Modification and optimization of robshot method and its possible use with DVD2SVCD

[r6d2]

Originally posted by Holomatrix
@r6d2 - "garbage in, garbage out" Yes, that could be what happened. Even though the DVD looks good, I've seen some 3.5 GIG VOB DVD's look better than some 6.5 GIG VOB DVD's.
Sorry, Holomatrix, I did not mean the DVD was a bad source. I meant maybe you gave not D2Sroba the right parameters... Try as Telemike says, Mode=auto and Q=30.

Q=25 and BR=940 may have another meaning... CCE may be thinking that to you BR is more important than Q.

[r6d2]

Originally posted by ric1234
Small problem however : my last CDs are never filled up, generally I have 750 Mb out of 805 Mb. [...] So I was wondering if other people have the same results with D2SRoBa ? Anyone knows how to get closer to fill up last CD ?

This one should make it into the FAQ :D.

Ok, let's see. CCE OPV uses a desired Q.factor as input. It does not allow you to specify AVG.BR. So, Tylo uses a binary search to find the Q(sample) for a sample of the movie which will then use to do the whole encode.

This Q(sample), even after binary search, is a guess. It depends on the sample. There is no guaranty it will represent the whole movie. If it does or not depends a lot on the source material. Extensive testing has been made to find a way of getting the best sample possible for any movie, with different GOP sizes, sample sizes, and all these tests have been inconclusive.

The best precision I've got is 1%, but with a sample size that renders OPV useless. I've not given up yet, but this is the state of the art.

But the good side on this is that with the default params, D2Sroba will almost always find a Q(sample) that is higher than the Q.factor (this is because using GOP=1 you have sorta worst case scenario).

So, encoding with that Q, if that Q satisfies you (you may look at the binary search samples on the Video folder), you will get a constant quality throughout all the movie, which will likely not fill you last CD. That is the price you pay for doing just one pass.

Since you're getting the quality you are satisfied with, this should be no problem. Yes, it's not the best quality you could get out of those CDs, but... As Dirty Harry said: You have to ask yourself a question: Do I feel OK paying this price? :)

Regards,

[ric1234]

@r6d2, well, your explanation is pretty clear, I understand it much better now, thanks.

Originally posted by r6d2
But the good side on this is that with the default params, D2Sroba will almost always find a Q(sample) that is higher than the Q.factor

Ok, this is safe. But now let's say I want to take more risks and reduce the Q so that *maybe* I will overflow my last CD... (for the reasons explained above). What would you think of my idea about the "cutting" ?

since there is a risk of overflow, would it be possible, after the final .mpv is created, to check its size BEFORE the "muxing and cutting" part, and if it appears that it is going to overflow last CD, then automatically add 1 CD image for the "cutting" ? This would be great because most of the time, you would just loose some end credits and I'm ok to discard these.
So for instance, you would select 2 CD but end up with 3 images : 805 + 805 + 30 Mb and just discard the last one.

But this raises one question : which software should do the size check ? D2SRoBa or DVD2SVCD ? Because I wonder if D2SRoBa has not given back the hand to Dvd2svcd at this stage and therefore can not interact with the process anymore... Unless D2SRoBa keeps monitoring the process until the encoding is finished and can squiz a little tweak in the number of CDs before the muxing starts... Or unless @dvd2svcd (who has done so much already) would consider adding this feature in a next version... ?

Thanks and Regards,
Ric

[r6d2]

Originally posted by ric1234
What would you think of my idea about the "cutting" ?

Well, I think it's a great idea. So great, that DDogg claimed he did the very same thing on this very thread, june 12: :D
Quoting DDogg:
The one pass vbr section of the plugin really works well although I am finding it is difficult to get a completely full second cd. I have been increasing the Q adjust upward to actually force a slight oversize on a two cd encode to cause a third partial cd to be created in the muxing step. That small third cue/bin is normally just a few minutes of titles which I can discard thus allowing the second cd to be full.

Seems that he got the muxed files cut automatically... Did you try it? Doesn't DVD2SVCD cut properly in your case?


