Poor rip issues

[levick]

Morning all,

I have been given a series of 10 minute short movies that have had the flag references destroyed.
My request is to retrieve the 23.976 fps again so I can convert to PAL 25 fps.

Now it gets interesting the films have been authored after conversion to 29.7 fps so I can not lock to the 2:3 pull down sequence for the whole clip. So I am hoping to use a script for avisynth. I need to say use the first 200 frames do the pull down in a certain order then the next 200 frames the pulldown needs to be altered again and so on etc etc so with every cut to a different scene I will need to alter the pull down configuration.

Can any one assist with a way to look at the first sequence of frames apply the pull down correctly and look at the next scene reapply another pull down sequence giving me the magic 23.976 pure progressive stream..

If I have not made my request clear please ask..I would like to "rip" into this project so to speak.

so it could look like (omitting the separating of frames and weave instructions).


check frames 0,200 selectevery (10, 0,1, 2,3, 4,5, 8,9)
check frames 201,500 selectevery (10, 0,1, 4,5, 6,7, 8,9)

and so on until done.

Thanks in advance...

[manono]

Hi-

Now it gets interesting the films have been authored after conversion to 29.7 fps so I can not lock to the 2:3 pull down sequence for the whole clip.

You seem to be saying that it's been hard telecined from progressive 23.976fps to interlaced 29.97fps. If so, then make the D2V for "Honor Pulldown Flags" and IVTC it. Nothing difficult about that. You sure as heck don't want to be messing with hundreds of "SelectEverys". Those SelectEverys you included seem to show it's been telecined. No big deal. Have you tried to IVTC it using AviSynth's Decomb or TIVTC?

If this answer doesn't help at all, then how about a small 10 second sample from the source uploaded somewhere? Use the [ and ] buttons in DGIndex to isolate a small section with motion and upload it for us to get.

[jagabo]

Along the lines you were thinking:


source = MPEG2Source("video.d2v")

v1 = Trim(source, 0, 200).SeparateFields().SelectEvery(10, 0,1, 2,3, 4,5, 8,9).Weave()
v2 = Trim(source, 201, 500).SeparateFields().SelectEvery(10, 0,1, 4,5, 6,7, 8,9).Weave()

Return(v1 + v2)

[levick]

[quote="manono"]Hi-

Now it gets interesting the films have been authored after conversion to 29.7 fps so I can not lock to the 2:3 pull down sequence for the whole clip.

You seem to be saying that it's been hard telecined from progressive 23.976fps to interlaced 29.97fps. If so, then make the D2V for "Honor Pulldown Flags" and IVTC it. Nothing difficult about that. You sure as heck don't want to be messing with hundreds of "SelectEverys". Those SelectEverys you included seem to show it's been telecined. No big deal. Have you tried to IVTC it using AviSynth's Decomb or TIVTC?

This is the problem I have, the flags are no longer present. However the repeated fields are hard coded into the mpeg stream...otherwise this would be a easy exercise as you mention.

[levick]

Along the lines you were thinking:


source = MPEG2Source("video.d2v")

v1 = Trim(source, 0, 200).SeparateFields().SelectEvery(10, 0,1, 2,3, 4,5, 8,9).Weave()
v2 = Trim(source, 201, 500).SeparateFields().SelectEvery(10, 0,1, 4,5, 6,7, 8,9).Weave()

Return(v1 + v2)

Now I believe this maybe the information I am looking for...thankyou Jagabo...

[manono]

This is the problem I have, the flags are no longer present. However the repeated fields are hard coded into the mpeg stream...otherwise this would be a easy exercise as you mention.

It still just sounds like an easy IVTC to me. Part of the problem in figuring out from your description just what you have is that much of what you've written so far is gibberish and makes no sense. Anyway, if, according to your first post, the cadence changes every 200 frames, then that's 30 SelectEvery sequences per 10 minute video, and a silly waste of time to try and figure out when:

LoadPlugin("C:\Path\To\DGDecode.dll")
LoadPlugin("C:\Path\To\TIVTC.dll")
MPEG2Source("C:\Path\To\Video.d2v")
TFM()
TDecimate()

can do the job better and much more easily. I ask again, have you even tried to IVTC it yet? Do you know what an IVTC is? You're in PAL land and can be forgiven if you don't know. Also, if you insist on doing it your way (why?), rather than trying to figure out all the SelectEvery combinations manually, use the built in ShowFiveVersions filter together with the Pulldown filter:

http://avisynth.org/mediawiki/ShowFiveVersions
http://avisynth.org/mediawiki/Pulldown

[levick]

[quote="manono"]

Part of the problem in figuring out from your description just what you have is that much of what you've written so far is gibberish and makes no sense.

Sorry, you see it as gibberish that was not my intention, Yes I understand what IVTC is, some one has badly butchered these clips and I would like to achieve a perfect reproduction of the orginal filmed material.

The frame counts were examples only, I have seven 10 minute clips to re do, about half will accept the normal
IVTC conditions perfectly and the other are so bad its just about not worth trying but I wish it give it one last attempt.

For example on a high motion scene change (car racing) the repeat fields apply in the first frame set...
so its noisy blured frames to start with and most filters I have tried are missing the extra fields....
Next scene change the repeated fields return to field pair 4,5..


I do thank everyone for posting and giving me some great ideas, I will now go and exercise them

Cheers.

[manono]

IVTCs can be tweaked to handle out-of-the-ordinary video. In addition, if you find frames that don't IVTC properly no matter what you do, override files can be created to force those frames to do it the way you want. Both Decomb and TIVTC have the ability to create override files. Also, YATTA is designed to handle just these kinds of problems. It was originally designed for anime and it doesn't get any worse than that.

http://ivtc.org/

If you have a sample showing the kinds of problems you're encountering, we'd be happy to take a look.