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

[dvd2svcd]

Here's an excerpt from the CCE users guide:
What is a video information file ? Generally at least two
passes are required to create a variable bitrate stream for specifying
an average bitrate. This is because the complexity of
images is checked and saved in a file by the first pass, and the allocation
of the bit amount for each frame is calculated according
to the stored information on the complexity of the images, and
encoding is executed in the next pass. Unlike a general encoder,
Cinema Craft Encoder does not distinguish between a pass for
encoding and a pass for analyzing complexity, but always analyzes
the complexity of images during encoding. Therefore,
a video information file does not always have to be recreated
to change the setting of an average, minimum and maximum
bitrate.
A video information file has a history of information for a plurality
of times of encoding, and information is accumulated each
time encoding is executed. This information improves the image
quality for encoding by the Multipass VBR system.
Creating and recreating video information file A video
information file is created by encoding by the CBR or the Onepass
VBR mode. Setting at this time should be close to the
setting for the Multipass VBR mode. If CBR is selected to
create a video information file, set the bitrate for CBR to the
same value as the average bitrate to be specified for executing
Multipass VBR.
A video information file has information for each image frame.
Therefore, values when a video information file is created are
used for parameters which have a major influence on the encoding
of image frames. These parameters are as follows.

Setting chapter point
Setting GOP configuration (N, M, GOP header interval,sequence header interval)
Progressive frame flag
Block scan order (Zigzag or Alternate)
Aspect ratio

To change the above parameters, the video information file must
be recreated. To change other parameters, the video information
file does not have to be recreated. If the setting of the
bitrate is a major change, however, it is better to recreate the
video information file because a better encoding result can be
obtained with less number of passes. When a bitrate is set to
twice or more or half or less than the average bitrate, recreating
the video information file is recommended.


So you see, what you say just cannot be, or you just have been extremely "lucky". How did you compare the two VAF files?

[1loser]

Originally posted by dvd2svcd
Therefore, values when a video information file is created are
used for parameters which have a major influence on the encoding
of image frames. These parameters are as follows.

Setting chapter point
Setting GOP configuration (N, M, GOP header interval,sequence header interval)
Progressive frame flag
Block scan order (Zigzag or Alternate)
Aspect ratio

To change the above parameters, the video information file must
be recreated. To change "OTHER" parameters, the video information
file does not have to be recreated. If the setting of the
"BITRATE" is a major change, however, it is better to recreate the
video information file because a better encoding result can be
obtained with less number of passes. When a bitrate is set to
twice or more or half or less than the average bitrate, recreating
the video information file is recommended.I'm not changing any of these. On the second pass I add the filters and change the resize from simple to bicubic (same aspect ratio). This cuts the encoding time of the first pass in half. From what I just read, I would think that I CAN use the same vaf file. Where does it say if I change the filters or resize from simple to bicubic I need to create a new vaf file? I realize I'm just a NEWBIE but until I hear from Cinema Craft, I believe that the method I'm using is valid and it's always given me good results.

[1loser]

Originally posted by Bach
are you blind? Did you read my last post and don't understand it, or you just want to keep posting nonsence? I must repeat: it is IMPOSSIBLE to re-use the vaf file in such situation. If you edit and change the avs file AFTER the vaf creation, this vaf is a reference about a stream of pictures which simply doesn't exist anymore. You are not simply changing "other" parameters. You are changing the whole source, and this is simply twice worst.
Assuming you are editing the avs file correctly, I BET your second pass is a new_vaf_creation+n_pass. The other possibility is that you forget to press "save" after editing the avs file and just "think" you have a new avs, but all that you have is the same old avs.

@dvd2svcd

please, delete 1loser's last two posts. There is no place for nonsence and misinformation in the Advanced forum. What a shame :rolleyes: Now you're sounding like an arrogant person. I read your post and I read the CCE guidelines that dvd2svcd guoted.Originally posted by Bach
If you think that what you say is possible, then why should I create a new vaf for every movie? I.e. I could just re-use the x-men's vaf in the 1st pass of LOTR conversion. Can you get the point?You are talking foolishly here. I may be stupid but NOT as stupid as you are implying. I consider the vaf file as a finger print for THAT movie and as noted by the CCE guidelines it's only necessary to create a newer vaf file for THAT movie when certain things change and NOT necessary when other things change. Did you read those CCE guidelines carefully? I think you are not humble enough to realize that even a NEWBIE like myself could be right. I'm not saying I am right for sure but your attitude sucks!!! I personally think your last post should be removed instead of mine. I'm surprised you didn't ask him to delete my user acoount also. If you want to be a leader than act it a proper manner and not like you're better than others.

