I have a very bizarre problem the cause of which I can't determine. I am the proud owner of the box set of Three Colors trilogy. For my own viewing convenience, I want to transcode each DVD with the KDVD presets so that I can fit all three movies onto a single DVD5 disc (I don't have DVD9s, nor do I want to buy any at this time). Each movie is less than an hour and 40 minutes long, so fitting all three on one disc should be no problem.

In my first attempt to do this, using the unmodified KDVD Full D1 preset with motion search precision set to Higest quality, no filters used, and video source settings to match my DVD source, I was able to achieve a suitable file size for each movie even at CQ 75. However, when I played back the transcoded files, two of the three MPEGs had very jerky motion, as if many of the frames had been cut or dropped. The files were all encoded to be 23.976 fps (internally 29.97 fps). I checked my source files, and they were fine. I re-encoded the files and the result was the same. I finally thought of a good explanation for this: TMPGEnc was reading the source files with the telecine flag switched on, so even though the source files were technically encoded with 23.976 fps (progressive), they were read as 29.97 fps (interlaced). (This is standard practice for film DVDs and what is meant by the 23.976 fps (internally 29.97 fps) parameter in TMPGEnc.) Therefore, TMPGEnc was trying to convert the source running at 29.97 fps into 23.976 fps without IVTC, just simply decimating frames, frames which happened to be original progressive frames (not the telecine-generated duplicates) for two of the three movies (the decimation actually worked out okay for the last movie). The best and only way that I could attempt to fix this was to use DVD2AVI (now DGMPGDec) and AviSynth to frame serve the source at the proper 23.976 progressive rate (i.e. using "Forced Film" in DGMPGDec). After transcoding once more, now with the AVS scripts, the output MPEGs played just fine as expected, so I was right. But here is where my problem for this forum starts.

I transcoded using the same parameters as I had since the start (except for the video source settings, which I changed to match the now progressive source). Remember that originally the encoded files had acceptable sizes at CQ 75. But now, after frame serving, the M2V streams alone totaled 5.05 GB. I set the CQ to the KDVD specified minimum of 70. After re-encoding, I was fortunate to see a drop of almost 900 MB. But this was not enough, because sum of the sizes of the M2V streams, along with three 192 Kbps AC3 streams (one for each M2V), and the rest of the overhead from my DVD-lab PRO project, came to be 4.72 GB, which was still quite a bit too large for burning. (Yes, I've tried burning the disc.) I've done everything right, at least as far as TMPGEnc and KDVD go. But when I don't frame serve, the file sizes are fine, and when I do frame serve, the file sizes are inflated. Any thoughts, guys?

Hi d17, and welcome

Have you given CQMatic and CalcuMatic a try
This way, you can get a very close file size to what you need.
Check the CQMatic forum for download links.

-kwag

There is no mystery there. The file size can have been lower with the frameserving also. In fact it is unpredictable because you changed the conditions (wrong IVTC movie -> correct progressive movie). That is like if you try to encode two movies completly different.

Do as Karl recommended but I suggest you to also learn a little about script ans filters because, if you are already at 70 and this is too much, then when the correct size will be found, the result will be barely unwatchable.

Using filters will make the filesize to reduce, while the CQ used will be increased.

Of course I'm not expecting the resulting file sizes to come out the same. Your "encoding two different movies" analogy is correct, but where you take it is either wrong or unclear. If I'm transcoding a 29.97 fps interlaced source into 23.976 fps progressive output, and the method of frame rate conversion is decimation -- NOT IVTC -- of course the set of frames that will be encoded will be quite different. This is where you are right: I am essentially encoding a completely different movie than I am encoding with frame-serving. But my problem/question has little or nothing to do with how encoding a decimated source and encoding a frame-served source are different. Let's forget for a moment that I can transcode a telecined source and say/pretend that the source is progressive WITHOUT the telecine flag. Thus, I would encoding directly from 23.976 fps progressive to 23.976 progessive. Because the total length of the source videos are significantly less than six hours, shouldn't I be able to achieve a final total file size less than that of a full DVD5? Or, were the KDVD/KVCD presets never designed to transcode from progressive sources and work (as intended) only with inverse-telecine frame rate conversion?