Unless D2SRoBa keeps monitoring the process until the encoding is finished and can squiz a little tweak in the number of CDs before the muxing starts... Or unless @dvd2svcd (who has done so much already) would consider adding this feature in a next version... ?

I've asked Tylo to have D2Sroba do some more stuff for DVD2SVCD, for instance, resize the CDs to the expected average size after the encode in order to get equally sized CDs (this is good so you don't end up with a last image of just 15 min).

I don't know if he has had time to do that, since he's on vacation, but yes, D2Sroba may continue to get its hands into DVD2SVCD.

DVD2SVCD has not been updated for a while, I don't know if @DVD2SVCD is still improving it.

Regards,

[ric1234]

Ooops ! Sorry DDogg, that is you idea in fact ! I couldn't remember this post... ;)

Seems that he got the muxed files cut automatically... Did you try it? Doesn't DVD2SVCD cut properly in your case?Well no, in my case, it doesn't cut properly all by itself.
I tried it twice and each time I ended up with an 800+++ Mb last CD image...

Anyway, I thought a bit further about it and I manage to re-cut the movie after the whole SVCD creation is finished. Here is what I did (that was a 3 CD conversion) :
- First I ran the whole D2SRoBa process with Q adjust = 8. I ended up with 3 images : 805 - 805 - 869 Mb
- Edit "dvd2svcd project file.d2s"
- Found the following lines (don't know which ones are really used) :
[Settings]
Number of images=3
Volumes=3
[MovieInfo]
Images=3
- Replaced 3 by 4 and saved
- Start Dvd2svcd and Recover from "muxing and cutting"
And it worked perfectly, I then ended up with 3 images of 805 Mb plus 1 image of 64 Mb. Luckily, that last sequence was from 10 sec after the start of the end credits ! :D

But I must say Q Adjust=8 is probably too much, I think I should better use Q=6 because that one was *very* close from the loss of the end of the movie... (the sample was 4%, so I assume quite accurate)

Anyway, I really hope the "Adaptive cutting" feature will be added in the future ! That would be great :rolleyes:
Meanwhile, if there's an easier way to do it manually, I'd be happy to take it.

Cheers,
Ric

[r6d2]

Originally posted by ric1234
But I must say Q Adjust=8 is probably too much, I think I should better use Q=6 because that one was *very* close from the loss of the end of the movie... (the sample was 4%, so I assume quite accurate)

You really mean 8, not 0.8? Gee! That's pretty unusual. What Tylo does is simply substract this number to the calculated Q (take a look at the log). Normally the Q you need should be around the Q automatically calculated, untweaked. 8 is indeed too big. The only explanation I can find is that the Q was very, very low, and the movie was either maxed out or might have fitted on 2 CDs with a reasonable Q.

[ric1234]

I know, it seems strange to me too to have to Adjust Q so much. This is why I tried on several movies, but I had pretty much always the same results... But again, I think -8 is too much, and I'd better go for -6 (but still it is far from -1.6 proposed in D2SRoBA)

Anyway, here is the log :
Settings:
- CD size : 800
- Number of CD's : 3
- Worst Q. factor : ---
+ One pass VBR Robshot/Bach : 1
- Sample percentage : 4
- Sample GOPs : 1
- Adjust Q. : 8
- Safety Zone : 34
+ Do credits bitrate tweak : 0
- Start time : 0:00:00
- Bitrate : 1000
--------------------------------------------------------
- Movie length : 02:39:42 (239572 frames, 25 fps)
- Audio size est : 383315200. bytes (192 + 128 kbps)
- CD user data size : 833237920. = (800 - 5)*1024*1024 - 380000
- SelectRangeEvery : every 375, select 15 frames
- Num. sample frames : 9585
--------------------------------------------------------
Calculations:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Number of CDs : 3
- Target mpv size : 2059004058. (1718 kbps)
--------------------------------------------------------
- Binary search for Q:
- Estimated mpv size : 2329687909. (Q=32, 1944 kbps, 93208132 sample sz)
- Estimated mpv size : 2057591804. (Q=48, 1717 kbps, 82321880 sample sz)
- Computed mpv size : 2193639856. (Q=40, 1831 kbps, (Q48+Q32)/2)
- Computed mpv size : 2125615830. (Q=44, 1774 kbps, (Q48+Q40)/2)
- Computed mpv size : 2091603829. (Q=46, 1746 kbps, (Q48+Q44)/2)
- Estimated mpv size : 2073434799. (Q=47, 1730 kbps, 82955740 sample sz)
--------------------------------------------------------
- Q Weight : (2059004058. / 24.9944705268649 - 82321880) / (83682662. - 82321880)
- Adjust Q : (48 + 1.0 - 4.15222145488089e-002 - (8)) = 40.9584777854512
- Determined Q. : 40 As you can see, I am in very reasonable Q values and bitrates. With this settings, OPV's running with a Q factor = 40.
A previous test with Q Adjust=1.6 was running OPV at Q factor 47, but I had 50Mb left on last disk. Now at Q=40, the improvement in quality is quite sensible.
Ric

