D2SRoBa Plugin

[r6d2]

Originally posted by known_03
r6d2, I'm sure you will find a new cyberhome, i'm one looking forward in reading the idiot guide.
I told you I'd be back! :D
Hey, what happened to the dog picture?
I decided to use a photo of myself. My dog was barking very loud about me being using his personality.

[r6d2]

Originally posted by lordkinbote
I'd be hard pressed to find, on what I saw of your site, anything that even comes close to ONE of those things. Man, they're weird.
Believe it or not, I just received this mail from Lycos as a reply to my appeal:
Dear [name removed],

It seems as though your account was removed in error and has been
reinstated. I apologize for any problems this may have caused you.

Sincerely,

Tom
The sadest thing of all is that Yahoo! Geocities, my new cyberhome, turned out to be much easier to set up and mantain. And they provide interesting site statistics free of charge.

Too bad, Lycos. I think you lost a customer because of stupidity of procedures and lack of prompt and adequate response.

[lordkinbote]

Nice. Well, good for you, too bad for them.

[tylo]

Hi all, home sweet home..

Uploaded v2.5.2, - a maintainance release with new fixes.
http://home.no.net/tylo/

[DDogg]

That is great news. Thanks for doing this so quick. Now my request would be a very simple one. Would either you or r6d2 do a short piece on this automation of facar? There is no documentation or explanation except for like three words :)

Update: Tylo, I just did a test D2SRoBa opv 1 disk encode with 2000 set in the disk size (DVD2DVD). D2S was also set to 2000. The resulting MPV was 4.04 gigs.

D2SRoBa v2.5.2 - DVD2SVCD plugin, by tylo
OS Version: WIN_XP
--------------------------------------------------------
- DVD2SVCD Shutdown
- 2003-09-12 20:53:03
--------------------------------------------------------
Settings:
- Encoding mode : OPV RoBa
- CD size : 2000
- Number of CD's : 1
- Worst Q. : ---
- Maximum Q. : 64
- Sample percentage : 1.0
- Adjust Q. : 1.3
- Sample GOPs : 1
- Safety Zone : 34
--------------------------------------------------------
- Movie length : 02:51:52 (247259 frames, 23.976 fps)
- Audio size est : 577515182. bytes (448 + 0 kbps)
- CD mpv data size : 4000066093. (3103 kbps, 1 CDs)
- SelectRangeEvery : every 1200, select 12 frames
- Num. sample frames : 2484
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 1
- Target mpv size : 4000066093. (3103 kbps)

[tylo]

Now my request would be a very simple one. Would either you or r6d2 do a short piece on this automation of facar? There is no documentation or explanation except for like three words
I've sent r6d2 a InnoSetup script, that will install everything automatically. I want him to maintain this, so go to his site to look for updates.
/Edit: See the FACAR thread or http://www.geocities.com/r6d2_stuff

/Add: Usage: Select "FACAR DetectBorders" in Frameserver tab - do a preview in Convertion tab. Close preview window, the Frameserver tab will automatically be selected. Choose "FACAR", and detected top and height is already filled in. :D Two new directories in 'Avisynth2.5 Plugins' are added. If it doesn't work, Check D2SRoBa.ini - It should have 'autocrop_report=AVISYNTH_FACAR'.

@DDogg: I'll look into your reported error.

@r6d2: Add 'EnableDirDoesntExistWarning=yes' in the [Setup] section.

@all: update D2SRoBa v2.5.3 : just a small final close bug fix.

[r6d2]

Tylo,

Is it really needed to shut down CCE during search? Why not just change the Q, the file name and press "Encode"?

[tylo]

Is it really needed to shut down CCE during search? Why not just change the Q, the file name and press "Encode"?
It is easier and safer to use the EclCCE comand line arguments -ecl <file> -batch to run CCE 2.66+, than emulating GUI input (which is a little different than for CCE 2.50). Thus, CCE 2.50 doesn't close during Q-search.

[JoLander]

Hi,
I just found another issue with the current versions of D2SRoBa and DVD2SVCD.

I tried to convert a DVD -> DVD and started D2SRoBa.
Everything worked fine until the CCE process has started.
CCE continued to run and was not shutdown by D2SRoBa.

I did some tests and found the following:
I usually use DVDDecrypter and create an ISO image from a DVD for further processing.
This image I mount with Alcohol120% and use the image-drive in DVD2SVCD.
To speed up the process, I disable DVD-Ripping.
With this setting, D2SRoBa did not interrupt CCE.
When the DVD-Rip is switched on, D2SRoBa stops CCE and everything works correct.