If the material is widescreen (not fullscreen), and if you use a correct script with efficient filters, yes. Without that, don't expect to put more than 4 hours on the disc (even 3 if fullscreen material)

Majority of sources we use are progressive (native or grant to a correct avisynth script). Some people encode true interlaced sources directly into interladed mode, to prevent problems and gain some quality (but bitrate needed is more).

But the most important word is : widescreen

Okay, but I'm still not quite understanding you. I'm transcoding a DVD source, which is of course at 720x480. Widescreen or full screen should not matter, as in either aspect ratio types, the source and the output are fixed at 720x480. As for filters, I don't quite see the point. The KDVD Full D1 template is advertised as allowing an average of around six hours of video fit on a 4.7GB disc. Nothing is said about filters or about using tools other than TMPGEnc. What I'm trying to do isn't even ambitious. If most source videos encoded with KVCD/KDVD are progressive, and the average results come out as advertised, then why am I having these file size issues? What filters could I use? The only one that seems any bit useful in my case is noise reduction. Remember that I'm simply trying to transcode DVDs, and I want to preserve the video as much as I can. All other filters in either TMPGEnc or AviSynth modify the video in ways that I find undesirable and unnecessary. I would gladly listen to you if you would please try to be clearer. If you tell me to use filters, then say what filters I should use and how they would help me (and please keep in mind my aims in this project). You treat my situation as a matter of course, and while I'm certainly no newbie to video encoding, I simply can't understand what is in your mind that makes you so enlightened about what I'm doing wrong. To put it simply: KDVD promises to fit around six hours on a DVD5 disc. I encode from M2Vs with an active telecine flag, TMPGEnc decimates some frames, and I get files that are of expected size. Then I try to encode a properly progressive source, and the file size almost skyrockets. And now you simply confuse me with saying the "majority of sources we use are progressive" (which should mean that I shouldn't be having these size issues), some ill-explained business about widescreen vs. full screen that does not seem to be applicable to my situation, and more ill-explained business about filters that apparently without them make me be able to fit only four hours on a disc. You have not directly answered my question, only gone on some unhelpful tangents. What you say on the surface even seems self-contradictory. So, not to insult you or disregard your experience, knowledge, and good intentions (all of which I think are in abundant supply), but do you actually know what you're talking about, or are you simply throwing out suggestions in attempt to be helpful without a clear understanding of the why?

A black border compress a lot better than a picture area. So widescreen or fullscreen is very important.

And it is not say that you can have this with a source directly taken from the DVD. Isn't it ? In fact no precision is given on the source at all.
With the templates you can have up to 6 hours, but all depend on the source. It is not said that you can use whatever you want and obtain a great 6 hour DVD. They are templates, not miracles

(edit: to use an analogy when an adds says "this car can run 130 mph and pull a 2 ton weight", you don't expect it can run 130 mph WHILE you are pulling a 2 ton weight. do you ?)

You make a confusion between VCD templates that gives you exactly 80 minutes of video on a CD and a KDVD template that can give you up to 6 hours on a DVD (and not "an average around" 6 hours).

See there :
http://www.kvcd.net/forum/viewtopic.php?t=3483

Only if they are used badly.
But if you are more confident with a very light script, then use simply Certainly your interrogation about what you are doing wrong while you are doing nothing wrong. You just expect something that can be done the way you think.

KDVD is "avanced video conversion" and that's a way to tell that it's not just a matter of using a template and feeding tmpgenc with the original vob sources. There's a little bit more behind this. That's all.

But surely not of the expected quality ! Restart from the begining : I suggested you to use filter because I suspect that you will have a too bad result without them. I don't tell you can't try without this, I just gave an advice.
Don't be fooled by your first encode that was by chance near to the target (in word of filesize). Yes it was closed, but it has problems. Now you have to forget this first "problematic" encode and just understand how to obtain a correct one.

You asked if the template are or are not designed for progressive sources. I answered that almost all our sources are progressive (that was the answer to your question, right ?). But when did I say that our sources are not filtered ?

Again you asked a general question about the templates ("my sources are less than 6 hours so the template should put that on a DVD-5) and I answered a general answer about the templates ("yes it is possible, but only if they are widescreen").

