D2SRoBa 3.3.3 - Plugin

[r6d2]

Originally posted by pacohaas
When the CCE (2.67 btw) window pops up, i can see that it says it's doing everything right (Q=1 300-2450), but yet this is the outcome. Anybody know what gives?

According to the lovely search button this has come out before. The anwer is you have to take BR Viewer's numbers with a grain of salt. CCE is right and local BRs are compliant.

BTW, just to clarify, "Pure Q=1" requires MAX=infinity, otherwise if constrained by SVCD specs its outcome will be, as w00kie and DDogg say, more CBR like.

[Trahald]

bitrate viewer is not extremely reliable as far as its bitrate graph goes. it idenifies perfect cbr streams as vbr (ie cbr svcds and even vcds identified as being cbr by everything else).. really the only thing you could use is the average bitrate it gets and take for granted the encoder did not exceed the max during the encode.

i did short tests using the d2sroba sampling avs made during an encode.. results...

stream                        bitrate    bv avg    calc avg
Q 1 @ max 2500 kbps stream    34541 KB    2015     2101.35

CBR 2500 kbps stream          40171 KB    2344     2443.86 

Q 1 @ max 9000 kbps stream    71973 KB    4194     4378.58

CBR 9000 kbps stream         144615 KB    8436     8797.87

bv avg is what bitrate viewer got as average and calc avg is what i got using calc.exe. this is a 3 second clip, so probably not enough full gops for a good test. ill do a longer clip and report back in a bit

[pacohaas]

good to hear, i was just using the tool that was suggested above by DDogg. I want to trust CCE, sure, but watching it encode, i see it go slightly higher than it should at times, and even with a CBR encode, it doesn't show the "current bitrate" as always the same.

But, considering the CCE settings are used and specs not broken, back to my first question, Could there possibly be a checkbox that would just allow CQ encodes such as this one to continue instead of wasting bits on a huge CBR encode?

It would be nice though, to have a tool to inspect the bitrates as BV is supposed to, don't you agree?

[tylo]

Could there possibly be a checkbox that would just allow CQ encodes such as this one to continue instead of wasting bits on a huge CBR encode?

The whole idea with D2SRoBa is to get out the best quality, bounded only by either a fixed number of CDs, or a Q limit. What you ask for is an option not to fill the CD as much as possible under certain circumstances (for some reason). It is not wasting bits when encoding e.g. CBR 2560 vs. OPV max 2560, giving avg 2200, when both fit on same number of CDs. The former will give an overall higher quality encoding. If you don't want to fill you CDs, use some other tool.

[Trahald]

It would be nice though, to have a tool to inspect the bitrates as BV is supposed to, don't you agree?

yerp... if im feeling adventurous i'll play with visual studio (which means i probably wont do it or i'll take forever )

anyhow.. here is the second test results are pretty much the same

stream                        *size      bv avg    calc avg
Q 1 @ max 2500 kbps stream   296591 KB    2015     2165.22

CBR 2500 kbps stream         334757 KB    2342     2443.85 

Q 1 @ max 9000 kbps stream   573169 KB    4009     4184.35

CBR 9000 kbps stream        1205123 KB    8431     8797.85

*my other chart should read size where it says bitrate

this was ~1095 mins worth of video. an accurate bitrate viewer would really solidify the conclusion.. but assuming that some of the frames would need to be above 2500 to hit q1, the conclusion you have to make with a 2165 average is that there are frames also being generated below 2165 (for the math to add up) which would indicate a vbr stream. the 9000 max prooves it further since the filesize came out larger (the frames that were truncated by the 2500 max were allowed the room to hit the bitrate needed for q1)

if anyone can find a good bitrate graph that would be great.

[r6d2]

Originally posted by pacohaas
[B]good to hear, i was just using the tool that was suggested above by DDogg. I want to trust CCE, sure, but watching it encode, i see it go slightly higher than it should at times, and even with a CBR encode, it doesn't show the "current bitrate" as always the same.

This has something to do with BIAS and CBR implementation in CCE. It is not absolutely CBR, just "very CBR".

Surpasing the spec limits momentarily is not unusual. TMPGEnc also does it. But rest assured that VBV is respected, that's what the encoder has to do well.

[telemike]

Tylo-

can you put a ink backup for version 2.70RC2 as the last stable version for newbies to use until version 3 is stable and debugged? Thanks.

[Trahald]

yeah.. r6 put it well.. cbr will just be 'very cbr' any encoder software/hardware obligation is only to keep it vbv safe so there are no overflows underflows within the decided "constant" bitrate.

i got another bitrate viewer. ( http://www.elecard.com/download/#StreamEye <edit - this program installs its own mpeg2 decoder.. install at your own risk>) it picks up the average better than bitrate viewer, but still denotes cbr streams as (slightly)vbr. in fairness, a bitrate viewer will never show a cbr stream as cbr because at a minimum the b frames will TEND to be smaller than I frames (although not always) and encoders have no obligation to make all the frames or even GOPS all the same bitrate.

nice details here:
http://www.gadegast.de/frank/mpegfaq/mpe1736.html

[DDogg]

tylo, or anybody else, where are we with open issues? I don't use the audio stuff or the auto mode or the bitrate adjust so everything seems pretty good to me, but it sounds like that is not the general opinion? If we can list them then maybe they can get checked off when time permits. Now it just seems like there is a bunch of maybe/kinda/sorta stuff floating around.