[dvd2svcd]

Well, Bachs attitude may suck, but I understand it, you don't seem even to try to understand that what you're doing can't be done. My guess to why this might work for you is that you probably have patched CCE with the CRC patch to avoid CRC errors. No matter what you will stop this nonsense at once, because you are misleading other users. Feel free to contact Cinema Craft and get confirmation.

Both, cool it or I will lock this thread (which will be a pity).

[chainsaw135]

@1loser Hey man there is nothing wrong with being wrong. The thing we dont like around here is when someone is wrong and wont admit that mistake and still wants to shove useless information on to our users. This community is all about new idea's and the discussion of new or even old ideas. We have a place in this community for people that throw around idea's so with that said i hope you dont take that as an attack as I'm not attacking your intelligence but on that note you are sounding really "silly":) right now and your views make no sense. Thats my 2cents.

[Linux]

One other viewpoint to this vaf-reusing.

Even if adding filter is changing more or less every picture that will be compressed, there ought to be some correlation between the old and the new movie. Every frame in the picture is in the same place so the relative compressness between the easy and the hard part of the movie should stay the same.

I believe that the original idea to this was to get a known quality corresponding to the choosen Q-Factor.

If one is only intressed with better quality but maybe not exactly the same every time, I think that it would be intressting to make some tests using 1loser's split filter method before saying it is nonsense.

What I meen is that even if a academical and teoretical view say it is nonsense it maybe a good aproximation.

[1loser]

Originally posted by Bach
Thus, it is IMPOSSIBLE to re-use the vaf file after you edit the avs file. If you make ANY changes to the avs file you WILL get a check sum error. So Bach's statement could have read that it is IMPOSSIBLE to re-use the vaf file after you edit the avs file UNLESS you have patched CCE with the CRC patch to avoid CRC errors. My earlier post on page 3 SAID I was using the CRC patch.

I encoded "Jackson Brown Going Home" twice.
1) using 2 different avs scripts (no filters on the first pass and simple resize) than reusing the same vaf file but adding all filters and using a bicubic resize for the second pass.
2) using the same script (with all the filters in place and bicubic resize on the first pass). The results where about the same. I'm not really sure if one was better than the other, each person would have to judge for themselves.

I'm not recommending this method to anyone because it's NOT the proper way to encode video and that's what this forum is all about. At the time I made my first post I didn't realize it was only possible because I was using the CRC patch. The ONLY point I wanted to make is that it was infact possible because I've done it about 20 times with results that I PERSONALLY thought were good but I know VERY little about video encoding and that's why I'm hear to learn how to do things in the proper manner.

I just want to stress this point, I don't recommend using the patch so you can use this method and I already removed the patch from CCE.

Update: I had to add the check sum patch back because even when I keep the avs file and ALL the settings the same I get the check sum error at 35 % completion. That's why I used it in the first place. Maybe if I did a fresh install of windows XP I wouldn't have that problem anymore

[emilius256]

This are my test results with RoBa for the movie Patch Adams PAL time 1.50:29
2 Audio Tracks 160 Kbps and 1 Subtitle.
Bitrate Max. 2530 Min. 300 Max. Avg. 2230 Cd Size 800.
4 pass, Image Quality Priority 24, Anti Noise Filter 2, Q.Factor 30, Bias 20 [X]Linear Quantizer Scale, [X]ZigZag Scanning Order, [X]Progressive Frame, Intra DC Decision 10.
In the AVS script Bicubicresize is used and no other filter.
The Average Bitrate used was 1680 and it came out 2 800 CD.
The quality is really awesome, when i look in my pc there's a little more "Mosquito noise" than a normal 5 pass, but on my tv is not really notable.
So good, good thing.

[Clixo]

@emilius : to remove that mosquito noise use temporalsoften2(3,5,7), is is making wonders here whith that problem

[emilius256]

Thank you Clixomano, i'm re-encoding the same movie with temporalsoften2.

[TsR]

Hi!