Greetz,
Joe

[Trahald]

i was wondering if there was a better way to add the video file. ie.. when d2sroba takes over it right clicks cce and adds the file that way. ctrl-o works and wont grab the mouse arrow from me if im using it

also haveing some problems with batch mode.. d2s will hang when d2sroba tries to close it the first time.(this is even when im not using hte pc) but its intermittant. also a few times d2sroba set odd values into the bitrate part of the .ini file during batch (not when the d2s first runs but after d2sroba closes d2s then reopens it the first time).. d2s complains about it (it sits there waiting for input to load default bitrates or keep the bad ones) if i can duplicate the issues ill try to post the logs .

the bitrate part looked something like this

MinsHigh1=19
MinsHigh2=0
MinsHigh3=0
MinsHigh4=46
MinsHigh5=99999

iirc

last thing, tylo .. after reading the thread and your explanation of adjust q.. would i be correct in saying if i wanted the q to be a higher number (ie if it was a choice between 34 and 35 ... d2sroba will always pick 35) if i choose a low adjust q (preferably 0.0)? id rather the images be too small than to big.. im not that anal about filling RIGHT to the edge.. 20-30 megs on the short side is fine.. just dont want 20-30 over

ok.. last last thing.. can you make it so the preview window doesnt show at all? (like it does after the final encode terminates but remains in the task bar).. the novelty of watching it work is pretty much over.. i trust the q it comes up with. since the display cant be moved i could do without it.

anywho.. thanx for the great plugin, wook

[tylo]

@jolander: Please don't write current/latest ver. in bug reports. did you use 2.5.3? Anyone else had this with 2.5.3?

@w00kiee: I've now replaced the RightClick with menu opening for cce266+. The batch mode is hardly tested, so thanks for you report. I'll try some more myself. On the Adjust Q thing, use 0.0 to be (quite) sure not to get too big images. Future release will do an conditional extra pass if the size is too big, to make it fit.
Status window: I may do something, but now set status_x in the ini file to something high, and it will disapear.

[DDogg]

Future release will do an conditional extra pass if the size is too big, to make it fit. Good news!

tylo, did you find or get a chance to fix that bitrate problem in dvd mode I reported (not honoring disk size)? It is really killing me :)

[JoLander]

@Tylo,
sorry for not using the version numbers.
(they were 'current' at the date of my posting).

Actually I using D2SRoBa 2.5.3 and DVD2SVCD 1.2.1 B2.
With this versions I have the reported problem.

Greetz,
Joe

[tylo]

§§§§§§§§§§
§§§§§
§§§§ - Introducing. . The D2SRoBa v2.6.0.
§§§§
§§§§§ - The Ultimate Encoding Plugin.
§§§§§§
§§§§§§§ - Visit - http://home.no.net/tylo
§§§§§§§§ ------------------------------
§§§§§§§§§ - True Encoding Pleasure

:D

[telemike]

Can't wait to try it!!!!!!!!!!! :D

[homerjay]

was going to give my pc a night off but will have to punish it some more now :p thanks for your continued improvements :D

[DDogg]

Tylo, wonderful stuff! Very clean and understandable interface.

A few questions and a report:

1> If 0% is chosen as oversize, do you look at it as 0% or do you use that as a "always"? Same question on undersize.

I noted a small anomaly (DVD2DVD mode)marked in bold at the bottom:
========================================================
D2SRoBa v2.6.0 beta1 - DVD2SVCD plugin, by tylo
OS Version: WIN_XP
--------------------------------------------------------
- DVD2SVCD Shutdown
- 2003-09-19 11:46:04
--------------------------------------------------------
Settings:
- Encoding mode : OPV RoBa
- CD size : 2000
- Number of CD's : 1
- Worst Q. : ---
- Maximum Q. : 64
- Sample percentage : 1.0
- Adjust Q. : 1.3
- Sample GOPs : 1
- Sizing pass on : 1 (-2.0 .. 1.0)%
--------------------------------------------------------
- Movie length : 01:25:14 (122614 frames, 23.976 fps)
- Audio size calc : 286385720. bytes (448 + 0 kbps)
- DVD2SVCD mpv est : 1756030291. bytes (2747 kbps, 1 CDs)
- SelectRangeEvery : every 1200, select 12 frames
- Num. sample frames : 1236
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 1
- Target mpv size : 1756030291. (2747 kbps)
--------------------------------------------------------
- Binary search for Q:
- Estimated mpv size : 2085489941. (Q=32, 3262 kbps, 21022604 sample sz)
- Estimated mpv size : 1559701666. (Q=48, 2439 kbps, 15722440 sample sz)
- Estimated mpv size : 1777601029. (Q=40, 2780 kbps, 17918956 sample sz)
- Computed mpv size : 1668651347. (Q=44, 2610 kbps, 16820698.=(Q48+Q40)/2)
- Computed mpv size : 1723126188. (Q=42, 2695 kbps, 17369827.=(Q44+Q40)/2)
- Estimated mpv size : 1746220575. (Q=41, 2731 kbps, 17602628 sample sz)
--------------------------------------------------------
- Q Weight : 1.0 - (1756030291. / 99.2022653721683 - 17602628) / (17918956 - 17602628)
- Adjust Q : (41 + (0.687394043638874) - (1.3)) = 40.3873940436389
- Determined Q. : 40
--------------------------------------------------------
- Encoding Movie
- 2003-09-19 11:48:53
--------------------------------------------------------
- Actual mpv file size:
09/19/2003 12:28 PM 1727508452 Encoded_Video_CCE_NTSC.mpv
--------------------------------------------------------
- Estimation result: 0.% on target ( 1727508 / 1756030291.)
- Executing VBR Sizing Pass.
(rest of test still in progress)

[tylo]

Yepp, seems there is a problem related to that I have norwegian date format on my machine. In your case, it threrefore WILL do the Sizing pass (it thinks you're way off), but I think it will work OK otherwise. I will upload a fix for this tonight hopefully. For your other question: 0% Above/Below means: don't tolerate any oversize/undersize, so both set to 0.0 will guarantee always do the pass. I wonder If it would be better to set 98% and 101% as limits? Thanks for reporting so fast!