[homerjay]

hello ric1234

just a question for you - what version of cce are you using
i was on 2.50 for above tests but i updated to 2.67 last night and with basic settings it actually estimated the correct size....

could be coincidence - i dont know

[ric1234]

Hello Homerjay,

I am using CCE 2.66... But you're right, this could be a track to explore...
I still have a 2.50, I'll give a try tonight to see how much difference there is with 2.66.

Ric

[r6d2]

Originally posted by ric1234
[B]I know, it seems strange to me too to have to Adjust Q so much. This is why I tried on several movies, but I had pretty much always the same results... But again, I think -8 is too much, and I'd better go for -6 (but still it is far from -1.6 proposed in D2SRoBA)
[...]Now at Q=40, the improvement in quality is quite sensible.

I see your values are "reasonable". However, I have no idea why this happens to you, and since it is on several movies, gee, I'm lost.

I personally had some strange stuff happening once, D2Sroba said Q=48 was good and went for a 1 CD rip. The movie looked not that bad, actually, but something smelled bad to me anyway. Its was an ~80 min movie, 4:3 format, BR was 1100 or so. I found it more suitable for MPEG1 and TMPGEnc.

You know, D2Sroba is magic, in a sense (as Tylo says), but it's not black magic. As a matter of caution, don't just trust anything a software tool will say (Try MS Word's grammar corrector to see what I mean :)). DVD encoding is an artisan work. Use your instinct too, and all you already know.

Before I made my choice, I tried different worst Q values, with several movies, and finally I kept 30 as my standard. No trouble ever since.

(It also happened to be the value recommended by Bach, the author of the Roba method).

BTW, Tylo uses 38.

[ric1234]

I am among these people who are not absolute Q fanatics (uh don't beat me :) ), I mean I always put in the balance the number of CDs as well.

Sometimes if the Q factor increases a little bit but it saves me 1 CD, then I'll be ok with it. Anyway, I respect both approachs.

In fact I wonder if my way of thinking would not be something like (and this could another idea proposed to Tylo ;) ) :
Estimate for 2 and 3 CDs (for instance, could 3-4) and I want Q between 10 and 40.
And the decision would be :
If one of estimated Q is in between => ok, take it ; if not, which Q-estimate is the closest to one end of the interval ? => then take that one.
So that if 2CD => Q=45 and 3CD => Q=1, then I'd go for Q=45
The only difference with the "Num CDs=auto" current method (I think) is that if it was 2CD => Q=45 and 3CD Q=11, then I'd keep Q=11 (that means I don't want to go too low in Q and waste space on the last CD... Is there a way to do this ?)

Regarding the accuracy of Q-estimation, I'll continue my tests and post feedbacks (if other people would like to post feedbacks, shall we start a new thread for this ?)
In this context anyhow, this is why I was asking for an "auto overflow cut" feature ( please ? :rolleyes: )

See you, Ric

[r6d2]

Originally posted by ric1234
[B]I am among these people who are not absolute Q fanatics (uh don't beat me :) ), I mean I always put in the balance the number of CDs as well.