I’ve got some thoughts about the RoBa method in D2S. Tried the beta versions a couple of times on some chapters and it works fine. Some things that I’m missing though is an option to manually set the bitrate for the number of CDs you gonna use. “Calculate the average bitrate that should be used to fill 1CD (avb1)”… I suppose that D2S does this automatically and that’s fine in most cases. But it would be nice to be able to calculate it yourself with a prog like Fitcd (depending on other mpeg tracks, menus, pics etc), and of course you can but that makes it a whole lot more time consuming. Cuz you gotta do all the steps of encoding, multiplexing, and burning with chapters yourself.

Another thing is the part where you calculate the bitrate for the first pass (“avr_bitrate=mpv_size*8.192/(movie_time_length*60”). If you encode just one chapter or part of a movie it would be nice to be able to set the movie time length manually to, instead of D2S having it set to the chapter length. If D2S set the length you will end up with a very high average bitrate which won’t be the case in the final movie.

It also seems like D2S sets the max bitrate in CCE to the same value as max average bitrate in D2S and not to the max you set there. That is not what you want.

So if it possible I would like to see those two options implanted. Would make the encoding with this method a lot easier. For me anyway…

While I’m at it I’ve got one more suggestion. I’ll post it here even though it isn’t in the right place… It’s support for over scan. I know, it isn't hard to edit the avs script and I do that, but there would be nice to be able to have an easy preview before I encode. This because I live in Sweden and almost uses subtitles every time. As it is now the subtitles always moves down a little bit form the preview in D2S so I gotta load the avs into virtual dub to see that they’re at the right place and not to low. So support for over scan is an option I would very much appreciate.

Thanks for a great program!

[Cidici]

Originally posted by clixomano
@emilius : to remove that mosquito noise use temporalsoften2(3,5,7), is is making wonders here whith that problem

im very interested in this mosquito noise "panacea" u suggest.
could u please explain me where i should put it ?

thank you in advance

[Clixo]

@cidici: http://forum.doom9.org/showthread.php?s=&threadid=22096&highlight=temporalsoften2.

the search button is very helpful also

[switu]

Hi
@ DVD2SVCD
is it possible to implement this "TemporalSoften2" to DVD2SVCD
Thanks

[Cidici]

Originally posted by clixomano
@cidici: http://forum.doom9.org/showthread.php?s=&threadid=22096&highlight=temporalsoften2.

the search button is very helpful also

you were definitively right, but i dont understand one thing : the mosquito noise im talkin about is the one added by the CCE processing and it's *not* present in the original movie (it's a dvd source), so how could a avisynth filter remove it ?
i suppose the filter works in this way :
original movie -> avisynth + filters -> cce -> mpeg output

so it couldnt remove something that has not been added yet.

am i wrong ?

[Linux]

Originally posted by TsR
Some things that I’m missing though is an option to manually set the bitrate for the number of CDs you gonna use. “Calculate the average bitrate that should be used to fill 1CD (avb1)”
If this is what you want you dont want this method. Use ordinary VBR/CBR and set a fixed value for the number of CD on every line.
If you encode just one chapter or part of a movie it would be nice to be able to set the movie time length manually to, instead of D2S having it set to the chapter length. If D2S set the length you will end up with a very high average bitrate which won’t be the case in the final movie.
To do this you just set the CD size to some corresponding size like 15 Mbyte
It also seems like D2S sets the max bitrate in CCE to the same value as max average bitrate in D2S and not to the max you set there. That is not what you want.
I think D2S is working properly. What max is you talking about.
Have you set the Max avg 200 kb/s lower than Max or have you tested to set Avg Max higher that Max which is inpossible?

[trott]

So if I'm reading this method correctly...No let's start over: while I can see the validity of not using linear quantization scale, as for the rest of it I have 2 problems:
1: to really get the 'what's the average bitrate cce would recommend provided I give it the quality I want my mpeg to become' question answered, should you not put the maximum bitrate during the cq-encode pass to the maximum of the encoder? If I say I want q=25 but 2500 kbit/sec can not reach this quality during certain scenes, the entire thing falls apart, no? (though maybe not by much)
2: my major gripe, though I may be misunderstanding something: At the end there's some calculating involved, which basically goes: (bitrate recommended by cce - bitrate to put the entire thing on x number of cd's)/bitrate to put the entire thing on x number of cd's.
Then you check whether this equation is less than or equal to 0.1.