You are supposed to not be a newbie (what I understand immediatly, even before you tell it). The answer are focused on that. Now if you need to have anything explained from scratch, just ask !
That's funny because more often people are worried because we answer to them as if they are a 5 years old boys

I have near 100 KDVD on my shelves that can testify I know what am I talking about. Don't worry about this
This is more to you to ask for details, like you are doing here, if you do not catch something I said than to doubt about what am I saying.

As you hail from France, I can only assume that English is not your most fluid language, and you may admit yourself that your posts have shown that your English is less than perfect. Therefore, some leniency may be granted if you have a slightly hard time explaining things precisely and in detail. But as I said, I trust your experience, since your 8500+ posts in this forum should pretty much give you the title of "expert," if not "guru." But I think that the problem all along has been a lack of adequate clarity and decent explanation, not just from you but from KVCD.Net itself.

Allow me some serious criticism: Many impassioned newbies on the DVD back-up scene will sooner or later happen upon the KVCD project and declare it to be "the best secret of the ripping community." The fact that one can fit up to 50% more video on CD (and maybe more with a DVD) while maintaining a decent level of quality should be considered a minor miracle; don't deny that. But when the inexperienced (or even those who simply take a quick look at KVCD) go to download the templates, they are taken to a page that features a chart with seemingly useful information (dvd-models.html). I go to download the template for KDVD Full D1, and right above it is small text saying "~6 Hour DVD Quality." Right next to that, above KDVD Half D1, is "~10+ Hours Great Quality." What does that mean? When the point is tried to be conveyed that way, it says that with KDVD Full D1, I can expect to fit about six hours on a disc. It doesn't mean up to six hours, and it does not explicitly mean a consistent average of six hours, but it suggests maybe as low as five hours and maybe as high as seven hours. But when I get only a little above four hours, that’s a major difference, and it’s wholly inconsistent with what the download page promises. If it’s meant to say “up to six hours,” then that’s what should be written there. And the page says nothing about filters or additional special measures. The majority of the KVCD community might be applying additional tweaks to get six hours or ten to eleven hours, what the download page says is simply misleading. Similarly, if one were to download KDVD Half D1, one would expect to get about ten hours on a disc, if not more, because that’s definitely what “~10+ hours” communicates. When someone like me experiences perceived size issues because of what the advertising makes one believe, he/she has to then turn to the forums for help, but the unpleasant surprise is that you have pay $2.99 to get that help. Now I realize that the site and project need some financial support, but when the user is given misleading information that all but forces him/her to go to fee-based forums, the whole project is cast in a devious light. Site admins: If you don’t want this, then please make the download page (and other pages advancing the promises of the KVCD project) much clearer. That’s the end of my criticism of the site.

When you mentioned the black bars, your “widescreen” point was finally clarified. Perhaps my mistake was not in considering that in the first place, but you should’ve mentioned the black bars when you first tried to make your point. Of course, instead of widescreen, you could’ve made the point simply by saying “letterboxing.” Saying “widescreen” includes anamorphic widescreen (the 16:9 flag in the MPEG-2 spec.), which is what I was trying to do. Now that you have made this point clear, I may try it. Although the one question that I have for you now is whether in your opinion you think that adding black bars would make the resulting picture quality better or worse than using filters to soften the image.

I realize (and your assessment of my personal knowledge and experience in video encoding should make you aware that I realize) that several variables will make output file sizes very hard to predict, especially since KVCD/KDVD uses CQ instead of two-pass VBR. The source is the key, and transcoding the Three Colors DVDs shows that first hand: Blanc has the shortest running time by a few minutes yet is by several hundred megabytes the largest of the transcoded M2Vs. I would certainly have a hard time fitting three copies of Blanc on a DVD5, even with filters and tweaks. And since CQ is in itself a form of variable bit-rate, I certainly couldn’t make the same assumptions about KVCD/KDVD as I can with the regular VCD standards. So your point here is a little off the mark. However, like I said in my general criticisms about the site, if KDVD promises “~6 hours,” I expect “~6 hours” or even five hours, but certainly not only a little over four hours.