I guess "?" mode is for you then, with sort of a threshold for the desired Q. Good idea. It would also reduce the cases where "you have to be there" to just select your option. Good for batch mode.

Regarding the accuracy of Q-estimation, I'll continue my tests and post feedbacks (if other people would like to post feedbacks, shall we start a new thread for this ?)

I agree with that. In fact, I think D2Sroba deserves its own thread already. Hope Tylo will initiate one.

[ric1234]

Originally posted by homerjay
i was on 2.50 for above tests but i updated to 2.67 last night and with basic settings it actually estimated the correct size....
Hello homerjay, here is the result of my tests with both CCE 2.50 and CCE 2.66.
As you will see, the prediction is more accurate with 2.66 than 2.50. But since even 2.66 is 50 Mb from filling up last CD, maybe I should try with 2.67...?

Image sizes with 2.50 : 805 + 805 + 735 Mb
Image sizes with 2.66 : 805 + 805 + 758 Mb

The source and settings were exactly the same in both encodes :
--------------------------------------------------------
Settings:
- CD size : 800
- Number of CD's : 3
- Worst Q. factor : ---
+ One pass VBR Robshot/Bach : 1
- Sample percentage : 4
- Sample GOPs : 1
- Adjust Q. : 1.6
- Safety Zone : 34
+ Do credits bitrate tweak : 0
--------------------------------------------------------
- Movie length : 02:39:42 (239572 frames, 25 fps)
- Audio size est : 383315200. bytes (192 + 128 kbps)
- CD user data size : 833237920. = (800 - 5)*1024*1024 - 380000
- SelectRangeEvery : every 375, select 15 frames
- Num. sample frames : 9585
--------------------------------------------------------
- Number of CDs : 3
- Target mpv size : 2059004058. (1718 kbps)
--------------------------------------------------------

Calculations for CCE 2.50 :
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Binary search for Q:
- Estimated mpv size : 2477298953. (Q=32, 2068 kbps, 99113880 sample sz)
- Estimated mpv size : 2189631693. (Q=48, 1827 kbps, 87604644 sample sz)
- Estimated mpv size : 2064171948. (Q=56, 1723 kbps, 82585144 sample sz)
- Estimated mpv size : 2006266958. (Q=60, 1674 kbps, 80268432 sample sz)
- Computed mpv size : 2035219453. (Q=58, 1699 kbps, (Q60+Q56)/2)
- Estimated mpv size : 2049873611. (Q=57, 1711 kbps, 82013084 sample sz)
--------------------------------------------------------
- Q Weight : (2059004058. / 24.9944705268649 - 81426788.) / (82585144 - 81426788.)
- Adjust Q : (57 + 1.0 - 0.821504513429513 - (1.6)) = 55.5784954865705
- Determined Q. : 55
--------------------------------------------------------
Replace values in ecl file:
- opv_q_factor : 55
- opv_brate_max : 1716 ==> 2350
- vbr_brate_avg : 6000 ==> 1718
--------------------------------------------------------
- Encoding Movie
- Actual mpv file size:
30/07/2003 04:49 1986452308 Encoded_Video_CCE_PAL.mpv
--------------------------------------------------------

Calculations for CCE 2.66 :
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Binary search for Q:
- Estimated mpv size : 2329687909. (Q=32, 1944 kbps, 93208132 sample sz)
- Estimated mpv size : 2057591804. (Q=48, 1717 kbps, 82321880 sample sz)
- Computed mpv size : 2193639856. (Q=40, 1831 kbps, (Q48+Q32)/2)
- Computed mpv size : 2125615830. (Q=44, 1774 kbps, (Q48+Q40)/2)
- Computed mpv size : 2091603829. (Q=46, 1746 kbps, (Q48+Q44)/2)
- Estimated mpv size : 2073434799. (Q=47, 1730 kbps, 82955740 sample sz)
--------------------------------------------------------
- Q Weight : (2059004058. / 24.9944705268649 - 82321880) / (83682662. - 82321880)
- Adjust Q : (48 + 1.0 - 4.15222145488089e-002 - (1.6)) = 47.3584777854512
- Determined Q. : 47
--------------------------------------------------------
Replace values in ecl file:
- opv_q_factor : 47
- opv_brate_max : 1716 ==> 2350
- vbr_brate_avg : 6000 ==> 1718
--------------------------------------------------------
- Encoding Movie
- Actual mpv file size:
28/07/2003 02:22 2010735312 Encoded_Video_CCE_PAL.mpv
--------------------------------------------------------