/Add: BTW: Now when we have the conditional sizing pass, I think we can normally increase the 'Adjust Q' to at least 1.6. Four out of five times (at least) you will get a closer fit, and one of five it will then oversize. (numbers are good guesses only). Found the bug - Should be uploaded within 20 mins.
/Add: 1.6.0 beta2 uploaded. :)

[r6d2]

Originally posted by tylo
§§§§ - Introducing. . The D2SRoBa v2.6.0.

Congratulations, Tylo, for a job well done.

PD: I'm not back, just hanging in from vacations ;)

[Holomatrix]

Sound like you guys have been busy :) New versions look great :)

[DDogg]

tylo, got a few problems on a svcd encode so here is the report. I ended up with two mpgs and images. bbMPEG_Muxed_File01.mpg was 68,023 k so it was way off by quite a margin even after the sizing pass which suggests there is a internal calulation problem. Encoded_audio_1.mp2 was 123,314 k. The Target mpv size value seems whacked? I thought you were generating the second pass using the d2s internal calculations? It did not seem so, but I was in and out. Pulldown.mpv after sizing pass was 730,788 k.

Btw, I ran this test twice and both times I got a large second image so I know this is not a fluke. I am pretty confused at the moment :)

Notice I actually used a 5% sample.

D2SRoBa v2.6.0 beta2 - DVD2SVCD plugin, by tylo
OS Version: WIN_XP
--------------------------------------------------------
- DVD2SVCD Shutdown
- 2003-09-20 15:20:17
--------------------------------------------------------
Settings:
- Encoding mode : OPV RoBa
- CD size : 800
- Number of CD's : 1
- Worst Q. : ---
- Maximum Q. : 64
- Sample percentage : 5.0
- Adjust Q. : 1.3
- Sample GOPs : 1
- Sizing pass on : 1 (-2.0 .. 1.0)%
--------------------------------------------------------
- Movie length : 01:45:13 (151371 frames, 23.976 fps)
- Audio size calc : 126268769. bytes (160 + 0 kbps)
- DVD2SVCD mpv est : 749720815. bytes (1900 kbps, 2 CDs)
- SelectRangeEvery : every 240, select 12 frames
- Num. sample frames : 7572
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 1
- Target mpv size : 749720815. (950 kbps)
--------------------------------------------------------
- Binary search for Q:
- Estimated mpv size : 749127403. (Q=32, 949 kbps, 37473444 sample sz)
- Estimated mpv size : 1124580419. (Q=16, 1424 kbps, 56254652 sample sz)
- Computed mpv size : 936853911. (Q=24, 1187 kbps, 46864048.=(Q32+Q16)/2)
- Computed mpv size : 842990657. (Q=28, 1068 kbps, 42168746.=(Q32+Q24)/2)
- Computed mpv size : 796059030. (Q=30, 1008 kbps, 39821095.=(Q32+Q28)/2)
- Computed mpv size : 772593226. (Q=31, 978 kbps, 38647270.=(Q32+Q30)/2)
--------------------------------------------------------
- Q Weight : 1.0 - (749720815. / 19.9908874801902 - 37473444) / (38647270. - 37473444)
- Adjust Q : (32 + (0.974711626842667) - (1.3)) = 31.6747116268427
- Determined Q. : 31
--------------------------------------------------------
- Encoding Movie
- 2003-09-20 15:29:46
--------------------------------------------------------
- Actual mpv file size:
09/20/2003 04:35 PM 729933828 Encoded_Video_CCE_NTSC.mpv
--------------------------------------------------------
- Estimation result: 97.3% on target ( 729933828 / 749720815.)
- Executing VBR Sizing Pass.

[Holomatrix]

Seemed alright for me. Right on target. One file 811 Meg mpg.


