, 15:50
Some news on WMV9 VCM in matroska, together with Vorbis and/or MPC, AAC :

Toff was successful in getting this to work when making 2 identicals encodings, one being AVI the other MKV. He then copied the complete Codec Private data from the AVI and feed this info to the AVIdecompressor, instaed of using the content of KaxCodecPrivate strcture from matroska track header, as he found out that the one from the AVI was 135 bytes while the one in matroska header was only 80 bytes.

In the end he got it to work this way, we are investigating now where exactly the problems is, means

- if VdubMod is reading the Codec Private data from the AVI header inccorectly when writing them into matroska header

- if we read them back incorrectly from the DShow parser

We keep you updated ....

ooops!
10th May 2003, 16:30
ChristianHJW, are there any comparision files available for download. So we can see what you mean?

ChristianHJW
10th May 2003, 17:46
Originally posted by ooops!
ChristianHJW, are there any comparision files available for download. So we can see what you mean?
#define 'comparison' ??

All we want is to play WMV9 from matroska in VfW compatibility mode, and for some stupid reason we currently can't .... :( ...

ooops!
10th May 2003, 22:26
Sorry ChristianHJW,

I should have been more specific. What I should have said is something like:-

Are we able to download any test files. So we (more like I) can see how you have achieved what you have achieved?

What people are able to to with Matroska is leaving me behind somewhat!

ChristianHJW
11th May 2003, 06:06
Originally posted by ooops! Are we able to download any test files. So we (more like I) can see how you have achieved what you have achieved?
My 'test file' is a private cam video from my son, and i wouldn't like to spread it. Simply download VdubMod 1.5.1.1a and encode with WMV9 VCM into .mkv as output format, and you'll have your own test file, because thats what i did. Unfortunately it wouldnt play right now same, for the reasons mentioned above, like my test file doesnt play either.
What Toff had done to make it play can not be spread as a single test file, as he was making a quick hack of our DShow parser to be able to feed the private codec data from the AVI to the AVI decompressor filter when calling the matroska playback graph. You had to be a programmer with indeep DShow experience, like Toff is, to make this, and then your code will only work for this specific example video. The test he has done is no practical solution, it was just undertaken to find out where the problem is.
What people are able to to with Matroska is leaving me behind somewhat! .... same for me :D .... the guys are awesome, if you will have a look at the new DShow parser filter ( to be released this weekend ) and compare it to the last version, you will be stunned by the development speed in whick they develop matroska further ... :) ...

shlezman
11th May 2003, 07:26
Originally posted by amirm
WMV9 is not based on any existing technology, MPEG-4 or otherwise. Yes, we are a patent holder in MPEG-4 and helped developed that technology some 6 years ago. But folks have learned a lot about compression since then and you are seeing the results of that learning in WMV9.
Amir
M$ also has it's hands deeply into the h.264 which is based exactly on the new stuff learned in the past 6 years. It seems to me that M$ took the good from both (MPEG4/H.264) and mixed up a new codec with some stuff that only a company with gr8 pair of "Kahones" can afford to do, then implemented it in the only way M$ knows, slow and heavy :(

ChristianHJW
12th May 2003, 00:04
WMV9 VCM playback from matroska files, with Vorbis or MP3 ( VBR/CBR ) audio, is working fine now with the new DirectShow parser, available from http://www.matroska.org/announce.html

Sirber
14th May 2003, 16:35
@Ooops!

I've just recieved the samples. I'll check them shortly. Thanks again :)

mingcl
14th May 2003, 19:17
Originally posted by ooops!
Well, Amirm has certainly put the Mpeg4 suspicion to bed on Microsofts behalf. Thanks.

Do you if the encoder will get quicker any time soon.

My poor old electricty meter has to spin round for so long, it's gone into orbit!

Apart from this I really like it. And I've had no complaints from clients when they view their promos at full screen.

You might want to try the performance bar on the compression page. You might see some quality loss for the faster encoding modes. But the loss is content dependent, and you might not see the loss for some contents.

Also, if you don't care about the file size so much and instead you want constant quality throughout the clip, 1-pass VBR is a good choice too. This will make the encoding much faster because of 1-pass.

Thanks,

Ming
Microsoft

ooops!
14th May 2003, 19:58
Thanks Ming,

Personally I have infact, tried 'all' the sliderbar settings using the same 720x576, 1min 17sec 'video only test file'. With 1pass and 2pass at just 630kbps.

And I am amazed at the difference in encoding speeds.

At the lowest setting, the 1st pass took 1min 45secs and the 2nd pass about the same.

At the highest setting, the 1st pass took 1min 45sec and the 2nd pass took 12min 45secs!

