dvd ntsc=> pal conversion & adding subtitles best way?expert opinion

[Abond]

@all. would be great if you could have a look at the dgpulldown thing (the two files i uploaded) because it is really strange ... i have a perfect 23.976 .m2v file which after dgpulldown turns into parts that are jerky.
i do not want to do the assume25fps thing because it speeds up things and this is a big audioproblem when listning to the audio, very noticiable as most parts are spoken (long parts of dialog).

Read http://forum.doom9.org/showthread.php?t=108518&page=2 the 6-th post from the bottom. On the next post there is dgpulldown version that may work for you.

[poescp]

Dear All,
here is a longer part of the video, please do download it and check what you think of it and how i could convert it to PAL 25fps, with using avisynth and not using any speedup option.
http://www.dieschafe.net/goodnews/yxcv.demuxed.m2v

[FulciLives]

Dear All,
here is a longer part of the video, please do download it and check what you think of it and how i could convert it to PAL 25fps, with using avisynth and not using any speedup option.
http://www.dieschafe.net/goodnews/yxcv.demuxed.m2v

OK ...

I downloaded yxcv.demuxed.m2v and ran it through DGIndex [Version 1.4.6] to get a D2V project file. I would like to point out that when I did this I got the following error message:



This is probably due to the way you "cut" the M2V file. So I wouldn't worry about it really.

I then created an AviSynth AVS script using as a basis the last script you posted. This is what I came up with:

LoadPlugin("C:\Program Files\AutoGK\DGMPGDec\DGDecode.dll")
LoadPlugin("C:\COPY_DVD\decomb521vfr1.3\Decomb521VFR1.3_P4-Athlon.dll")
mpeg2source("C:\TEST\yxcv.demuxed.d2v")
Telecide(order=1,guide=1)
Decimate(mode=3,threshold=2.0)
Crop(0,64,-12,-66)
AddBorders(6,4,6,6)
Lanczos4Resize(720,576)
ConvertToRGB24()

This is almost exactly the same script. I changed the programe paths of course to match mine and I dropped VSFilter and the TextSub as I don't have the subtitles. I am using the same version of DeComb and the same settings. I only very slightly adjusted the Crop and AddBorders commands so as to get a perfect aspect ratio conversion. I am a stickler for that. I also changed the last line to RGB24 instead of 32 as I understand that 24 is the way to go.

I noticed that this does make the frame rate 23.976fps but I do see what I guess is some blending or ghosting here or there so this may NOT be the best IVTC method for this bit of footage. However it doesn't look that horroible considering the fact that the original doesn't have a standard 3-2 pattern. Oh yes the blending only happens at the end of the clip as far as I can tell ... the part where the guy walks past the screen and then the next and what is the last scene where he walks to his chair.

Anyways I went with that ...

OK so I loaded the script into TMPGEnc Plus and encoded it.

Here is how I did it:

1.) Start TMPGEnc Plus and kill the Wizard if it pops up.
2.) Click on the BROWSE for VIDEO SOURCE [bottom area] and select the AVS file.
3.) Click on LOAD at the bottom there and select the DVD (PAL).mcf file.
4.) Click on LOAD at the bottom there and select the unlock.mcf file.
5.) Click on ES (Video only) for the STREAM TYPE [bottom area]
6.) Click on SETTINGS at the bottom which brings up the MPEG SETTING window.

There are a few things you need to change here. On the TAB called MPEG SETTING the ASEPCT RATIO is at the deault of 4:3 so change that to "16:9 Display". The Frame Rate is at the default of 25fps. CLICK ON the SETTINGS icon next to that and change it to "23.976fps". The Encode Mode is on the default of INTERLACED so change that to "Non-Interlaced".

It should now look like this:



Now you also have to change a few other things. For instance you probably want to do a 2-pass VBR encode so you need to change the RATE CONTROL MODE to reflect that. You also might want to change the DC COMPONENT PRECISION to 10 (although some people prefer to use 9 unless they are using very high bitrates i.e., a CBR or AVG of 6000kbps or higher). You also probably want to change the MOTION SEARCH PRECISION to something other than the default ... I suggest using the setting called, "High quality (slow)".

Since this is a very short sample you have to do a CBR as it isn't long enough to do a 2-pass VBR so I set it up as a CBR using 8000kbps as the video bitrate. I used 10 for DC COMPONENT PRECISION and I used "High quality (slow)" for the MOTION SEARCH PRECISION.