========================================================
D2SRoBa v2.6.0 beta2 - DVD2SVCD plugin, by tylo
OS Version: WIN_2000
--------------------------------------------------------
- DVD2SVCD Shutdown
- 2003-09-20 19:02:09
--------------------------------------------------------
Settings:
- Encoding mode : OPV RoBa
- CD size : 800
- Number of CD's : 1
- Worst Q. : ---
- Maximum Q. : 128
- Sample percentage : 3.0
- Adjust Q. : 0.8
- Sample GOPs : 1
- Sizing pass on : 1 (-2.0 .. 1.0)%
--------------------------------------------------------
- Movie length : 01:31:34 (131742 frames, 23.976 fps)
- Audio size calc : 87915916. bytes (128 + 0 kbps)
- DVD2SVCD mpv est : 721872091. bytes (1051 kbps, 1 CDs)
- SelectRangeEvery : every 500, select 15 frames
- Num. sample frames : 3960
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 1
- Target mpv size : 721872091. (1051 kbps)
--------------------------------------------------------
- Binary search for Q:
- Estimated mpv size : 476097358. (Q=64, 693 kbps, 14310892 sample sz)
- Estimated mpv size : 665901783. (Q=32, 969 kbps, 20016176 sample sz)
- Estimated mpv size : 930042497. (Q=16, 1354 kbps, 27955916 sample sz)
- Computed mpv size : 797972140. (Q=24, 1161 kbps, 23986046.=(Q32+Q16)/2)
- Estimated mpv size : 712087734. (Q=28, 1036 kbps, 21404468 sample sz)
- Computed mpv size : 755029937. (Q=26, 1099 kbps, 22695257.=(Q28+Q24)/2)
- Computed mpv size : 733558852. (Q=27, 1068 kbps, 22049863.=(Q28+Q26)/2)
--------------------------------------------------------
- Q Weight : 1.0 - (721872091. / 33.2681818181818 - 21404468) / (22049863. - 21404468)
- Adjust Q : (28 + (0.544301430352402) - (0.8)) = 27.7443014303524
- Determined Q. : 27
--------------------------------------------------------
- Encoding Movie
- 2003-09-20 19:33:32
--------------------------------------------------------
- Actual mpv file size:
20/09/2003 10:25p 678854768 Encoded_Video_CCE_NTSC.mpv
--------------------------------------------------------
- Estimation result: 94.% on target ( 678854768 / 721872091.)
- Executing VBR Sizing Pass.

[tylo]

I am pretty confused at the moment
I was too when I read this, but when I looked closer at both of your logs, it stroke me that I have made a blunder with the calulations since v2.5.0: In the log line:
DVD2SVCD mpv est : 749720815. bytes (1900 kbps, 2 CDs)
the numbers 1900 kbps and 2 CDs are picked up from DVD2SVCDs calculations.

The current math:
roba_video_br = d2s_video_br * roba_cds / d2s_cds
(1900 * 1 / 2) = 950 --> Wrong

The math should be:
roba_video_br = (d2s_video_br * roba_cds / d2s_cds) - (audio_br * (d2s_cds - roba_cds) / d2s_cds)
(1900 * 1 / 2) - (160 * (2 - 1) / 2) = 950 - 80 = 870

In holos case:
DVD2SVCD mpv est : 721872091. bytes (1051 kbps, 1 CDs)

Current math:
(1051 * 1 / 1) = 1051

Should be:
(1051 * 1 / 1) - (128 * (1 - 1) / 1) = 1051 - 0 = 1051,
which is the same, so therefore it works for him!

Beta3 will be uploaded tonight with these fixes.

Btw: both of you should increase the Adjust Q. Dogg, was closer on target (97.3%) because of the error, holo 94%: I will change the default value to 1.6.

[DDogg]

That makes sense and would explain why on some encodes I was getting perfect sizing and on some it was messed up. Just to double check, I am doing the same encode again, but this time I made sure the time-bands in the d2s bitrate tab were set correctly to properly reflect one CD instead of two for this particular length of source. I notice the bitrate now looks correct in your log. Will let you know later, but I expect it will be fine.

[homerjay]

pretty glad to see this has been picked up :D
did a conversion last night that had been having a prob with oversizing so set -3.2 to force undersized ( 1cd ) but forgot to take tick out for do conditional sizing pass
the result was that it created a second mpeg 36mb in size :rolleyes:

[DDogg]

Yep, right on the money now that I made sure the time-band was adjusted to 1 cd. CD_Image_File_CD1.bin = 823,717 kb

========================================================
D2SRoBa v2.6.0 beta2 - DVD2SVCD plugin, by tylo
OS Version: WIN_XP
--------------------------------------------------------
- DVD2SVCD Shutdown
- 2003-09-21 11:18:02
--------------------------------------------------------
Settings:
- Encoding mode : OPV RoBa
- CD size : 800
- Number of CD's : 1
- Worst Q. : ---
- Maximum Q. : 64
- Sample percentage : 1.0
- Adjust Q. : 0.0
- Sample GOPs : 1
- Sizing pass on : 1 (-2.0 .. 1.0)%
--------------------------------------------------------
- Movie length : 01:45:13 (151371 frames, 23.976 fps)
- Audio size calc : 101015016. bytes (128 + 0 kbps)
- DVD2SVCD mpv est : 707105106. bytes (896 kbps, 1 CDs)
- SelectRangeEvery : every 1500, select 15 frames
- Num. sample frames : 1515
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 1
- Target mpv size : 707105106. (896 kbps)
--------------------------------------------------------
- Binary search for Q:
- Estimated mpv size : 726440520. (Q=32, 920 kbps, 7270596 sample sz)
- Estimated mpv size : 562038625. (Q=48, 712 kbps, 5625176 sample sz)
- Computed mpv size : 644239573. (Q=40, 816 kbps, 6447886.=(Q48+Q32)/2)
- Computed mpv size : 685340046. (Q=36, 868 kbps, 6859241.=(Q40+Q32)/2)
- Estimated mpv size : 698883604. (Q=34, 885 kbps, 6994792 sample sz)
- Computed mpv size : 712662062. (Q=33, 903 kbps, 7132694.=(Q34+Q32)/2)
--------------------------------------------------------
- Q Weight : 1.0 - (707105106. / 99.9148514851485 - 6994792) / (7132694. - 6994792)
- Adjust Q : (34 + (0.403307522489496) - (0.0)) = 34.4033075224895
- Determined Q. : 34
--------------------------------------------------------
- Encoding Movie
- 2003-09-21 11:20:54
--------------------------------------------------------
- Actual mpv file size:
09/21/2003 12:27 PM 684765848 Encoded_Video_CCE_NTSC.mpv
--------------------------------------------------------
- Estimation result: 96.8% on target ( 684765848 / 707105106.)
- Executing VBR Sizing Pass.

