Tricky deinterlacing

[Mounir]

Well guys i have a source that's been a challenge for me lately.
Problem is whenever i deinterlace it somehow the length ends up too short (the video part) in the Sony Vegas timeline and the A/V sync is impossible. I tried with qtgmc and tfm and keeping it 29.9 fps (using dgindex).I'm not asking the moon just have the A/V in sync

https://mega.co.nz/#!B4YWAQDR!poPReiEj9CEJEfLZJe5JhqOesPMCyOLwX43jzf_rynE

Thanks for your help

[johnmeyer]

Well guys i have a source that's been a challenge for me lately.
Problem is whenever i deinterlace it somehow the length ends up too short (the video part) in the Sony Vegas timeline and the A/V sync is impossible. I tried with qtgmc and tfm and keeping it 29.9 fps (using dgindex).I'm not asking the moon just have the A/V in sync

https://mega.co.nz/#!B4YWAQDR!poPReiEj9CEJEfLZJe5JhqOesPMCyOLwX43jzf_rynE

Thanks for your helpDeinterlacing is the wrong thing to do. This is, more or less, telecined film with (again, more or less) a standard 3:2 pulldown pattern.

So, QTGMC is the wrong tool.

TFM/TDecimate are the tools you want to use. However, the source has a few fields that don't have matches. You can see this if you run the source through a script that is nothing more that "separatefields()". With normal film material that has been telecined to 29.97, you will get two matching fields followed by three matching fields ("matching" in the sense that they are from the same moment in time). The default TFM/TDecimate settings will throw out that one extra field, leaving you with 23.97 progressive film.

That is what you want to get.

However, those occasional orphaned fields that come from a unique moment in time and which do not have a corresponding top or bottom field, must be detected and then bobbed to fill in the missing field. I don't have time to construct that script, and don't know if it can all be done with nothing more than TFM/TDecimate.

[Mounir]

I look forward to your script Johnmeyer !

[thecoreyburton]

DGDecode_MPEG2Source("video.d2v")
IVTC=Last.QTGMC(FPSDivisor=2)
TFM(PP=2, clip2=IVTC.SelectEven)
TDecimate()

It doesn't bob the source or account for VFR, but if it's mostly a film source with a few orphaned fields as johnmeyer previously said that should do what you're after.

[Mounir]

It return " invalid arguments to function Qtgmc", any idea ?

[thecoreyburton]

Sorry, it was a typo on my behalf. I edited the post, that should work.

I usually chuck in DAA at the end if I don't want to manually check for any residual combing or aliasing. It's probably bad practice, but if you prefer the results here's what you'd use:

DGDecode_MPEG2Source("video.d2v")
IVTC=Last.QTGMC(FPSDivisor=2)
TFM(PP=2, clip2=IVTC)
TDecimate()
Daa()

[Mounir]

Now it return: "TFMPP: clip2 does not have the same number of frames as input clip"

[ChiDragon]

Remove either FPSDivisor=2 or SelectEven.

[Mounir]

I did a quick encode i still don't have good A/V sync (audio way off the mark) and the audio track it's longer of 10sec 07 at the end

[Izuchi]

Use delay audio when muxing or realign the audio using an editor? I don't think the fps is the problem here given the quality of the sample you posted.

[Mounir]

i'm affraid a simple delay audio isn't the solution with this source unfortunately.I may "realign" and sync the audio at point A but later at point B it'll not be in sync anymore

[Mounir]

Is there a solution for the video/audio sync guys ?

[thecoreyburton]

Sounds to me like either the audio or video are somehow being played at the wrong rate.

When you say it goes out of time, does the audio end up faster or slower? If you see the action and then hear the sound, the audio's faster. If you hear the sound and then later see the action that caused the sound, the audio's slower.

Additionally, does the audio sound higher / lower pitched in the output than the original, or is it the same?

And what is the input video's frame rate and the output video's frame rate? Post both, if you can.

[Mounir]

As far as i can tell the pitch it's the same to me
The audio is late at the beginning and at the end even more late