Ric

[telemike]

Hi-

I am using CCE 2.50 with the following Avisynth filters: Spacedust and TemporalCleaner.

I have been getting good results, however in my backup of my son's Star Trek Nemesis (2 cd encode) I notice at times that Picards face would shift. It seemed CCE kept part of the frame from moving in sync with other frames.....kinda freaky

[r6d2]

Originally posted by telemike
[B]Hi-

I am using CCE 2.50 with the following Avisynth filters: Spacedust and TemporalCleaner.

I have been getting good results, however in my backup of my son's Star Trek Nemesis (2 cd encode) I notice at times that Picards face would shift.
@Telemike, use the PicardFaceUnShift() filter :D

@Ric1234,

I noticed that you use max=2350. Would you try max=2530?

[telemike]

Most likely it was CCE rendering at a low bitrate.

I may try again at CVD resolution. Out of habit I have been using 480x480 svcd resolution.

[r6d2]

One of Bach's brilliant posts which didn't make it after he left was this:


Quoting Bach:
Let me try to make this CVD hype more clear:

=>quality comparison:
->paramethers:
average bitrate (avr);
bitrate/block (bpb);
visual quality(q).

Results:
a) for bpb=constant:
q of DVD > q of SVCD > q of CVD > q of VCD

b) for avr=constant: This is not so static but, most of times, we have the following:

b.1) if avr > 3500 kbps:
q of DVD > q of SVCD > q of CVD > q of VCD

b.2) if 3500 > avr > 1800:
q of SVCD > q of CVD > q of VCD > q of DVD

b.3) if 1800 > avr > 1000:
q of CVD > q of VCD > q of SVCD > q of DVD

b.4) if 1000 > avr > 600:
q of VCD > q of CVD > q of SVCD > q of DVD

if avr<600, any standard resolution looks bad with mpeg2. Maybe we should try 64x64 resolution

So, it is not simply a "what looks better" question.

I don't know the theory behind it, but coming from Bach it should have a good one. If any other of this forum's founding fathers would elaborate on this it would be great.

[Holomatrix]

I used to create CVD's in the b.3 senario range but used to get thoes "analising" Lines, (diagonal lines) can't remember the right word rigt now, through the video at certain parts so I went back to SVCD and the problem went away.

[r6d2]

Originally posted by Holomatrix
I used [...] to get thoes "analising" Lines, (diagonal lines) can't remember the right word rigt now, through the video at certain parts so I went back to SVCD and the problem went away.
My personal experience is BR even down to 1550 (Edit: 16:9) still look good with SVCD, but I surely would like to know the math behind Bach's statement.

PS: The word is "aliasing", but you were kidding, right? :D Anyway, I'm pretty sure there should be a solution to aliasing with AviSynth, but I'm clueless about it.

[Holomatrix]

not sure what you mean by the "Math" of it.
All I get from:

b.3) if 1800 > avr > 1000:
q of CVD > q of VCD > q of SVCD > q of DVD

is, if Max Avg bitrate (avr) is less than 1800 but greater than 1000 then @ the resolutions of each of the 4 formats, quality of CVD is greater than VCD, VCD greater than SVCD .....

Sorry if you are looking to get more out of thoes statments, I didn't really go to deep with them.

We'll at least for me the solution to the 'Aliasing' :) was to encode to SVCD instead.

[r6d2]

Originally posted by Holomatrix
not sure what you mean by the "Math" of it.

By math I mean the theory that justifies this...