[Holomatrix]

Were did I go wrong? It started doing the 'Executing VBR Sizing Pass'
again. Was it the Sample percentage 5.0 or Adjust Q. 1.6 to high ? if it's going to start this VBR Sizing Pass then I should have just selected a Milti-pass VBR 1 encode.


========================================================
D2SRoBa v2.6.0 beta2 - DVD2SVCD plugin, by tylo
OS Version: WIN_2000
--------------------------------------------------------
- DVD2SVCD Shutdown
- 2003-09-21 16:46:16
--------------------------------------------------------
Settings:
- Encoding mode : OPV RoBa
- CD size : 800
- Number of CD's : 1
- Worst Q. : ---
- Maximum Q. : 128
- Sample percentage : 5.0
- Adjust Q. : 1.6
- Sample GOPs : 1
- Sizing pass on : 1 (-2.0 .. 1.0)%
--------------------------------------------------------
- Movie length : 01:45:57 (152430 frames, 23.976 fps)
- Audio size calc : 101721722. bytes (128 + 0 kbps)
- DVD2SVCD mpv est : 706489146. bytes (889 kbps, 1 CDs)
- SelectRangeEvery : every 300, select 15 frames
- Num. sample frames : 7635
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 1
- Target mpv size : 706489146. (889 kbps)
--------------------------------------------------------
- Binary search for Q:
- Estimated mpv size : 556603845. (Q=64, 700 kbps, 27879488 sample sz)
- Estimated mpv size : 822093981. (Q=32, 1034 kbps, 41177508 sample sz)
- Estimated mpv size : 652516515. (Q=48, 821 kbps, 32683616 sample sz)
- Computed mpv size : 737305248. (Q=40, 927 kbps, 36930562.=(Q48+Q32)/2)
- Estimated mpv size : 686090565. (Q=44, 863 kbps, 34365292 sample sz)
- Estimated mpv size : 704069599. (Q=42, 885 kbps, 35265836 sample sz)
- Computed mpv size : 720687423. (Q=41, 906 kbps, 36098199.=(Q42+Q40)/2)
--------------------------------------------------------
- Q Weight : 1.0 - (706489146. / 19.9646365422397 - 35265836) / (36098199. - 35265836)
- Adjust Q : (42 + (0.854400444353716) - (1.6)) = 41.2544004443537
- Determined Q. : 41
--------------------------------------------------------
- Encoding Movie
- 2003-09-21 17:54:09
--------------------------------------------------------
- Actual mpv file size:
21/09/2003 08:55p 623552972 Encoded_Video_CCE_NTSC.mpv
--------------------------------------------------------
- Estimation result: 88.2% on target ( 623552972 / 706489146.)
- Executing VBR Sizing Pass.

[DDogg]

It started doing the 'Executing VBR Sizing Pass' again.
Holo, clarity bro :) Are you saying you did not want the sizing pass, or that it did the sizing pass twice meaning three passes? Also I am surprised it is 18% off target, especially with a 5% sample. Frankly I would nearly think that impossible so maybe you found another pest for tylo to squish

[Holomatrix]

Seems to have been better with the 3% sample, true, anyway it didn't do it three times, I just didn't wan't it to do a VBR bitrate encode instead of the OPV encode. If at the end it's going to give me a Multipass VBR encode anyway (most of the time) then I should have set DVD2SVCD up that way in the first place and not use the plugin.

[DDogg]

Just tweak you undersize parameter to 5%. Then the sizing pass will not run normally unless the encode is off by 5% which is rare. That's why I remarked on your 18% error. I expect that will show there is some problem that tylo will have fixed soon from what he said above. This is just the second beta with the conditional sizing pass and it is near perfect already. Hang in there. We are about to have a very special new tool. I am sure you will be very pleased.

BTW, there is a heap of difference between OPV with sizing (RoBa) versus multipass. It is a completely different animal.

[r6d2]

Originally posted by Holomatrix
Seems to have been better with the 3% sample, true, anyway it didn't do it three times, I just didn't wan't it to do a VBR bitrate encode instead of the OPV encode. If at the end it's going to give me a Multipass VBR encode anyway (most of the time) then I should have set DVD2SVCD up that way in the first place and not use the plugin.
@Holo, for some reason this RoBa stuff does not fit easily in everybody's minds, but:

1. Did you uncheck "Conditional Sizing Pass"? If you don't, and the plugin misses the Q (as in this case), it will do it.

2. Multipass (VAF+1), is not likely to give you a better encode than OPV + 1VBR. So the plugin gets a better job done in more or less the same time.

3. Sample size does not matter! Choose 1% and you'll be good enough. There is not sample size which will give you a "closer Q". It is just a gamble, as it is the adjust Q.

No one has found a recipe that will work for every movie, so in the meantime, you're free to gamble, or just let the plugin do a conditional second pass.