If I encoded my 137min 'test film' using this setting it would take around 22hours to complete the second pass!

I have to admit that when encoding at such low bit rates, using the highest setting produced outstanding results.

I still maintain the application requires a shot of adrenalin!

Cheers.

Sirber
15th May 2003, 11:40
I watched parts of your test 137min CD. The overal quality is nice, with few squared. Did you encode it in 1024x at ~600kbps?

kilg0r3
15th May 2003, 11:53
@Sirber
don't give your Rv9 soul away for WMV! :eek: ;)

Sirber
15th May 2003, 11:59
My best choice for codec is RV9. But oops! sent me a 137min test of WMV by mail, so I checked it. :)

ooops!
15th May 2003, 21:20
> Sirber

Hello,
The video size = 720x576.
Codec = Std Microsoft WM9 Encoder (Not WMV9)
Video codec = 615kbps + overhead 2passVBR
Audio codec = 64kbps + overhead 2passVBR

In order to see the image to be displayed correctly, you must have the latest WinMedia player installed.

Since creating the CD-R version that you now have, I have been experimenting with the WMV9VCM codec, hence the original post.

I can confirm that the quality of this codec is superior to WM9. When I have concluded the test I will forward you a copy 'video only' so you can see for yourself.

The only problem is the encode took 3hours 21mins to complete the first pass and over 22hours to complete the second pass.

Sirber
15th May 2003, 21:35
Using MPC, the output is in 1024x for the wmv. Any clue why?

CaptainCarrot
15th May 2003, 21:45
I can confirm that the quality of this codec is superior to WM9. When I have concluded the test I will forward you a copy 'video only' so you can see for yourself.
:confused:
AFAIK WM9-Encoder and wmv9vcm use/are the exact same codec. So i'd guess there must be some hidden settings tab in wm9encoder that you haven't found yet, cause they should both produce the same quality (at least on codec level, wm9encoder might mess up things before the actual codec takes over).
I haven't done any encodes with wm9encoder because it crashed the first time i started it, and shortly after that the beta-codec was released, so i didn't look into that and simply used vdubmod.

CaptainCarrot
15th May 2003, 21:50
Originally posted by Sirber
Using MPC, the output is in 1024x for the wmv. Any clue why?
I think the wmv-format has an AR-flag just as matroska will have (ok, has, but not supported yet) so the player knows that the 720x576 should be 16:9 and thus plays it at 1024x576. It's just like encoding it at 720x576 with rv9 and telling producer to resize to 1024x576.

ChristianHJW
15th May 2003, 21:54
Originally posted by CaptainCarrot has an AR-flag just as matroska will have (ok, has, but not supported yet)
It is supported already, on both Linux ( mplayer ) and in TCMP on win32 .... just we cant set it from the main win32 app writing matroska files ( VdubMod ) ... :D !

Sirber
15th May 2003, 22:06
Originally posted by CaptainCarrot
I think the wmv-format has an AR-flag just as matroska will have (ok, has, but not supported yet) so the player knows that the 720x576 should be 16:9 and thus plays it at 1024x576. It's just like encoding it at 720x576 with rv9 and telling producer to resize to 1024x576.

Ooops! didn't mention it... Anyway the results where good, with a few sqares. But 24h to encode it, it's kinda long...

mingcl
15th May 2003, 23:27
Originally posted by CaptainCarrot
:confused:
AFAIK WM9-Encoder and wmv9vcm use/are the exact same codec. So i'd guess there must be some hidden settings tab in wm9encoder that you haven't found yet, cause they should both produce the same quality (at least on codec level, wm9encoder might mess up things before the actual codec takes over).


The video quality produced from WMEncoder 9 and WMV9 VCM should indeed be pretty much the same for all settings/modes at the video codec layer. (The final file size might be different because of audio and file container.)

If this helps, an easier way is to match the parameters in the command line mode for both. In VCM, it is in the "parameters" field at the bottom of the compression page. It can be mapped (wih some minor modifications) to the script/command line mode of WMEncoder 9.

Ming
Microsoft

Ramirez
16th May 2003, 01:08
LOL now we have two M$ representatives on the board,:D I guess nobody sent you here either? ;)

Sirber
16th May 2003, 02:42
lol

We will need another one from Real then :D

*** Just kidding, don't take it personnaly

ChristianHJW
16th May 2003, 04:44
Originally posted by Ramirez LOL now we have two M$ representatives on the board,:D I guess nobody sent you here either? ;)
LOL !! Maybe my April 1st joke about matroska container being sold to IBM scared some people at M$ and Real ( the possibility alone ) ... in any case, welcome mingcl :) !

mingcl
16th May 2003, 05:27
Thanks a lot for the nice and warm welcome:) I look forward to having more discussions with you all.
Thanks,

Ming
Microsoft

ooops!
16th May 2003, 10:45
Thanks again Ming.