[Holomatrix]

Do you really think there is a real theory behind this, I think it is just his 'Opinion' on the testing he has done. If you know Bach then you would know that he thought his Opinion was from God's mouth to his ear :) :) Maybe that's just my opinion ? :)

[ric1234]

Originally posted by r6d2
@Ric1234,
I noticed that you use max=2350. Would you try max=2530? Yes sure, I did the test last night but unfortunately it completely failed... I ended up with a 3rd .mpeg file of... 5.2 Gb ! (yes Gb!)
I don't know why it did that but I must have screwed up something since I used Recovery to save a bit of processing time...

Anyway, first I thought you definitely had spotted the bug (my 2350 value comes from 2750 - 192x2 for 2 audio tracks, to keep director's comments sometimes). But then I did the test with CCE2.66 and max=2430 (2750 - 192 - 128, director's comment still) and the estimation D2SRoBA makes was a little worse than before :
- Adjust Q : (50 + 1.0 - 0.13411485580138 - (1.6)) = 49.2658851441986
- Determined Q. : 49
I'm looking for a determined Q of 44 on that encode I think.

So... I don't know... What I'll do is for a next conversion with 1 audi only, I'll set max back to 2530, because I know the test I did was not max=2530 precisely as you suggested.

And to bring a good news down here, look what reply I found in the Basic Forum from @dvd2svcd himself :
Originally posted by dvd2svcd
Any similar tool for AVI to DVD-R?
Soon I might have something for you. Keep an eye on my homepage. Hopefully soon we might get rid of all the disk space hassle, let's all buy DVD writers !

Ric

[r6d2]

Originally posted by Holomatrix
If you know Bach then you would know that he thought his Opinion was from God's mouth to his ear :) :) Maybe that's just my opinion ? :)
Well, in fact I think he got his opinions from God's mouth to his ears, very well supported indeed (he'd not have taken them otherwise :)), but he refrained to give us the "supporting" parts because we were lesser creatures unable to understand. :D

Originally posted by ric1234
Hopefully soon we might get rid of all the disk space hassle, let's all buy DVD writers !

Yeah, I also read DVD2SVCD's post. Can't wait to see what's in his new version.

Regarding the max=2530, it has to do with the "threshold" concept Bach explained about it. When the threshold is too low, then OPV gives a damn about quantization, so it's like encoding with CBR.

The point is that you have to set thresholds which are educated guesses according to the resolution and specs of the format you're using, as well as the compressibility of the movie, to have OPV (or TMPGEnc's CQ mode) do a good job.

Otherwise, you will get an encode anyway, but it won't be "constant quality", even though you may be convinced about it.

[ric1234]

Another subject, does anyone know if D2SRoBa takes into account the thread priority defined in Dvd2Svcd when it runs CCE ?

Because I generally set CCE's priority low but right now, it seems to take to much CPU over other stuffs I'm doing... wonder if it's not running in "normal" priority ?

Ric

[r6d2]

Originally posted by ric1234
[B]Another subject, does anyone know if D2SRoBa takes into account the thread priority defined in Dvd2Svcd when it runs CCE ?

Because I generally set CCE's priority low but right now, it seems to take to much CPU over other stuffs I'm doing... wonder if it's not running in "normal" priority ?

Don't wonder, see the properties in task manager. And no, D2Sroba does not use DVD2SVCD priority. You have to set it up in CCE using ECL. Set up once and it's persistent.

Regards,

[El Leon]

Hi, when I fill the CD size with 990 and number cds with 2, the target bitrate rise to over 3400KBps and the final Q is 1, the final image size is bigest than 2 cds.
I want to use 99 mins CDs, any one can help me?
Thank you

[DDogg]

@ Tylo, maybe my memory is going but I thought you were honoring the frame selection in dvd2svcd? If so I fear it is broken. This was in a normal multipass mode. I'll also check in one pass. [edit: also not working in 1pass]. I guess you just have to pick up the AviSynth_Script_file_FS1.AVS instead of the normal one as it has the trim command already in it.

Also, I wonder if you could allow a "cut" or "0" bitrate in the tweak section? Sometimes, especially on low bitrates, I would just prefer to cut some of the credits instead of encode them, even at a low bitrate.