Cheers,

[DDogg]

Sample size does not matter! Choose 1% and you'll be good enough. There is not sample size which will give you a "closer Q". It is just a gamble, as it is the adjust Q. Now that is very interesting! Coming from you I take it as the gospel, but if you get the itch to type I wouldn't mind a bit more education on the topic. Is this more math stuff about Babylonians that is gonna cross my eyes again? :D

[tylo]

Now, things are getting a little out of hands again. :rolleyes:

First, r6d2, shouldn't you be on vacation :). Sample size % does matter! On average, you will be closer on target when you increase sample %, but on certain movies, you will get a worse result. It is statistical probablity that comes into play here. However, I agree that you actually may use 1%-2% as max because the statistical probability varies so much depending on the movie at hand.

Still, I haven't got it as bad as 11.8% off, like holo did, at least not on a 5% sample (DDogg, you really are terrible at math :p). There shouldn't be any differences from earlier versions on this part, so you may rerun it on any older version to verify that you get this bad estimation result.

Another interesting thing is that DDogg got a bitrate of 896 from DVD2SCVD's one CD computation. In beta3, I compute it to be 870, which is then too low (see post above). This is probably due to the fact that there is less overhead data on a one CD encode than on a two CD encode in total. It's not straight forward to take the DVD2SVCD bitrate calculation for e.g. two CDs, and translate that to what the bitrate will be for one CD (sorry, r6d2). Fortunately, by trial and error, it think it is possible to figure out a formula (which also takes subtitles into account), that will make this translation more accurate. I'll do some investigations. :)

/Add 1: As a resort, we can let DVD2SVCD recalculate bitrate with the number of CDs that we have decided to use, if it is different from what DVD2SVCD calculated initially. The number of CDs would in that case be chosen on slightly wrong premises... So we must recover from 'Video Encoding', and breaking it a second time. (r6d2, I think you suggested this once). Then we can search Q, and continue as before. It wouldn't work well with the '?' mode as this would lead to even more sample tests, but otherwise this may be the way to go...

/Add 2: I think it's better to have a 'Adjust %' settings instead of the 'Adjust Q'. I.e. the Q-search should compare sample_size*(frames/sample_frames)*(100+adjust_percent)/100 with the target size. This will make it easier for users to quantify how much to compensate for the typical estimation error you get, and is also a better way. Much less complicated than the Q adjust implementation too. :)

[r6d2]

Originally posted by tylo
First, r6d2, shouldn't you be on vacation :). Sample size % does matter! On average, you will be closer on target when you increase sample %, but on certain movies, you will get a worse result.
You are seeing the same kind of data as I am and are getting into different conclusions ;). My statement comes from the following analysis:

If sample size does matter, then the bigger the sample the more accurate the target BR gotten (it's BR we're ultimately estimating, not Q). This is not but a common sense statement.

Well, try this: encode n samples with 2^(n-1)% sample size each. For non babylonians ;), this is 1%, 2%, 4%, 8%, etc. Tabulate the results obtained (Sample size, BR) in Excel. Add a [100%, Actual_BR] pair to your list and draw a curve in X-Y graphic mode.

Common sense says that since you're doubling the sample on each try, the error should halve accordingly.

You'd be surprised. It does not in the general case. Some movies get it even more accurate with 1%, even with 0.5%!

The conclusion is common sense did not work. So anything about sample size you may find adequate for one movie will not be for the next one.
It's not straight forward to take the DVD2SVCD bitrate calculation for e.g. two CDs, and translate that to what the bitrate will be for one CD (sorry, r6d2).
Actually, I did manage to find a way to estimate the BR for any number of CDs within 0.5% precision of FitCD's calculations (which is even more precise tan DVD2SVCD's). See my RoBa Q Finder for reference. It does not support subtitles yet, though, but it is a starting point for your further investigations.

/Add 1: As a resort, we can let DVD2SVCD recalculate bitrate with the number of CDs that we have decided to use, if it is different from what DVD2SVCD calculated initially. [...] So we must recover from 'Video Encoding', and breaking it a second time. (r6d2, I think you suggested this once).

Yes, but if you take my suggestion of determining first the number of CDs, then the video BR and then the Audio BR, you'll save all the hassle. D2S will have the correct number of CDs when it runs Video Encoding.
Then we can search Q, and continue as before. It wouldn't work well with the '?' mode as this would lead to even more sample tests, but otherwise this may be the way to go...

If you use Newton, sampling different resolutions, resizers (and even denoisers) would be very straightforward with the new FACAR. Just change a parameter. And since you can use 1% samples and it converges faster, the user won't notice the additional samples.

I'm still on vacation, but this stuff is too hot! :D

[Holomatrix]

Originally posted by r6d2
@Holo, for some reason this RoBa stuff does not fit easily in everybody's minds, but:
I think I've missed a few pieces of the puzzle :)


1. Did you uncheck "Conditional Sizing Pass"? If you don't, and the plugin misses the Q (as in this case), it will do it.
[/B]

Should I always have it checked then? If I want to be most acurate?

[DDogg]

Gee, it must be true...Whiskey does impair math ability!

