Whats the BEST way to create 720x480 XSVCDs ??

[SansGrip]

Wulf's numerous posts on this matter have suggested skipping dvd2avi altogether and loading the vobs directly into TMPGenc.

I assume that would involve using the installed DirectShow filter/decompression codec for YV12. I've had problems in the past with MS arbitrarily overwriting my codec decision for that task and replacing it with a version that seems only able to produce 16-bit output. I try to stay away from DirectShow as much as possible in this process.

It would be interesting to test whether this does indeed produce correct pixel value ranges. Provided a properly functioning decompressor is installed it should, but there's no way of being sure without testing it. TMPGEnc's "Output data as Basic YCbCr not CCIR601" option will have a bearing here, so anyone doing this should figure out whether this setting should be on or off, otherwise their pure black won't be 16, it'll be either 0 or around 30.

This is the "frameserving" process I was specifically referring to.

When I hear "frameserving" I automatically think Avisynth. Sorry about that .

Sure decomb would give even more consisent results but the best way to get 24fps material is to simply not telecine it at all. I see no reason to telecine something just to inverse telecine it immediately after.

That's generally true, but I've noticed that with some material (most recently American Pie) occasionally scenes do seem to slip through the net and interlace artifacts are clearly visible, despite DVD2AVI claiming >99% FILM. Using Avisynth and Decomb would catch these aberations.

Also, as you noted, I kinda doubt the extra conversion in avisynth would even yield a noticable difference. It certainly could never cause enough damage to offset the %20 decrease in quality you get when encoding in ntsc vs ntscfilm.

This is certainly the case when space-limited. If not one could of course simply increase the bitrate a corresponding amount, but in almost all cases one is encoding to a specific size of media.

I've heard some argue that keeping interlaced material interlaced is the optimal solution for TV viewing, but having tried both I honestly can't see any difference (I refer to MPEG-2, of course).

Basically my point is don't skip dvd2avi unless you actually have a valid reason, and increased quality certainly isn't a valid reason since you actually get the opposite.

If one could say for certain that TMPGEnc's built-in IVTC is capable of properly producing 24fps material from the telecined source, and that the DirectShow filter, the VFAPI plugin and TMPGEnc maintain correct -- i.e. within CCIR-601 -- pixel value ranges, and that the entire source material is contained within a single vob, and that one is able to accurately syncrhonize the audio stream with the IVTC'd encoding, then it should be possible to encode directly from the vob with no quality loss.

But that's a lot of "ifs" .

Regards,
SansGrip

[kwag]

If one could say for certain that TMPGEnc's built-in IVTC is capable of properly producing 24fps material from the telecined source, and that the DirectShow filter, the VFAPI plugin and TMPGEnc maintain correct -- i.e. within CCIR-601 -- pixel value ranges, and that the entire source material is contained within a single vob, and that one is able to accurately syncrhonize the audio stream with the IVTC'd encoding, then it should be possible to encode directly from the vob with no quality loss.

But that's a lot of "ifs" .

Regards,
SansGrip

Maybe it's time to write a driver ( that would be fun ) that can read a VOB and send 24fps directly to TMPEG? Doing the Force FILM in the background, instead of having to wait for DVD2AVI to create a pseudo avi (pointer) file. This way we could bypass the DVD2AVI process completely. As far as containing the complete movie on a VOB, there's an option on DVDDecrypter to do that. So the complete movie appears as a single VOB. And audio would be processed separately as it's usually done, or just demux the AC3 with DVD2AVI. So I see that wulf109 has a valid point after all

-kwag

[adam]

If one could say for certain that TMPGEnc's built-in IVTC is capable of properly producing 24fps material from the telecined source, and that the DirectShow filter, the VFAPI plugin and TMPGEnc maintain correct -- i.e. within CCIR-601 -- pixel value ranges, and that the entire source material is contained within a single vob, and that one is able to accurately syncrhonize the audio stream with the IVTC'd encoding, then it should be possible to encode directly from the vob with no quality loss.
But that's a lot of "ifs" .

Regards,
SansGrip

Right but the same result can be gotten through dvd2avi w/forced film but without the "ifs" and in substantially less time. That's all I was really trying to convey.

Unless you are a PAL user than I honestly can't think of one reason to skip the dvd2avi/mpeg mediator step, and even if you are PAL its still a good idea.

[adam]

So I see that wulf109 has a valid point after all

-kwag

Where? You can possibly get the same result only IF you do an IVTC which can take hours. So you may get the same quality in a longer amount of time but no guarantees. I don't see any benefit in that.

Loading the vobs directly or frameserving via dvd2avi shouldn't make any difference because its still VFAPI doing the processing. Dvd2avi simply allows you to keep the 24fps as opposed to removing the overhead to get back to where you started. At the very least it just saves time.

Now a filter or plugin to skip the dvd2avi step and still get 24fps would be nice.

[kwag]

So I see that wulf109 has a valid point after all

-kwag

Where? You can possibly get the same result only IF you do an IVTC which can take hours. So you may get the same quality in a longer amount of time but no guarantees. I don't see any benefit in that.

Unless a driver is built that will deliver, for example, 24fps on the fly and then we're reading the true DVD color space directly without any overhead layers like VFAPI.

Now a filter or plugin to skip the dvd2avi step and still get 24fps would be nice.