[davidh44]

Originally posted by DDogg
@ Tylo, maybe my memory is going but I thought you were honoring the frame selection in dvd2svcd? If so I fear it is broken. This was in a normal multipass mode. I'll also check in one pass. [edit: also not working in 1pass]. I guess you just have to pick up the AviSynth_Script_file_FS1.AVS instead of the normal one as it has the trim command already in it.

Frame selection in DVD2SVCD + D2SRoBa + 1 pass VBR has always worked for me. I cut out the credits using frame selection.

[homerjay]

Originally posted by DDogg
Also, I wonder if you could allow a "cut" or "0" bitrate in the tweak section? Sometimes, especially on low bitrates, I would just prefer to cut some of the credits instead of encode them, even at a low bitrate.

thats a brilliant idea :D i second that request

actually just remembered a question i wanted to ask - i thought while playing with d2sroba that i once set the Q i wanted to encode at and got the conversion to start immediately but cannot remember what i did :o anyone got any ideas :confused:

[ric1234]

When you use the bitrate tweaking, you have to run a multi-pass VBR, right ? I mean from my understanding of D2SRoBA, you can not have the best of both worlds : OPV with Q estimation + bitrate tweaking, am I right ?

Because this is exactly why I proposed to Tylo the feature of "Automatic overflow cut on last CD" ("Adaptive cutting"). In my idea, it was to achieve the same result as what you are talking about : cut the end credits, but also, we could still use OPV.

What I do in my conversions is : set a big Q-Adjust (-6) to force overflow and then re-cut last mpg file to loose 2-3 mn of end credits and make sure my last CD is filled up completely. But right now, I re-cut manually (TMPg or Dvd2svcd Recover).

The only problem is that it is not very precise... :)
Ric

[r6d2]

Originally posted by homerjay
i thought while playing with d2sroba that i once set the Q i wanted to encode at and got the conversion to start immediately but cannot remember what i did :o anyone got any ideas :confused:
Hi, @homerjay,

I think this can happen in two cases (first one not verified though):

1. You just got lucky and the Q you set gave D2Sroba the precise BR for the disk size for that movie. So, no need to do more testing (very unlikely).

2. You say you were playing... Maybe from previous runs there were already lots of D2Sroba_Q? files on the Video dir. and the process was straightforward since D2Sroba uses them unless explicitly deleted.

[homerjay]

i choose door number two r6d2 thanks :D
there were definately lots of previous attempts in there, thats got to be it lol

[DDogg]

@davidh44, yep, it was some type of glitch as I was doing a lot of restarts. It is now working and this allows credits to be whacked off in frame selection..

@ric1234, you can already do this. Check the "Use Frame Selection" box in the conversion tab. A lot of folks do not even know it is there although DVD2SVCD spent countless (read 100s) of hours perfecting it. It is truly an amazing piece of under appreciated work. Then, since I was incorrect above, Tylo's one pass process will pick up the trimmed avs script. It works well.

[tylo]

Hi all, I'm back from a nice vacation. :cool:
A lot of activity has been going on here since I left! Great to see that you folks are also doing support now!

I have started a new thread, as a few of you suggested.
I'll post planned new features and fixes very soon there - just need to catch up a bit first.

Please, continue discussions in the new thread.
Cheers - tylo

[DDogg]

I got a PM which I assume was meant to be posted. Since the senders mailbox was full, I could not verify that.


Hi,

Bach asked me to send you this PM - people here have locked his account - in order to clear some misunderstandings about his method.

-->>>>Robshot method: a tweaked 2+ pass method, done using some specific CCE settings.
The first pass is OPV, with Q.factor=60. In this pass, you DO NOT NEED to create the mpv file.
The second(+) pass is multipass VBR, reusing the vaf file created at the firts pass, after some bitrate tweak done in "advanced tab" in order to have green zones only, with the lowest average bitrate possible(according to CCE manual: green=good, gray=bad, red=shit).