Tylo, this thing is working so darn good now I hope it doesn't get broken trying to fix it. I did 5 encodes back to back using 1% sample size and every one was within 2 to 4%. An oddity was a reverse of yesterday as I inadvertently had d2s set at 1 cd for the source length and I set up the plug-in for 2 cds. Results with a 1% sample were 100.2%. I don't know if that has any relevance or was just a statistical anomaly.

Granted I don't use subs or multiple audio tracks but this thing is working really good and it is so much more elegant and streamlined in operation with fewer breaks than earlier versions. Er, I guess I'm just suggesting caution :)Should I always have it checked then? If I want to be most acurate? Holo, all the conditional says is IF the OPV is off by more than the X% you have selected THEN do the sizing pass. You make the election how much sizing error you are willing to accept as a trade-off for saving the time of not having to do the second pass.

It is more or less accurate to say the effect on quality by the second pass is neglible whether the size correction is up or down. It is only for sizing, so if you are willing to accept a 5% error of 30-35 megs too small then set it there. I personally think 3% is more realistic for those that speed is the most important. If you use a 1% sample the time saved could well contribute to the second pass if needed so you may save time in the long run if you tend to redo encodes because you are unhappy with the final size. With the sizing pass on you will always get it just right the first time.

[r6d2]

Originally posted by DDogg
It is more or less accurate to say the effect on quality by the second pass is neglible whether the size correction is up or down. It is only for sizing
I thought we had already settled the matter but seems that whiskey is bad for memory too ;)

If the MPV is oversized, the second pass is needed to reduce or it won't fit.

If the MPV is undersized by x%, the second pass will improve the BR by x%.

Example: If you are doing a 1 CD rip on a 90 min movie with audio=128 (hence video avg=1088), that "negligible" 5% error will give you just a 1034 output video, that's 54 kbps less than expected, which can really make a difference.

[DDogg]

I wish you had left that can sealed. Note the word negligible. I thought we had agreed that given the accuracy of the opv which seems to get it within 3% most of the time that the effect on quality was more theoretical than practical reality so I felt the statement would help alleviate confusion. Instead I guess it will now cause more. In order not to cause any confusion let me restate it:

IMHO, and as a general rule of thumb the effect on quality by the second pass is negligible whether the size correction is up or down unless the sizing correction is substantial like 5% or so.

Technically speaking, if the OPV is smaller than the targeted size then the size correction will increase the bitrate so that should increase quality. If the OPV is too big (rarely it seems) then the resizing pass will slightly reduce bitrate thus decreasing quality.

Practically speaking, I think the quality difference for a 2 or 3% upsizing correction is not going to be noticeable to anybody and if the OPV pass is 1% too big then most probably a extra mpg/image file will be created with a few seconds of Credits and can be discarded. If you have used frame-selection to already cut credits this may not work out so be forewarned.

Not disregarding the above, as a practical matter, if you want the best efficient and quality it would be best to leave the conditional sizing pass on for all encodes. Sometimes that will cost you an extra pass, but given that it is fully automatic and hopefully at off times or night who cares?

[Trahald]

because the estimated Q's are integers and even a 1% sampling is going to be fairly accurate, raising the % a few notches CAN end up with a less than accurate Q (by dumb luck of which movie you pick). although obviously if you really think it out.. basically you would have to say 1% sampling can be better (more accurate) than 100%... which uhhmmm.. would be impossible. not nearly impossible. but impossible. its just that 100% is the only % that wouldnt be effected by the fact the Q input has to be an integer. the derived q would be the same as the ultimate q (unless your pc is busted :D )(obviously were talking about the final encode being OPV and not multipass-VBR)

if you do the excel thing and take % starting from 1% sampling to 100 % and draw an average line through it you would see the trend from less accuracy to more accuracy (would work better if you factor in multiple movies)

[DDogg]

As an FYI adaptive filters like Peachsmoother will negatively effect the accuracy of sampling prediction because the rapid fire scene changes of a sample can cause adaptive filters to return different results in the sample versus the full encoding pass. That pretty well wrecks the sampling accuracy. Peachsmoother is the only one I have verified this with, but I assume there may be a few others out there so be forewarned.

[r6d2]

Originally posted by w00kiee
its just that 100% is the only % that wouldnt be effected by the fact the Q input has to be an integer.
Nope, even for 100% "samples", i.e, the whole movie, the integer nature of Q will not give you a filled CD in the general case. Increasing Q by 1 unit may give you anywhere between 20-50 less MB, depending on the source.
if you do the excel thing and take % starting from 1% sampling to 100 % and draw an average line through it you would see the trend from less accuracy to more accuracy
Nope, that is an unexisting trend and it won't show up. Try it yourself.
(would work better if you factor in multiple movies)
Sorry, this assumes movies behave alike, which in my experience, they don't.

[r6d2]

Originally posted by DDogg
IMHO, and as a general rule of thumb the effect on quality by the second pass is negligible whether the size correction is up or down unless the sizing correction is substantial like 5% or so.

Actually, believe it or not, the aforementioned error being "negligible" depends on the number of CDs. 5% more BR is important in 1 CD, but not in 2 or 3.

I know this sounds like black magic again, but let me explain. On the same example above, with 2 CDs, audio=192 (10% rule), video avg=2240. Error=5% means output BR of 2128, which is as good as 2240.