Your next point seems valid, but it’s still a little misguided. Yes, I did think that I did something wrong because the results did not meet my expectations. However, the fact that I didn’t do enough to get my desired results is an act of doing something wrong. It’s like plugging in a television and getting all fuzzy channels because the antennae was forgotten is making the mistake of negligence (or too much eagerness to watch that TV). Certainly, I knew that using filters would alter the size of the output files, but I was not made aware that I had to do that. Not even the FAQ talks about the use of filters to achieve the desired file sizes (more shame on you, KVCD.Net!). (It only says that using noise reduction will result in a cleaner image and smaller files instead of something like “’Why are my files coming out way too large, even when I use CQ 70?’ In many instances, the use of filters is required to get the expected file size” or something like “Make sure that output is encoded in 4:3 aspect ratio with letterboxing if you have a widescreen source.”) My faulty expectations were created by KVCD.Net’s nebulous promises, so please shift some blame to the site and the project.

Forgetting the first encode may be good advice, but the reason why I couldn’t forget it is because of the dramatic difference between the results of encoding with frame decimation and frame serving. And more than that, I was furthermore alarmed with the fact that I could no longer fit my videos onto a single disc, even using CQ 70 (remember, I could fit all three on one disc with the first encode, which used CQ 75). These two things are what compelled me to sink down that $2.99 and ask for help. So I wouldn’t say that I was so much “fooled” by my first encode as mislead by nebulous promises of KVCD in the first place.

The question that I had originally asked was not whether or not the templates were designed for progressive sources. That was a closely related, side question. My original question was “What am I doing wrong?” or better yet, “Why are there these unbelievably enormous differences between my two different encoding procedures?” Intermediate and expert video encoding enthusiasts alike know that normally, similar encoding parameters with similar sources yield similar results. The amount of file size inflation that I experienced was 25%, which I don’t think should be considered normal. I am troubled when you bring up the point that you had never mentioned that your sources were not filtered. Once more, this continues to show how potentially misleading the KVCD project is.

Finally, the reason why I alleged that you went on unhelpful tangents was because you didn’t get what my original question was (and if I had made not made the original question clear, then I apologize), and when you mentioned things you didn’t explain them very well and used unclear/imprecise language, like the whole widescreen point. Therefore, I was simply confused, and your lack of explanation and clarity made you seem like you were just throwing out ideas instead of taking the time to educate me, which I would’ve welcomed, since I tried to clearly convey that I was lost in this issue. “Throwing out ideas” is typically a hallmark of someone who only half understands the subject at hand and merely wants to convey good intentions. I have right and reason to doubt you, because remember that sometimes even the experts are stumped.

I think that that about covers everything that I have to say, and I hope that this heated discussion need not continue as it has done so. So to conclude, I have four options before me: I can change from trying to encode anamorphic widescreen to letterboxed widescreen; I can use the recommended AviSynth filters to clean/soften the image; I can simply try a slightly lower CQ value (against the recommendation of the KVCD community, of course); or I can just try two-pass VBR. I lose some quality any way I choose, but I really have no other choice. So which option do you seriously think will give me the best quality in the end? Anyone in the forum is free to answer this one. Thanks.

Hi there d17,
You can wait for Phil's answer and/or for the kvcd.net Admins answer but anyway, here goes an "all good intentions" post from newbie.
When I bought my DVD burner I had the same dilema: should I stick with non-anamorphic letterboxed encodes or should I move on to anamorphic encodes.
I knew that if I took the anamorphic path I would end up with bigger file sizes that would lead to less quality on my encodes.
At the time I was already using Avisynth scripting for quite some time and I already had some knowledge on which filters could help me.
So I did a very simple test run.
I did a 1 x movie on a DVD-5 anamorphic from a PAL source and I did a 1 x movie on a DVD-5 letterboxed non-anamorphic from the same PAL source.
No Avisynth filtering except for the resizing on the non-anamorphic encode.
I have to say that I was very much well impressed with the letterbox encode but I fell in love with the anamorphic encode.
Plus I wouldn't say that kvcd methods are simple Tmpgenc templates.
No, sir.
Karl keeps them there so that newbies can have a smooth start on the real deal.
I used them for less than a month and then I completely forgot they even existed.
Once you grasp the concept you won't need templates any longer and you will be able to use the same methods with other encoders out there.
As said in the begining of the post this is an "all good intentions" post .
Oh, and I'm sorry if English is not my middle name .
Cheers

__________________
Rui

Lol.
I'm working for Nortel (a Canadian company so) and spent several hours a day on phone with Ontario people or our Indian partners in Bengalore. I hope my english is good enought to not be fired in the next quarter
But I agree, it's not my native language and this can explain some lack of accuracy in my sentences.