Now, though this looks interesting, could you not just simply say: divide the recommended bitrate (by cce) by 1.1? (or, simply put, take 91 percent of the recommended bitrate as your average bitrate, and see whether it fits on let's say 2 cd's. If not, take 3. If it still doesn't, take 4. and so on.)

Basically, you're giving a complicated calculation which means: let cce calculate the bitrate given a certain quality, and then make sure your bitrate is minimum 91 percent of that bitrate and fit it on the amount of cd's needed for that bitrate.

What it boils down to is: why take 0.1 as the result of your equation? Because when using only something like fitcd and common sense (oh let's see, a high-action movie, better not let that bitrate get below 2200) I could get the same-quality encode, no?

Presumably this is a way to get something resembling the compressibility check in gordian knot for divx?

I might be completely wrong here, though...(or confusing the hell out of everybody :))) )

[Cidici]

Originally posted by Bach
1: to really get the 'what's the average bitrate cce would recommend provided I give it the quality I want my mpeg to become' question answered, should you not put the maximum bitrate during the cq-encode pass to the maximum of the encoder? If I say I want q=25 but 2500 kbit/sec can not reach this quality during certain scenes, the entire thing falls apart, no? (though maybe not by much)No, because you are using CVD or SVCD resolution, and want to use compliant bitrates. However, for DVD resolution, it is better to set max=9800, which is DVD compliant anyway.2: my major gripe, though I may be misunderstanding something: At the end there's some calculating involved, which basically goes: (bitrate recommended by cce - bitrate to put the entire thing on x number of cd's)/bitrate to put the entire thing on x number of cd's.
Then you check whether this equation is less than or equal to 0.1.

Now, though this looks interesting, could you not just simply say: divide the recommended bitrate (by cce) by 1.1? (or, simply put, take 91 percent of the recommended bitrate as your average bitrate, and see whether it fits on let's say 2 cd's. If not, take 3. If it still doesn't, take 4. and so on.)

Basically, you're giving a complicated calculation which means: let cce calculate the bitrate given a certain quality, and then make sure your bitrate is minimum 91 percent of that bitrate and fit it on the amount of cd's needed for that bitrate.you got the point. These are the results of that calculation, which is not complicated anyway... I was just avoiding to use the word "percent", since a recent research have shown that 4 of each 5 Americans (=80%) at college don't know what "percent" means... I was just trying to help them:DWhat it boils down to is: why take 0.1 as the result of your equation? Because when using only something like fitcd and common sense (oh let's see, a high-action movie, better not let that bitrate get below 2200) I could get the same-quality encode, no?fitCD and common sence are not reliable resources for compressibility check. Example: A very sharp calm picture will use a bitrate higher than that used by a blurry high-action movie. Common sence say the opposite.Presumably this is a way to get something resembling the compressibility check in gordian knot for divx?the main difference is that here the compressibility check is done while encoding.


Originally posted by Bach

you got the point. These are the results of that calculation, which is not complicated anyway... I was just avoiding to use the word "percent", since a recent research have shown that 4 of each 5 Americans (=80%) at college don't know what "percent" means... I was just trying to help them:D


i hope you were joking...

[Cidici]

today i noticed a strange behaviour of the roba method :
i encoded CUBE (sci-fi movie, 88minutes, PAL, 4:3) and finally i got .... a SINGLE 80min CD with an "amazing" average bitrate of 811Kbit !!
i used it with 4-passes but u can guess how the quality is: ugly, too many blocks, poor definition, noise and so on, it looks like a VCD

the first test i made with this method was really good (star trek first contact, 106minutes, PAL -> 2 CDs) but this one really sux.

i know the movie is quite short, but i was planning to fit the "usual" 2 cds, so i must re-encode it with the standard multipass method

so i think this time the roba method failed :/

[trott]

No, because you are using CVD or SVCD resolution, and want to use compliant bitrates.

True, but we're looking for the optimal value recommended by cce here, and then we'll be trying to approach that value as best as possible. Therefore, should we not be working within the limits set by the encoder rather than those imposed by the svcd specs?

i encoded CUBE (sci-fi movie, 88minutes, PAL, 4:3) and finally i got .... a SINGLE 80min CD with an "amazing" average bitrate of 811Kbit !!