So, quality wise, the error x% is important as the target BR is far from the max by spec. Or something babylonian ;) as:

Let y% the error you are willing to accept on Avg=Max. (for instance, 10%)

Target_BR * y%
Then x% = --------------
Max

Sorry guys, the can was open! :D:D:D

[DDogg]

I know this sounds like black magic again Not at all. Perhaps Babylonian anality, but definitely not black magic (cracking myself up):)

[Trahald]

hehe.. i guess i'll call this agreeing to disagree (closes the can)
if im bored maybe i'll do the chart (read -- never) :D

anywho.. so far with 2.6.0 beta3.. been flawless

good work tylo

[tylo]

Thanks, folks.

For the moment, adjust the time bands so that D2S selects the number of CDs you want, like DDogg did, and you'll get an exact target bitrate.

When you get back, r6d2, if you try with e.g the logs posted by DDogg, wherein the first, DVD2SVCD computed the bitrate for 2 CDs, and in the second for 1. - DVD2SVCD mpv est : 707105106. bytes (896 kbps, 1 CDs)
Using your sheet, I don't get close numbers. What may be wrong is the bbmpeg mux overhead, which I think is different depending on certain options given to it..? Also, for the 'Reduce size' DVD2SVCD uses 5 (in .d2s file). I will try a bit more to translate a given bitrate for X CDs to Y CDs. That may still be simpler than a full-blown computation of the bitrate.

Btw: In the next beta, I've added a checkbox which allows you to do the Audio encoding after the Video. It doesn't do anything more fancy than that now, but in a future version it can be modified to e.g. let audio bitrate be X% of the video bitrate (maybe within min/max) :)

[r6d2]

Originally posted by tylo
When you get back, r6d2, if you try with e.g the logs posted by DDogg, wherein the first, DVD2SVCD computed the bitrate for 2 CDs, and in the second for 1. - DVD2SVCD mpv est : 707105106. bytes (896 kbps, 1 CDs)
Using your sheet, I don't get close numbers. What may be wrong is the bbmpeg mux overhead, which I think is different depending on certain options given to it..? Also, for the 'Reduce size' DVD2SVCD uses 5 (in .d2s file). I will try a bit more to translate a given bitrate for X CDs to Y CDs. That may still be simpler than a full-blown computation of the bitrate.

Well, I don't have the oher parametes used by DDogg so I cannot check his D2S's calculations, but I still get 1kbps error comparing to FitCD in my movies (using RoBa Finder 1.2, dated 30/8).

The sheet is made using my usual muxing parameters: uncheck scan offsets and uncheck seq.header alignments. If I change that in FitCD with the DDogg log's data, I get 893 where D2S got 896. Still pretty close.

I'll try to include these parameters in the sheet.

[tylo]

Great! That's only 0.33% difference (or about 2.5MB for a typical movie). So, I gather you know (from FitCD) how much those two options impacts the muxing-size, in order to get it in the sheet?

I'll wait for your new sheet then, and see if I can reproduce it myself. Maybe you even can mail me the filled in sheet that shows it.
Thanks.

[r6d2]

Originally posted by tylo
Btw: In the next beta, I've added a checkbox which allows you to do the Audio encoding after the Video. It doesn't do anything more fancy than that now, but in a future version it can be modified to e.g. let audio bitrate be X% of the video bitrate (maybe within min/max) :)
@Tylo, what about this:

1. Let D2S start and do the DVD2AVI thing, frame selection. Abort.

2. Do the first sample encode. Determine number of CDs. Choose audio by the 10% rule and let the user specify a "manual minimum override". Let the user confirm the audio size. Set all NumCD variables to the number of CD found.

3. Resume D2S from audio extraction. Let D2S do audio encoding and picture conversion.

3. Let D2S start Video Encoding in OPV mode (if you check "Calc. bitrate as max", D2S will give CCE the AVG you need for both Q finding and for the eventual second pass.) Abort as usual.

4. Find Q.

5. Do the OPV pass and VBR pass if needed.

6. Resume D2S from pulldown/muxing as today.

7. You're done.

The flaw is at stage 2 you don't have the target BR. But since this is just the number of CDs estimation, a simple and inexact calculation like this will do:

(800*n)/Duration = Video_BR + Audio_BR

[DDogg]

Sounds feasible but it needs to support DVD2DVD also which may may have 3-6 subs, 2 audio avs or mp2 plus the scenarist muxing overhead. tylo, can I do any certain tests for you to help get you data? If so just instruct your humble servant :) I still urge caution :) It is just so good now. Perhaps different methods for dvd2svcd and dvd2dvd as dvd2dvd is normally only going to be one disk and people will not be using auto mode I would think.

[r6d2]

Originally posted by DDogg
Sounds feasible but it needs to support DVD2DVD also which may may have 3-6 subs, 2 audio avs or mp2 plus the scenarist muxing overhead.
Actually, the approach described above can be easily generalized for multiples audio and subs, and target media also.

While my RoBa sheet gives a much better BR estimation for the VBR pass than D2S, it just mimics FitCD, which does not support DVD targets well (that's Fit2Disc).

(Remember the discusion of DVD2SVCD and shh about the former using the latter's FitCD code released under GPL without intending to release the sources himself? ;))