[Trahald]

I use the audio after (i only ever get one track) and auto mode and the adjust. i dont set vbr pass (the size has been fine with this version so far) ive done an encode a night and no problems

ive been shying away from using disks=X option because i hear there are problems? i just use auto and put the auto q really high
as r6 suggested. tonight i'll try disks=2 and report how it goes.

sofar... no issues myself.

[ric1234]

Also, I would like to propose an update on the status of the followings with D2SRoBa v3 :
- is it now possible to minimize DVD2SVCD ?
- is it possible to iconize DVD2SVCD ?
- is it possible to minimize CCE ?
- is it possible to minimize D2SRoBa log window ?
- is it possible to run the whole in a separate user session in XP ?

Can somebody confirm the statut of these topics ?

[tylo]

I don't think the state is bad at all. I am not able to get any of the problems seen so far, but the reported problems are:

1. Sometimes when choosing a fixed number of CDs, it still compute number CDs. I believe this is an AutoIt3 related problem, but I have no proof of that yet. (reported by ric, r6d2). Earlier the oposite prob also did occur: the Auto mode did not compute num CDs, but that seems to be gone.

2. Problems with breaking CCE. Reported by telemike.

If anyone has seen other probs, please report.

The only thing I can do about these probs so far is to put out a version with more debug output in the log, to get more info about where/why it fails. I'll do that tonight

/Add:

Also, I would like to propose an update on the status of the followings with D2SRoBa v3 :
1 - is it now possible to minimize DVD2SVCD ?
2 - is it possible to iconize DVD2SVCD ?
3 - is it possible to minimize CCE ?
4 - is it possible to minimize D2SRoBa log window ? (/add)
5 - is it possible to run the whole in a separate user session in XP ?

1 - yes, 2 - yes, 3 - yes, 4 - yes, 5 - don't think so.
I'll try to add a minimize option to D2SRoBa for the next (debug-enabled) beta.

[ric1234]

Tylo, minor point but do you plan to do something to allow the use of a second audio track in "Audio encoding after Video" ? I think that was on hold too.

[tylo]

@ric: True. I forgot the 2'nd audio thing. I never use two audio tracks myself, so therefore I haven't tested with it. I am sure it's a small issue to fix, though. (btw: thanks for PM's)

PS: I only want two audio for animation movies (for the kids), like Finding Nemo, and then I create separate copies with different audio for max player compability + better video quality. When I think of it, that could be a great option in D2SRoBa. What do you think?

// It could be done like this: D2SRoBa creates the first audio after video creation, and then recover, as today. After D2S is done, it moves the created files (mp2,mpg, cue,bin) etc. and creates the second audio, recover again from muxing.

/Add: on second though, it may be difficult to get this to work by selecting both audio's through D2S. I think the second audio should be selected from D2SRoBa only, and let it extract that audio track itself... Oh, well, it was just a thought.

[ric1234]

In my case, I have used 2nd audio in some SVCDs when Q was really good and I could afford a 2nd low bitrate audio track for director's comments. So it was kind of a bonus
Personnaly, I wouldn't go for a 2nd set of discs just to have a different language, but some might like it...

[DDogg]

<cough> Maybe I can get away with this without wearing asbestos drawers...Everytime we get close to a stable release, more individual enthusiasms pop up. Nothing wrong with that at all except it seems like it always keeps us from getting a stable release candidate for the average Joe to use without problems.

<putting on flameproof underwear> IMO, and certainly sort of expressed before, adding the audio caused a train-wreck full of nasty old worms that delayed a stable release by nearly two months, IIRC. Personally I have never needed it and I suspect most do not. Is it a good feature? Sure it is. Was it worth all the damn grief it caused? Certainly not to me, and I suspect a ton of others.

I fear I may be sounding a bit prickish, but I have been involved with quite a few commercial devel projects and I sure think there should be a double solid release for the average user and a clearly marked experimental test copy for us that like to play around.

Having only one version available on the web-page tends to cause less experienced users unnecessary grief. I am afraid a lot of people have been put off from this great piece of work because of that.

OK, I got my days dose of grumpy out. Hope nobody minds my normal bluntness /add: I aspire to being called a "curmudgeon"

[tylo]

@DDogg: My main goal is getting this plugin stable, and great for everyone just by using the default settings. I can't agree more with you that it is much better with a stable simple version, than a buggy one with some extra features. V2 never got to a release version because I didn't feel it was bug-free enough. I hope and think V3 will - and soon.

ps: I'm honestly getting a little fed up with it (like I said I would), but I'll try to hang on until it's stable.