I'll make a note of the command line code.

If you had come to the forum earlier you may have read some of my first posts with Sirber and kilg0r3.

As we were amongst, the many people trying to settle, the 'which codec is better at low bitrates' debate.

I have found after extensive testing of my own, both at home and at work, that they each codec can out perform another during certain conditions. Infact there's a very detailed report about this if you are interested here: - http://www.doom9.org/index.html?/codecs-103-1.htm

Like many other on the forum I used to formulate my comparative tests using Matrix as a source. I have since changed my source to StarWars 2 (too much green and not enough colour).

However, when I conduct my tests I encode the 'video content only'. I also encode at full frame 720x576 (yep I live in the land of PAL) and use 2passVBR.

The main reason for encoding the 'video content only' is because I've found that one manufacturers codec at say 615kbpsVBR 2pass does not generate the same file size as another manufacturers codec at 615kbpsVBR 2pass. For instance WM9 at 615kbps generates the same file size as WMV9 at 619kbps (full bar), RM9 at 626kbps and DivX at 629kbps.

The differences may only be slight but when encoding at such low bitrates it could make quite a difference.

I would be interested to know if more people have experienced variances with the finished file size, after setting different codecs to the same bitrate speed.

The bitrate to file size differences become even greater when you encode true 16:9 (1.77:1) images.

CaptainCarrot
16th May 2003, 17:04
Has anyone of you ever tried to encode the credits of a movie seperately with low constant quality and then give the rest of the movie a higher bitrate? Cause I got the Movie and the Credits-file now, but vdubmod won't append the credits to the movie and tells me that the video streams have a diffrent data format :sly: . Any ideas?

ooops!
18th May 2003, 19:04
> CaptainCarrot

Yep, I've created a couple of movies with the credits at a different bitrate. I can't remember what application I used to join the two files together.

However, if it's your intention to put the whole lot onto a CD-R I would suggest not to bother joining them at all - Just burn them as separate files!

CaptainCarrot
18th May 2003, 19:12
However, if it's your intention to put the whole lot onto a CD-R I would suggest not to bother joining them at all - Just burn them as separate files!
That would mean that I'd have to split the audio files at the correct positions as well, and I don't wanna go through that. Besides, joining has worked for me so far with any other codec i used. I think I'll experiment a little with diffrent encoder settings.

ooops!
18th May 2003, 19:27
Eeee Gods!

What application did you use to extract you .vob files and what application are you using to create your encodes.

Also what are your finished encodes. DivX, Xvid, WMV9?

CaptainCarrot
18th May 2003, 20:42
Originally posted by ooops!
Eeee Gods!

What application did you use to extract you .vob files and what application are you using to create your encodes.

I usually do all my preparations with gknot, and then open the *.avs-files with vdubmod if i want to do something gknot can't

Also what are your finished encodes. DivX, Xvid, WMV9?
I use all of them, but since this thread is about WMV9, that's what i was talking about. See, maybe you got me wrong, but i have already done a lot of encodings with DivX, some with XviD, some RV9 (with diffrent tools, of course) and some with wmv9. With the first three mentioned i have no problems to encode the credits seperately at a lower constant quality and then merge the movie with the credits-file, and dixv and xvid should work exactly the same way as wmv9. Now my problem is that a movie encoded with wmv9 and 2-pass variable bitrate can not be merged with the credits encoded at 30% constant quality. Since this was actually a test for the diffrent codecs, i used the exact same *.avs files for divx and xvid-encodes as well, and with those everything works, whereas with wmv9 it doesn't. And now i want to know wether that's a wmv9 problem or i did something wrong.
So if you have already tried to merge a 2-pass vbr wmv9-avi with a xx%-constant quality encode and it worked, or didn't work, then that's something i'd like to know about.

ooops!
18th May 2003, 21:25
When I get back from work tomorrow I should have an answer for you.

Can you check and post the following file 'properties' information for both of your files. I just need to know the 'Image' pixel width & height sizes and the 'Frame rates'

Thanks

CaptainCarrot
18th May 2003, 21:48
Both files are 512x208 at 25 fps. The file information is identical. I just tried to encode the credits at 60kbit/s with 2-pass-vbr but couldn't merge them either. I'll try to reencode the main movie, maybe something went wrong with that.
If i copy the credits-file i can merge those two files.
Merging also works for the movie file and a copy of it.