When I calculate the quality using this method, I seem to get around 1600 kbits/sec. Now, as I seem to recall, mpeg-2 is a pretty good codec, but it was really designed for higher bitrates. (sat average around 3 Mbit/sec.) I think I read someplace that, due to the design of mpeg-2, it's not recommended to get below 1800 kbit/sec using an svcd resolution, even for scenes with a low amount of detail and a small amount of fast-moving frames. Maybe this method should have an addition: 'but do not go below X kbit/sec' ?

Maybe I should clarify, since the chosen quality be cce seems to be about right when doing entire movies...I created an avisynth script which selects only a part of the movie, much like it is done in the scripts generated by gordian knot for divx encoding. When doing a 1-pass q-based encode on this, I _always_ get a lower-than-expected bitrate...

Could the difficulty here lie in the fact that divx is really optimized for low bitrates and, let's be honest, really s*cks at bitrates above 2 Mbit/sec? (which is basically why I don't believe in its future: the moment dvd media becomes as cheap as a cdr, why do a divx movie at all?)
Mpeg was created for higher bitrates but we're using it at a low bitrate here, could this explain for the fact that doing a relatively low-bitrate encode on a movie which is basically only parts of the movie at set times a la gordian knot compressibility check results in a too-low bitrate?

I do believe firmly in the concept here, but do not believe it to be practical if you have to do a full encode first to get accurate results...

[vivaatloni]

I have used the method with a avi file (divx, very good quality). The movie was 1:43 min. The output was 2 cd's (800mb) with very high quality. It plays like dvd quality on my dvd player.

Very good guys...keep up the good work.

[KitKat]

I did a few tests w/ Whole Nine Yards over the weekend.

I abtained optimal results with:

CCE settings: iqp=5, Q=40, antinoise=12, bias=10, progressive[x], zigzag[x], linear[].
bitrate: ended up w/ ~1850.
machine:athlon 1.4 GHz (roughly realtime)
my opinion: 5-pass quality w/ 2-pass encoding time. Hence very good overall. Car-on-fire sceen: fire is excellent. Moving-crowd-plus-swaying-trees scene: not so good, noisy.

Anything (repeat: *anything*) higher than 5 for iqp gave me very obvious mosquitoes.
Using sharpen would only worsen (repeat: *worsen*) the mosquitoes.

Any comments/recommendations?

[KitKat]

A funny result:

A movie of 1h56. Max bitrate of 2350. At the end I got two .bins of the normal size (825,xxx) and a third one of 927,xxx! :eek: What the...? :confused:

[Linux]

Originally posted by KitKat
Using sharpen would only worsen (repeat: *worsen*) the mosquitoes.
Sharpen sees the mosquitoes as details.
This is the reason I seldom uses sharpen.
It only takes time to make the artificial sharpen.
I can get the same effect if I adjust sharpening on my TV.

[tylo]

Great news, folks: Finally a tool
that does Robshot/Bach in DVD2SVCD!

I know this thread has been silent for quite a while,
but I think it belongs here anyway. Responses regarding
this tool are welcome.

--
D2SRoBa is a DVD2SVCD plugin that automates the Robshot/Bach One-pass VBR method, described earlier in this thread.
Also, read this thread for background info on this tool: http://forum.doom9.org/showthread.php?s=&threadid=41033

Features:
- Predictable one-pass VBR encoding with CCE - fills last CD to the edge
- Automatic estimation of optimal number of CD's, based on Worst Q. factor
- Setting of exact number of CD's (overrides DVD2SVCD)
- Setting of a range of CD's (see below)
- Automatic credits bitrate tweaking - available through RB's EclCCE.
- Full batch/cmd line support with "one-click" to prepare it!
- Status window - tells what's going on
- Fixes the OPV max bitrate flaw in DVD2SVCD
- Works with CCE 2.50 and CCE 2.66+.


vStrip for DVDDecrypter, is another DVD2SVCD plugin, which allows DVDDecrypter to be used as dvd ripper.

Download latest versions at: http://home.no.net/tylo/
--

[VILLA21]

U forgot the download link? :)

[tylo]

Originally posted by VILLA21
U forgot the download link? :)
Well, I added it as an attachment, but I think it may take some time before it is put up. However, I made a page where you can download from: http://home.no.net/tylo/ :D

Please, report how this works for you.
Tylo

[tylo]

I just uploaded a new version at http://home.no.net/tylo with quite a few updates.
---
Not a single response on this tool so far, although a fair amount of people has downloaded it. It would be helpful to know at least if it worked on your system, or not.