[DDogg]

Oh man you definitely get an A+++ grade for perseverance. I think I would have turned into an ax murderer by now if I was doing it (not that I could) Well, like you said, the light is appearing at the end of the tunnel so maybe the end is in sight!

[tylo]

Yeah, thanks - there has been some sleepless nights...

GREAT NEWS: Found the (1) main bug just now: A silly logical error. No more lame excuses like "AutoIt3 is beta", or so. . Well, we're getting there

@telemike: I'll put in some more log in connection with detecting/closing the CCE window.


/Add: v3.0 Release Candidate 1 is uploaded:
- Fixed major logical error
- Fixed two audio after video encoding
- Changed: 'Use Min br' instead of 'Fixed bitrate' (which used D2S settings).

Info: Beta 3 had about 200 downloads.
http://home.no.net/tylo

[DDogg]

Test results v3.0 RC1. I used AUTO and checked all the stuff. The second disk was 691,005 KB so something is not quite right?

--------------------------------------------------------
D2SRoBa v3.0 RC1
- Author: Tylo
- System: WIN_XP, AutoIt: 3.0.74.0
--------------------------------------------------------
- 2003-11-18 07:14:54
--------------------------------------------------------
Settings:
- Output type : SVCD
- CD size : 800
- Number of CD's : auto
- Sample percent : 1.0
- Auto Q. (worst) : 32
- Maximum Q. : 64
- Adjust Q : 0.6
- Sizing pass : on=1 (-2.0 .. 1.0)%
- Audio after vid : on=1 (% of video, 10%, min 128 kbps)
- Adjust max br : on=1 (total 2754 kbps)
- Crash recover : No
--------------------------------------------------------
- Waiting for CCE window ...
- Detected a CCE encode window
- Detected the CCE main movie encode window... trying to kill it.
- Finished shutdown of CCE and D2S
--------------------------------------------------------
- Movie length : 01:34:21 (135740 frames, 23.976 fps)
- D2S audio setting : (128 + 0) kbps
- D2S video estim. : 2000 kbps, 2 CDs
- Select Ranges : every 1200, select 12 frames
- Sample frames : 1368
--------------------------------------------------------
Calculations:
- 1 CD: video_br=928 (928) audio1_br=128 video_sz=656733400. fill=100.0% cbr=0
- 2 CD: video_br=1936 (1936) audio1_br=192 video_sz=1370081748. fill=100.0% cbr=0
- 3 CD: video_br=2530 (2976) audio1_br=224 video_sz=1790447739. fill=86.1% cbr=1
--------------------------------------------------------
Computing target number of CDs:
- Using max bitrate=2562 for auto sample (adjusted for 2 CDs)
- Estim. mpv BR : 1396 (Q=32, err=-1.#%, size=988166561., sample sz=9958832.)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 2
- Target mpv BR : 1936 (max 2562) kbps, size=1370081748. bytes
--------------------------------------------------------
Search for Q:
- Estim. mpv BR : 1858 (Q=16, err=-4.0%, size=1314926081., sample sz=13251944.)
- Estim. mpv BR : 1958 (Q=13, err=1.1%, size=1385811335., sample sz=13966332.)
- Estim. mpv BR : 1925 (Q=14, err=-0.6%, size=1362584712., sample sz=13732252.)
- Determined Q : 13 = round(14 - 0.6 + (-0.6)/1.7)
--------------------------------------------------------
- Encoding Movie
- 2003-11-18 07:27:02
- Actual mpv file size:
11/18/2003 08:25 AM 1413197436 Encoded_Video_CCE_NTSC.mpv
- OPV pass result: 103.1% on target ( 1413197436 / 1370081748.)
--------------------------------------------------------
- Executing VBR Sizing Pass - 1936 kbps.
- Actual mpv file size:
11/18/2003 09:25 AM 1368708108 Encoded_Video_CCE_NTSC.mpv
- Sizing pass result: 99.9% on target ( 1368708108 / 1370081748.)
--------------------------------------------------------
- Executing: C:\Program Files\DVD2SVCD\BeSweet\BeSweet.exe -core( -input "G:\hunted2\Extracted_audio_1.ac3" -output "G:\hunted2\Encoded_audio_1.mp2" -logfile "G:\hunted2\Encoded_audio_1.log" ) -azid( -L -3db -C 0db -S 0db -c normal -s surround -g max ) -shibatch( --rate 44100 ) -2lame( -e -b 192 -m s )
--------------------------------------------------------
- Recover:
- 2003-11-18 09:38:55

As an aside comment, I have lately reduced the adjust Q to .6 yet I still see oversize on nearly all encodes done so far. Next I think i'll reduce it to 0. Might be something to do with my filter setup.