CaptainCarrot
18th May 2003, 22:22
AHA! Got you, WMV9! You're actually (at least) 2 diffrent codecs pretending to be one!

I reencoded my credits part at quality 80, and the encoding speed dropped to 10-20% of the "quality 30"-speed, an when I try to merge those two i get that error again. So wmv9 for low bitrates is not compatible with wmv9 at high bitrates. That means seperate encoding of the credits at a lower bitrate is not an option so far. Hey matroska-ppl, did you hear this? I said "so far" :D . Didn't I read somewhere that mkv could contain pretty much anything? Is that true for diffrent video codecs in one mkv-file, too (even if they pretend to be the same codec)? Would be nice, cause this would be one thing that could only be done with mkv.

slavickas
19th May 2003, 10:47
Originally posted by CaptainCarrot
AHA! Got you, WMV9! You're actually (at least) 2 diffrent codecs pretending to be one!

I reencoded my credits part at quality 80, and the encoding speed dropped to 10-20% of the "quality 30"-speed, an when I try to merge those two i get that error again. So wmv9 for low bitrates is not compatible with wmv9 at high bitrates. That means seperate encoding of the credits at a lower bitrate is not an option so far. Hey matroska-ppl, did you hear this? I said "so far" :D . Didn't I read somewhere that mkv could contain pretty much anything? Is that true for diffrent video codecs in one mkv-file, too (even if they pretend to be the same codec)? Would be nice, cause this would be one thing that coud only be done with mkv.

well i'm not sure but maybe other profile was used at low q

ChristianHJW
19th May 2003, 12:02
Originally posted by CaptainCarrot That means seperate encoding of the credits at a lower bitrate is not an option so far. Hey matroska-ppl, did you hear this? I said "so far" :D . Didn't I read somewhere that mkv could contain pretty much anything? Is that true for diffrent video codecs in one mkv-file, too (even if they pretend to be the same codec)? Would be nice, cause this would be one thing that could only be done with mkv.
You would have to wait a bit until our so called 'control tracks' are implemented, but then yes, np. We would handle this as 2 different video streams, and they'd be concatenated in one file and the control track would tell the player to first play the first, then the 2nd video stream, as if it was just one file. This feature was normally planned to allow using the same intro/outro scene if you encode several episodes of star treck on one CD, so it would be abuse for this purpose, but nevertheless work fine, and even gapless( or almost, depending on player speed ).

As you had 2 entirely different video streams here it also wouldnt be a problem if both use the same codec with the same FourCC, as both has their own headers so WMV9 could initiate correctly in both modes ....

ooops!
19th May 2003, 19:57
Hello again CaptainCarrot,

Yep, this is wierd, we tried at work to merge/join a VBR and a CBR file together and came across similar problems. Maybe you can't merge VBR with CBR WMV9 files - very wierd.

We had a little more success merging/joining 2No VBR files together (one 6min file at 630kbps and one 2min at 200kbpsVBR).

This is going to need a little more research. Maybe an application like EO Video might do the trick.

Just as a matter of interest, what movie are you encoding and what were your 2 finished file sizes. Oh, and approx how long did each one take you to encode. As this could be just the right kind of task for our new 'office slave' to get to grips with.

However, I will ask her, yes her, to try encoding at full frame (720x576 PAL). It should keep her quite for a while!

Over the weekend we encoded StarWars 2 (137min) again using WMV9. This time we set the video codec to 620kbps 'full bar' 2pass VBR and added basic Mp3 audio at 56kbps. The frame size was our usual 720x576.

God damm, did it look good. We were all very supprised. It looked much better than WM9. And it still all fits on a 700MB CD-R.

Amazing!

CaptainCarrot
19th May 2003, 22:58
The movie I encoded was skorpion king for 1 XCD (795MB) with AC3, so the video had 551MB. The movie file came out 547MB, but I don't remember the exact bitrate i entered. The credits came out 2,47MB with 30% quality and a little bit bigger with 60kbit/s.

I remembered that amirm posted somewhere that wmv9 has diffrent encoding features popping in automatically at certain bitrates, that's probably the problem problem.

It looked much better than WM9

I still don't understand that, since it's the same, just the container is diffrent.

I won't have time to look into anything really for some weeks, i'm too busy studying. Hope your "Office Slave" has some success testing.

ooops!
20th May 2003, 18:47
Yes, I can't understand why the 'full bar' WMV9 encodes look better than WM9 but they do!

Maybe somebody at M$ can let us know which settings we have to use in each encoder, to create files that look the same?

Anyway the proof is safely backed up on CD-R all 137min of it!