Tylo

[arlsair]

Ok, here some feedback:
I made a small test of the DVD Decrypter Plugin.
First I would find it better, when you also install a "vstrip.ini", so I can make the chances without fist start a process and abort it to create the .ini .
Also a solution (and a feature request) would be a installation routine, where you can ask for the DVD Decrypter directory (like RB`s AutoFitCD)

And then, why do you put the DVD Decypter Plugin and D2Sroba in one package. I would like to download and install only that, what I need.

Beside that all, good work guy. For me it works and I know, many people would like this.

[tylo]

Originally posted by arlsair
I made a small test of the DVD Decrypter Plugin.

Thanks for those suggestions. Done them. In addition, the plugin now optionally configures DVD2SVCD to use it. I will upload later today.

Still nothing on the D2SRoBa tool... ;)

Tylo

[arlsair]

About D2SRoba I think, it is only useful, if you do a 1pass VBR.
For a multipass VBR the internal routines of DVD2SVCD are faster, because the estimation for the amount of cds is done by a 1pass VBR which can reused for the following multipass VBR as a video information file. So there is no encoding of the 1% of the movie.

And for me and many others VBR is state of art.

[RobertR]

But dvd2svcd is only taking the lenght of the movie when calculating number of disks, while RoBa method considers also the compresibility of a material.

[arlsair]

No, there is a special RoBa mode.
You have to edit the dvd2svcd.ini and set RoBa Visibel=1.
Then you can choose in the encoder tab the roba routines.

But keep in mind, that this mode isn`t supported by the mods and dvd2svcd himself. So don`t ask dvd2svcd for help or feature request.

[RobertR]

Originally posted by arlsair
No, there is a special RoBa mode.
You have to edit the dvd2svcd.ini and set RoBa Visibel=1.
Then you can choose in the encoder tab the roba routines.


I must admit i didn't try program that Tylo wrote but i suposed that it's just different (maybe better) approach to RoBa method with dvd2svcd. And yes I know about the above settings in dvd2svcd.ini (i'm almost sure that i found info on it in this very thread).

But keep in mind, that this mode isn`t supported by the mods and dvd2svcd himself. So don`t ask dvd2svcd for help or feature request.

From what i understood reading posts here RoBa method is not quite stable and sometimes can give strange results (it's effect heavily depends on source material) and that was main reason to hide it. I might be wrong of course :D

[Holomatrix]

Originally posted by tylo
I just uploaded a new version at http://home.no.net/tylo with quite a few updates.
---
Not a single response on this tool so far, although a fair amount of people has downloaded it. It would be helpful to know at least if it worked on your system, or not.

Tylo

Hi, it seems good. I ran into a problem today were it wasn't clicking the Encode button for CCE 2.67.00.09, the mouse kept jumping down an inch, when I did a chapter test earlier it when through ok. Still testing :)

EDIT: I just noticed when doing a full movie that the birate did not get adjusted to my max CD size. I ended up with to many CD's at max bitrate

[tylo]

Originally posted by Holomatrix
Hi, it seems good. I ran into a problem today were it wasn't clicking the Encode button for CCE 2.67.00.09, the mouse kept jumping down an inch, when I did a chapter test earlier it when through ok. Still testing :)

EDIT: I just noticed when doing a full movie that the birate did not get adjusted to my max CD size. I ended up with to many CD's at max bitrate

@Holomatrix: About CCE encode button: This is the only critical part of the plugin: I need to calculate a mouse position (relative to the window) to hit the Encode button. I use an offset (about [-80, -30]) from bottom center of the window. I have tested with CCE 2.66.01.07, which seems to work ok. If this remains a problem I can put the offset values in the ini file.

@Holomatrix: There was a *MAJOR* bug in version 1.2: It didn't use the D2Sroba.avs file with the SelectRangeEvery() function in! It just used the first 1% of the movie (not very representative!) -- Please upgrade to version 1.3

However, I am not sure what you mean with "too many CD's" - D2Sroba computes the required number, which is "how many you need".

NB! You *may* get too big files if the distributed 1% sample of the movie is not representative, but so far in my testing, the mpv has rather become shorter than estimated - than longer. You may give D2Sroba bigger slack by increasing the bbmpeg_cutoff value in the D2Sroba.ini file.

@arlsair: For multipass, D2Sroba is solely used for estimating the number of CD's based on quality - that takes 2 minutes. Is speed really an issue here?