-->>>>RoBa method: This - not so good - name was given by DVD2SVCD. It shoud be called "Bach's 2+", since it is a very extensive modification of Robshot method. The thread about it is still there. A lot of settings corrections. A very different tweaked 2+ pass aproach.
The first pass is OPV, with the worst Q.factor of your choice (mine is 40. For CCE, Q concerns to quality and the smaller the better.), and you DO NEED to create the mpv file.
The second(+) pass is multipass VBR, reusing the vaf file of the first pass AND an average bitrate CALCULATED using the size of the first pass as reference.
-Example to ilustrate the method:
a)movie Time length: 140min.
b)first pass:
-- Setting:OPV, Q.factor=40.
-- result: mpv size=1433422KB, audio size=131250KB, total size=1754682KB, #CD to use=(total_size/cd_size)=1.91.
-- Decision: Calculate video averate bitrate (CAVG) necessary in order to fit 2 CDs.
c) do multipass reusing vaf from first pass and using CAVG.
WEAKNESS OF ROBA: if CAVG > 2*(1st_pass_mpv_size/time_length) or CAVG < 0.5*(1st_pass_mpv_size/time_length), you MUST recreate the vaf file in order to do the multipass encoding.

-->>>> Bach 1passVBR: could be also called "fast 2 pass" method. Here you will use the result of just OPV. However, since OPV have no size boundary - you cannot setup its average bitrate - it is necessary to perform a first sampling enconding in order to predict the value of Q necessary to fill in the CD(s). This is the essence of this method: predict the value of Q necessary for you desires and perform OPV with this Q. The primary step consists of a crucial choice between two things:
a)if you want quality: set the worst value of Q and do the sampling encoding, and use the number of CDs which CCE needs for that quality. Do some other sampling encodings in order to fine tune Q until you get a value wich fills the CDs.
b)if you want fixed number of CDs: set the number of CDs and do sampling encondings searching for the value of Q which fills your CDs.
The difference between (a) and (b) is that in (a) you let CCE decide the number of CDs to be used. So, (a) deserves to be called "fast RoBa method".
WEAKNESS OF BACH 1PASSVBR: it is impossible to achieve a value of Q.factor apt to fit 100% of the CD set. There is an error of +-2% inherent to this statistical prediction.

Comparing RoBa method with Bach method, it is possible to see that the weakness of one method is the strong point of the other. So, lets do both of then and have THE METHOD (at least for those "fill cd freaks": simply do Bach's sampling prediction in order to achieve the value of Q which must be used in RoBa. As simple as that.

[r6d2]

Now Bach's is going for anonymous postings? Is he embarrased to show his face again?

Honestly I don't see any misunderstanding cleared up by this post, except claiming authorship who nobody is disputing anyway.

[DDogg]

erm, don't go there please. Bach's account is locked so he could not post directly. I miss Bach's brilliant intellect a lot, even if he is/was a young/old cranky rascal at times. I don't think it was meant as anonymous and it may not have even been meant to be posted. I hope I did not make a mistake in doing so. I think the purpose was to help clear up some of the thrashing (probably mine :)) that we have been doing about these methods.

[r6d2]

Originally posted by DDogg
erm, don't go there please. Bach's account is locked so he could not post directly.
Bach was not striked, at least on the late period he was a member of the forum. His decision to leave was his own. Nobody expelled or exiled him. If he wants to come back, he may ask for his account to be unlocked. If the mods decide to grant him that, the decision is up to them. If they decide not to, he does not belong here anymore.

But hiding behind respected members of the comunity as you, I think is not a grown up thing, so what else is new.

He also may ask permision to nobody and sign up again as Chopin, Haendel, or any other famous musician, which would be another childish thing to do.

But he hasn't, so I think he does not want to.
I miss Bach's brilliant intellect a lot
I think we all do. Me personally learned a lot from him. But he made a lot of people work really hard after he left to reinstate his kwoleadgeable posts. I think the very first thing he can do is to apologize, before just claiming undisputed authorship of his brilliant ideas, or lecturing us about "what he really meant to say".

Just IMHO, and I shut up on this subject because the dead deserve respect.