Now let us go to the ADVANCED TAB as we need to change some things there ... actually the only things we need to change here is the VIDEO ARRNAGE METHOD. Make sure it is set to "FULL SCREEN" instead of the default which is "FULL SCREEN (KEEP ASPECT RATIO)". We do this because the AviSynth script already does the resizing we want etc. so we don't want TMPGEnc Plus to muck around with any resizing ... that is why we pick "FULL SCREEN"

We are now done with the settings. There is nothing to change on the GOP STRUCTURE TAB nor do we want to change anything on the QUANTIZE MATRIX tab. So just click OK on the bottom of the screen (of the MPEG SETTING "window") and this takes us back to the main screen of TMPGEnc Plus.

We are now ready to encode. So click on START in the upper left corner and let TMPGEnc Plus encode the video file.

In my case I called the AviSynth AVS file "test.avs" so in the end I get a MPEG-2 file called "test.m2v"

Then we run the M2V file through DGPulldown ...



We get a new file (in my example) called "test.m2v.pulldown.m2v"

This file is now 100% PAL DVD compliant but the "playback speed" is the same as the original NTSC video so there is no need to adjust the audio (i.e., you do not have to speed up the audio).

Playback is very smooth for me despite the bit of video blending that happens in those last very two scenes of the clip. Again that may have to just be left like that or you might have to fool around a bit more with the DeComb settings. Sorry to say I am far from an expert when it comes to using DeComb so you are on your own. If that "broken pattern" only happens here or there then you may just have to live with it (the slight blending/ghosting that happens when that pattern comes up in the video).

As for your earlier issues with encoding ... my guess is that before you were encoding incorrectly. If you aren't carefull with the TMPGEnc Plus settings it is easy to encode 23.976fps with pulldown for 29.970fps and then when you try to run that through DGPulldown ... that probably caused the stuttering or the "not smooth" playback issue.

The way I have set it up ... you end up with a video that is 23.976fps (as per the original after DeComb) and no pulldown is applied by TMPGEnc Plus ... which is what you want ... since you use DGPulldown to apply the pulldown from 23.976fps to 25fps.

I should point out that I didn't download and look at any of your samples ... the only thing I downloaded was the yxcv.demuxed.m2v file. I feel secure though that my results are better than what you were getting (based on your comments). I hope you are happy with it and I hope I gave you enough information so you understand now how to encode this yourself using TMPGEnc Plus.

- John "FulciLives" Coleman

P.S.
I have uploaded the final file test.m2v.pulldown.m2v to my website in case anyone would like to download it. The file size is 32.4MB so please only download it if you really need it ... otherwise I might go over my bandwidth rating and have to pay extra money at the end of the month. Also since the file is so large I can't leave it there for long but I will leave it for at least a few days. I will come back here and delete the link once I delete it from my website.

Here is the link to download it [right click ... save target as] ---> CLICK HERE

*** EDIT ***
One last thing ... I just watched the clip a few more times (my final PAL conversion) and it looks good to me ... better than I thought ... really the only scene or shot with any blending/ghosting is the 2nd to the last where the guy is walking past the Airport Monitors. I thought at first I saw some of that in the very last scene/shot as he is walking to the chair but I think that looks A-OK now that I've watched it closer.

Anyways let me know what you all think of it.

[poescp]

it's by far the best i have seen so far, however to be honest it still stutters slightly again and again, it is not really smooth all the way, especially noticeable at the scene where the car drives by, notice the 2 x slight stutter as it moves from right to left and also when the sliding doors open there is 1 or 2 stutters. one thing i did however notice is that the stutter only appears once dgpulldown is used. prior to that the video move perfectly. by the way i did it exactly as you described and also used your file you created (my version exactly ended up as yours).
again, it is by far the best i have seen sofar, but still could be smoother ... the challenge still exists ...

[vhelp]

The reason(s) for your "studders" will be evidensed shortly..

First - if I may ask, what source is this from, ??
ie, dvd; capture; hdtv capture; vhs; etc.

Second - After a quick analisys of your latest video mpeg .m2v source, I have
consluded that it is a MIXture of Cuts and Edits after the Telecine has been
done on it by those who own the source commercially. Instead of editing in the
raw film 24p, (though they could have edited in the 24p too, but) they quite
possibly performed additional edits to the now Telecined source, which is
before you [and us], today.

Third - your source containes bits and pieces of Field-shifted patterns.