Thanks for the feedback.
tylo
http://home.no.net/tylo

[Holomatrix]

Originally posted by tylo
@Holomatrix: There was a *MAJOR* bug in version 1.2: It didn't use the D2Sroba.avs file with the SelectRangeEvery() function in! It just used the first 1% of the movie (not very representative!) -- Please upgrade to version 1.3

However, I am not sure what you mean with "too many CD's" - D2Sroba computes the required number, which is "how many you need".


Ok, I'll try 1.3, I though I was using the latest version, whoops:) and for the "too many CD's" thing I am just saying that I've tried two movies and set DVD2SVCD as I normally would in the bitrate TAB for a two CD encode but ended up with 2 1/8 CD's. I did one movie as 1Pass VBR and the other Multi-Pass, same prob. I will try again using 1.3 and see how the calculations work out.
Cya :)

[arlsair]

Originally posted by tylo
...that takes 2 minutes. Is speed really an issue here?

Yes, you are right. The couple of seconds doesn`t matter

I can`t download the DVD Decrypter Plugin. I click on the link and nothing happens.

[tylo]

Originally posted by arlsair
Yes, you are right. The couple of seconds doesn`t matter

I can`t download the DVD Decrypter Plugin. I click on the link and nothing happens.
Sorry about that, I'll fix that soon.
Download: http://home.no.net/tylo/vStripDVDDecrypter_setup.exe

tylo

[arlsair]

Thanks for the link.
The readme of the DVD Decrypter readme is the old one for D2Sroba.

[Holomatrix]

Just to confirm something here, I can use this plugin, set Multipass VBR 1pass, set the number of CD's I want as per minute movie and due to the technique of the Roba method I should get the same quality as if I was doing a 3pass encode. correct?

EDIT: Well I just tried it with Back to the Future and it picked 3 CD's. Not what I was expecting. Oh, well maybe someone will have luck. Maybe I'm doing something wrong

[tylo]

Originally posted by Holomatrix
Just to confirm something here, I can use this plugin, set Multipass VBR 1pass, set the number of CD's I want as per minute movie and due to the technique of the Roba method I should get the same quality as if I was doing a 3pass encode. correct?

EDIT: Well I just tried it with Back to the Future and it picked 3 CD's. Not what I was expecting. Oh, well maybe someone will have luck. Maybe I'm doing something wrong
This is really not very well documented by me. You don't set the number of CD's. Actually, D2Sroba ignores both the minutes and the number of CD's specified in DVD2SVCD Bitrate tab. It only use the Q factor you set in the Encoder tab. The Q Factor is a direct meassure of the quality that you want. You specify the minimum quality you can tolerate (higher Q). You will then always get at least this quality, but the number of CD's depend on the compressablity of the movie + length. (compressablity varies almost as much as the length!)

D2Sroba will compute the number of CD's you need to achieve this minimum quality, and sets this number for all time intervals in the DVD2SVCD.INI (before the Recover phase). Also the CD sizes is set this way (given in D2Sroba.ini). I would guess that if you use a Q factor about 40, you'll get only 2 CD's with your movie. Also, try RB's AutoFitCD plugin - it will reduce number of CD's needed but maintain quality.

A one-pass VBR encode will not give the same quality as a 3-pass (with the same av bitrate), because CCE will improve on the bit allocation distribution for each pass. But you definitely get the best quality per minute encoding time - by far.

Edit: In short, the Roba method ensures the quality that you want - the number of CD's you'll get is just a result of that. Sometimes it will produce fewer CD's than you would expect, sometimes more.

tylo

[tylo]

Originally posted by Bach
thanks for your effort tylo. You are the kind of "newbie" that should ever post here:D

I have just realised your posts so that I have not tested your tool yet. Shame on me... I have done by hand all those steps that your tool seem to make automagicaly. I will use it tonight and report my impressions here tomorrow. Thanks again.
Bach,
Your ideas here made me write this tool. I had a lot of fun coding it too, because I "break" into another app (dvd2svcd), but it let me get away with it. (thanks dvd2svcd, for the recover function). I'll be reading your opinions, questions, or suggestions for improvements with interrest.

tylo

[switu]

tylo great work !!
are you planning to add avi2svcd support?:D

[MictXP]

I'm having trouble running this. If I run 1.3, I keep getting "need to run CCE with DVD2SVCD" error then closes. If I run 1.4, I get "Syntax: GetFileSize <file>" Pressing OK starts encoding, followed by that error again, which starts encoding again, etc. I've tried using CCE 2.5, 2.66.01.07, and 2.67. Any ideas?