After having read all that thread again I saw for instance that you are right when you say "around 6 hours" : this is what is indicated on the page where you can DL the templates, while on the homepage of kvcd.net it is written "up to 6 hours".

The same when you say that I go on some tangents rather than to answer to your direct question : in fact we are walking in the same room but not looking at the same wall. You want to understand why your 2 jobs where so different while I try to tell you how to just obtain a correct one.
But I think that in the end, what you need is a decent KDVD, and not really to understand the why.

For sure. And that's our job to explain them that it is not. Isn't it ? We are not priests of a secret church. At least I'm not.

As noted above, you are right, this page is erroneous. Even if, before to reach this page, all people are supposed to have read the home page where you can read "Using KVCD parameters to create DVDs (KDVD), will enable you to create 100% DVD compliant MPEG-2 streams, capable of playing on any standard DVD player. This will allow you to put up to about 6 hours Full D-1 720x480 on one DVD, or about 10 hours at Half D-1 352x480."
Again it does not say either that no other things than the template are used. By thinking the opposite you are just giving to the words the meaning you want them to have.

A filtered source is nothing more than a "highly compressible" source.
In a perfect world, you should be abble to use directly sources from your DVD, but we are not in a perfect world.

No fully true. You can try to have help somewhere else. VCDHelp for instance that talks about any format and/or conversion toppic. I'm not sure you will have the same help than you have here. This added value can justify a little fee.

Not offense. It's just a matter of having the same idea behind a word. Obviously you didn't have the same than I have.

Excuse me but "Anamorphic widescreen" are fullscreen only if the original A/R is 1.77 (that is perhaps the case for your present sources). But the majority of movies have A/R from 1.85 to 2.35 and have black borders even if you encode in anamorphic mode (the border is just smaller than if you encode them in letterbox).

Definitely better. The compressibility gain of the introduction of a black border in the picture is bigger than whatever filter can give to you.

I see you have audio stream at 192 kbp/s. Are they the original AC-3 streams ? Or did you convert them from 5.1 to 2.0 ?
If you converted them, then think about goind to 128 Kbp/s insteed of 192. Except if you are very sensitive to audio quality. Personnaly I give priority to video.

Good remarks. I think Karl continues to read this thread.

Let come back to something that I thought was "agreed" on both sides : your two encodings procedure were like encoding two sources completly different.
Things were crystal clear for me that, once this was understood, your question vanished by itself. Because I don't think you will ever ask "why Blanc takes several megabytes more than Rouge to encode ?".

I think that is where my answers started to go in a different way than you were expected . Sorry about that.

Are you skilled with the CQ mode in tmpgenc ?
Do you know that is it not linear and there is a part in the CQ curve where changing the things just a little bit may increase the filesize by several percents ? Tmpgenc is full of surprises like this

Don't think I do not understand your trouble. But once again this is only in your mind that took birth the idea that whatever source can be used. When I first read the KVCD specification, two years ago, the first think I had in mind was "how the hell do they torture the picture to obtain that ?".
Of course, my 15 years background in MPEG format are not a common situation. I agree with that.

Don't worry, you are a lot clearer than most of members

That's what my teachers told "different level of understanding". I was sure to be clear enought . Again, I'm sorry about that.

I took no offense

In your situation (because I live this kind of situation also !), what I do is first to lower the audio bitrate, then to use a very simple "Removegrain().Deen()" script, then I would go to letterbox.

I know this last statement is opposite to what I tell upper about black border vs filtering. So let me explain : with a black border you will have a better CQ but if you plan to watch your DVD on a big screen, perhaps not today but in the near future, you will regret to have them 4:3 and not anamorphic. Don't forget that in few years, 4:3 set won't exists anymore.

At least, this is my choice.

Note: I forgot; I'd cut the end credits also. These are bitrate killers.