That's what we're talking about

-kwag

[adam]

Now a filter or plugin to skip the dvd2avi step and still get 24fps would be nice.

That's what we're talking about

-kwag

Only as of your last post. I think the point of the discussion was whether or not there is any benefit to skipping dvd2avi, or another form of frameserving, and instead loading the vob directly into the encoder. As of yet, the answer still remains no.

[kwag]

Only as of your last post. I think the point of the discussion was whether or not there is any benefit to skipping dvd2avi, or another form of frameserving, and instead loading the vob directly into the encoder. As of yet, the answer still remains no.

If you look at the sources of DVD2AVI, that is if you know anything about coding, you'll notice that DVD2AVI uses some of the code from the MSSG group. So the VOBs are actually being decoded into memory buffers, etc.. Don't you think these layers ( or multi layers ) introduce some errors and/or variations? Of course they do! If a driver/plug or whatever you want to call it can be written to bypass all this bloated layers, we get to the true source material without any alterations. So the answer is "yes", there is a benefit if this is developed, and we remove legacy MPEG decoding layers, as there are in DVD2AVI.

-kwag

[adam]

I think you are missing my point Kwag. Theoretically, a driver/plugin can be made that can do anything, what good does this do anyone in the now? The key word in your post is, "if."

If such a driver is made then great, I will use it. Until such a driver is made there is no benefit to loading a vob file directly into TMPGenc because I consider direct show more of an altering/detrimental "layer" than anything that dvd2avi does. Wulf claimed there was a quality benefit, about 50 times...people believed him and they wasted their time so I spoke up.

No I don't know anything about coding. What does this or an undeveloped plugin/driver have to do with this thread? You changed the subject and now you are trying to argue about it.

Sansgrip, I look forward to your response if you have anything further to add.

[SansGrip]

Right but the same result can be gotten through dvd2avi w/forced film but without the "ifs" and in substantially less time. That's all I was really trying to convey.

That's also what I was hinting at .

That said, DVD2AVI on its own can't really handle hybrid material properly. These days I tend to turn off force film and use Decomb to ensure a clean IVTC.

[SansGrip]

So the VOBs are actually being decoded into memory buffers, etc.. Don't you think these layers ( or multi layers ) introduce some errors and/or variations? Of course they do!

Theoretically they shouldn't. Remember, this is digital, and we can in theory copy the data an infinite number of times without losing or corrupting any information. vob -> DVD2AVI -> Avisynth -> VFAPI -> TMPGEnc might be considered inefficient, but if one is careful it's possible to retain the information in almost exactly its original form.

If TMPGEnc allowed YV12 input then with Avisynth 2.5 we could do the entire process in the same colourspace. I'm not sure if CCE can accept YV12, but if it can it's one of the few areas where it remains somewhat superior to TMPGEnc for MPEG-1 (the other main one being speed, of course).

[SansGrip]

Until such a driver is made there is no benefit to loading a vob file directly into TMPGenc

Sounds to me like what kwag is suggesting is either a multiple-vob-reading DirectShow filter or VFAPI plugin. To be honest I'm not sure what advantage this would provide over DVD2AVI, since that's almost an exact description of DVD2AVI already.

I actually feel that both vob-via-DirectShow and d2v-via-VFAPI are suboptimal methods of encoding, since what you're effectively doing is taking DVD-quality material (i.e. material preprocessed specifically for very high bitrate encoding) and trying to reencode it at much lower bitrates without modification.

Ideally because of the bitrate difference one would want to apply additional preprocessing in order to prepare the material for encoding with the lower bitrates used by (X)VCD and (X)SVCD, and this is precisely what frameserving with Avisynth or VDub allows us to do.

It also allows us to attempt to compensate for the fact that the material has already been compressed in a lossy manner, for example by dithering areas below a certain detail threshold which can help hide existing blockiness in the source. Visible blocks are a particular problem when reencoding with DCT-based compression, since each iteration tends to emphasise the flaws of the last.

While this is just personal opinion, I would argue that attempting to reencode at relatively low bitrates material preprocessed specifically for DVD bitrates will usually result in worse quality than if one were to take advantage of the additional preprocessing provided by a filtering frameserver.

Regards,
SansGrip

[Zaun]

Wow, you guys are like, smart, and stuff. Im glad I got you thinking about all these ways to make dvd conversion better. BTW I have posted a new question, if any of you know the answer, please give me a shout. This was all vry interesting read thats for sure.


- xaun

[SansGrip]

BTW I have posted a new question, if any of you know the answer, please give me a shout.

I would help you if I could, but I don't use CCE .

[vhelp]

Evening all. .
.
.
So, who wants to create this dream-driver ??

I'm not MPEG guru, but surely there MUST be someone'near

I have some belief on this "color/quality" issue, cause I have a very
sensitive eye DV

-vhelp

[SansGrip]

So, who wants to create this dream-driver ??

I think it's already created. It's called DVD2AVI .

Regards,
SansGrip

[vhelp]

.
.
what!!TT@@TT@@

..you mean it's already ben cReAtEd (dvd2avi) he, he...

.
.
. . then what was the point in all this argument
later.

-vhelp

[SansGrip]

then what was the point in all this argument

Just thrashing out some technical issues .

Regards,
SansGrip