So, you have before you, a source that consists of 3:2 patterns; field-shifted
patterns; and cut-scenes. And the cut-scenes are whats probably throwing
you off the most. Because they are done to the Telecined source, inside the
3:2 patterns, without proper care. In all honesty, there really is no
clear rule as to how one should go about "cutting" inside the 3:2 patterns.
There are variuos reasons why, but not important at this time.

Now, the reason(s) why your source continues to fail all your attempts at
bringing it (assumably so) back to Progressive (or even Film) in a clean
mannor, is because of this sources inconsistant patterns, not to mention, "types"
this source was destined to be.. ie, Field-shift in some places, while Telecined
in others, and yet, cut here or there, etc.

So, to compile very breifly, you have:

** fail: Edited randomly across Telecine patterns
** fail: Field-shift is not IVTC appliable, and will jit/jut/studder
** fail: Cuts between Scene-Changes inside/outside of bounds

Thus, as I noted with you all earlier, the only sure way (and I'll repeat
here again) is to develope the proper filter() to:

A - determine source consistancy.. ie, is it 3:2 pattern or Field-shift, etc.
B - after determ., apply *each* filter() method to those known patterns.

Unfortunately, there is a missing "C", because this is for handling cut-scenes,
but there is no way to do so in a clean manor, because cuts do occur at any
point with-in a 3:2 pattern, PPPii. And, this latest source D/L has such
cuts. But, in normal situations, you won't have "C" to worry about. So, I
didn't inlcude it in the above A,B list. As I was saying..

This filter() should be able to do this automatically. But, if auto is not
applicable or required, then a user-entry type filter with known values to
process video source accordingly, might do.

It is assumed that most video sources are Film-based..
And, because of this, when starting any video conversion project that might be
IVTC 'able, it is wise (after determination steps) to start the IVTC at a known
pattern entry point. I always set my IVTC mechanism at the first "P" entry,
PPPii .. If you start your IVTC at the point of your video's starting position,
which is usually position 0 [zero], it is likely to fail the ivtc process. But
this fact is dependant on the actual filter() 's capabilities. Perhaps, some
filter() 's spot this, and "bandaid" these pieces together via "blends" so
that they may continue with the ivtc pattern processing through-out the re-
maining video source.

Below, is a partial compiled list of what I have found through-out the video
source posted for D/L by member, poescp.

frames: - [006] < PPPii PPPii .. PPPii > [100] 3:2 patterns
frames: - [101] <:> is a scene change, where frame is Interlaced
frames: - [101] < ii > [102] these were interlaced frames during scene change
frames: - [103] < PPPii PPPii .. PPPII > [207] 3:2 patterns
frames: - [208] < P\i > [210] was a scene cut*[1] (not scene change)
.
.

*[1] - this "scene cut" was obviously a part of a 24p->Telecine source, where
this Telecine became the new Original, and the editors began using this source
for additional editing work, and the sequence of frames that were a part of this
Telecine, PPPii PPPii, was Cut/Pasted into the .m2v source posted, hence, frame
208\209 is a scene cut, starting with frame 209, and merged finally with frame
number 208.

The "\" means a Cut frame.

So, frame 209 was cut from the *new* original Telecine source, PPPii pattern,
and probably went something like this, originaly:

----------1---------2---------2---------2---------2---------2
----------9---------0---------1---------2---------3---------4
0123456789012345678901234567890123456789012345678901234567890
Telecine -- [203] < .. PPPiiPPPiiPPPiiPPP .. > [220]

but after "editing stage", is now something like this, when they
merge unmatched pieces together:

----------1---------2---------2---------2---------2---------2
----------9---------0--------\1---------2---------3---------4
0123456789012345678901234567890123456789012345678901234567890
Film Edit - [203] < .. PPPiiPiPPiiPPPiiPPP .. > [220]

And, what you have, is the edited version, which is also the Commercial
version, on DVD (or whatever it is that you have, source'wise) today.

It is still a commercial (or, professional) version, but just an edited
version, and not to be known about it, because you are just a viewer in
possession of it, and are only suppose to be watching on tv, not diesecting
it and uncovering leasions or abnormalities, etc.

However, we here, are just opening up your eyes to these kinds of things
that are not tipically known everyday by the casual viewer, who at this
time, is trying to copy/transfer such video to another format, MPEG, and
is actually learning why the process is not so smooth-going.

And now, you are a day smarter than yesterday, hehe

-vhelp 4016