[tylo]

I'm not quite sure what may cause your problems: Here's a checklist:

1) Configure DVD2SVCD to use CCE and DVD2SVCD - AVI2SVCD is not supported yet, but will be once DVD2SVCD support works well..

2) Versions prior to 1.4 worked only for default "Folder structure" in Misc tab.

3) The plugin needs to be installed in a subfolder of DVD2SVCD (e.g. Tylo). D2Sroba expect to find the ..\dvd2svcd.ini file (DVD2SVCD home ini file). In this file It locates the project folder ('DVD2AVI Folder' entry in [Folders]). This should point at the correct folder after DVD2SVCD was shutdown. It then reads entries from DVD2SVCD.INI in the project folder.

I have only tested on Win2K, so it may be further problems with other OS versions.
PS. I will make a version 1.5 that test that all the expected files are present,
so that it will be easy to figure out what is wrong.

Other findings: Using 1% of the movie does not seem to be sufficient in many cases. The difference between estimated and real size of the mpv file is too large. By default, I now use 2% (mod in D2Sroba.ini) of the movie: SelectRangeEvery(goplen*100/2, goplen), which seems to be sufficient.

If you find what caused the problems, please let me know.

tylo http://home.no.net/tylo

[MictXP]

Ok, I found out what was causing the errors:

"You must use CCE with DVD2SVCD"
This was caused by using a Folder Structure different from the default settings. Upgrading to 1.4 fixed this problem.

"SYNTAX: GetFileSize <file>"
This was because I had a space in my folder structure name. It was: "D:\Temp DVD\movie" Changing this to "D:\movie" Fixed this error.

However, I am still having problems :confused:. Using 1.4, after CCE is finished encoding, the process returns to DVD2SVCD, and tries to recover the file. It then stops at "Are you sure you want to reuse the VAF file?" Choosing Yes just redoes encoding. Instead, changing the D2S file position from 32 to 34 lets your tool do it's magic, and produces a fine image. I imagine this is what should happen.

Thanks for your hours put into this. I'm running WinXP Pro SP1 right now, and will probably test this on Win 2003 shortly. I get quiet a speed gain from 2003 :D

[tylo]

@ MictXP: Thanks for the report. I realized what was wrong because of the GetFileSize error. The "spaces in path" bug was also in vStrip for DVDDectypter plugin. Fixed them in v1.5. However, the other problem you describe is strange. I do delete the vaf file, so there must be something else. Ticking off "Create vaf file" will only obscure more I think.. Anyway, try with v1.5. It does more checking.

@ Bach: As you see, there were still a few bugs, but we're getting there. With v1.5 it should be simpler to track down further bugs.

tyloOriginally posted by Bach
hi tylo,

AWESOME work. My first report was using version 1.3. With version 1.4 and the default settings I've got an accuracy of 99.8% (!) in file size prediction for 1passVBR/Bach method(movie: run lola run).
btw: It doesn't hurt to use 2% instead of 1%. The file size prediction will be just 6~12min slower, and the accuracy is realy higher.
btw2: why don't you use the command line features of eclcce instead of those keystrokes? It could make d2sroba even more robust, I guess.

[Holomatrix]

Can someone just quickly explain when you say 1PassVBR for this program are you using VAF (1PassVBR+VAF)? or should I uncheck VAF? Why does DVD2SVCD complain when you select 1PassVBR but not when you select multi-pass VBR #of passes= 1? Same thing?
Thanks

[tylo]

Hi,
@ Bach, thanks for those words - just doing my best. Uploaded v1.6 with EclCCE command line usage. Try it. If you use my other plugin vStrip for DVDDecrypter, d2sroba gives an option to select chapters from DVDDecrypter.

@ Holomatrix: I think this has been answered many times before. SEARCH. Oh, well - with One-pass vbr, there is no need to create vaf. Multi-pass with #pass=1 is really a two-pass vbr (=predictable file size). One-pass vbr + vaf is used if you want to do a subsequent multipass encode. (e.g. Robshot). Under certain criteria, you may then skip the first (vaf creation) pass.

tylo
http://home.no.net/tylo