help fixing 29.97fps video

[maudit]

Video: MPEG2 Video 720x480 (4:3) 29.97fps [V: mpeg2 main, yuv420p, 720x480]
Audio: Dolby AC3 48000Hz stereo 192kbps [A: ac3, 48000 Hz, stereo, 192 kb/s]
Subtitle [S: pgs]
Subtitle [S: Forced Subtitles (auto)]
Subtitle: UTF-8 [S: No subtitles]

this one: https://www.dropbox.com/s/seyvfi5c1luea37/00005.m2ts?dl=0

I don' know what command to give in avspmod in order to remove that kind of interlaced stuff.
I tried with "QTGMC(Preset="Slow").SelectEven()" but it does not work very well.

also i dont understand why when I play the original m2ts, it plays at 640x480.
I think I have to use the "NTSC 4:3 720 x 480 8:9 ITU OFF" SAR, right? so the res become 640. correct?

please help me to write a correct avs script.

[jagabo]

Use: QTGMC().SRestore() to get back to the original 25p.

[maudit]

Use: QTGMC().SRestore() to get back to the original 25p.

original 25? I can't get it, the video must to remain at 29.97 to match the audio.

[jagabo]

The sequence I gave will not change the running time, you can keep the original audio.

And why is your interlaced video encoded progressive? Is that really indicative of what you're starting with? How did you make that sample?

[maudit]

effectively that command QTGMC().SRestore() fixes the video and change it to 25p while the lenght remains the same,
so I can keep the original audio.
thanks a lot but just to understand, what that command did exactly to each frame?

how can it change the fps but keep the lenght?
I'm quite confused.

[jagabo]

QTGMC converts each field of the source to a frame. So your 29.97 fps video (really 59.94 half-pictures per second) to 59.94 fps. SRestore() removes the duplicate frames, preferentially the ones with blending. In the end you are left with 25 fps film frames with no blending, no duplicates, no drops, and no change in running time.

How can you change the frame rate without changing the running time? By changing the number of frames. Say you have deck of 52 playing cards. Flip them over at a rate of one card every second. How long does it take to flip them all over? 52 seconds. Now throw out half the cards. You have only 26 left. Flip those 26 card over at a rate of one card every two seconds. How long does it take to flip them all over? 52 seconds again! The NTSC video has duplicate frames/fields. The filtering gets rid of those duplicates.

[maudit]

QTGMC converts each field of the source to a frame. So your 29.97 fps video (really 59.94 half-pictures per second) to 59.94 fps. SRestore() removes the duplicate frames, preferentially the ones with blending. In the end you are left with 25 fps film frames with no blending, no duplicates, and no drops.

thanks!

[maudit]

excuse me if I up this thread, I have another video at 29.97fps, but in this case the command

QTGMC().SRestore()

only partially works, I mean that some frames remain doubled.

@jagabo, can you please give me another hint about what to write in the script?

here's a sample: https://www.dropbox.com/s/n4wegk4wgp93lut/sample_%20-%201%20-%20MPEG2%2C%20480i60_edit.mkv?dl=0

[Selur]

@maudit: Why 'QTGMC().SRestore()' and not normal IVTC handling (like 'TFM().TDecimate()') ?

[jagabo]

Yes, that video is simple telecined film. Use TFM().TDecimate().

[maudit]

@Selur

hi, and thanks for that hint... in facts the command

TFM().TDecimate()

it seems to works very well in that scene taken from the movie,
but as soon as the interviews begin, I start seeing interlaced frames.
..this is quite complicated for me, I've never handled such a file.

here's a longer sample including interviews parts:
https://www.dropbox.com/s/h89g45ot09c9d2x/longer%20sample_%20-%201%20-%20MPEG2%2C%2048..._edit.mkv?dl=0

ps: the original file is too big to post it, so these samples are made this way:
- I've extracted the m2v raw video file from the m2ts
- so I've incapsulated the m2v into an mkv
- then I've cutted the mkv in avidemux at two I-frames, simply copying it.

I hope I did not have changed the video's original properties.

[Selur]

Okay, the interview part is interlaced.
-> you got mixed content (partially telecined, partially interlaced).
So how I see it you can either:

• go for 23.976 fps then the movie scenes will be fine and the the video will miss frames ('TFM().TDecimate().AssumeFrameBased().QTGMC(Input Type=2)' or QTGMC().sRestore(frate=23.976))
• go for 29.976 fps and live with the duplicates in the movie scenes (QTGMC().SelectEven() or QTGMC().sRestore(frate=29.976))
• go for 59.94 fps and live with the duplicates in the movie scenes (QTGMC())
• apply ivtc to some parts and deinterlacing to other parts and create VFR content
• stick with interlacing and don't deinterlace

None of the solutions are really nice, but that is what I see as possiblities.
(personally I would use 'TFM().TDecimate().AssumeFrameBased().QTGMC(Input Type=2)' or stick with the interlacing if that is an option)

Cu Selur

[jagabo]

The easiest of those is to simply use QTGMC and encode at 59.94 fps. As long as your playback device(s) support the frame rate (some older devices didn't). If QTGMC is too slow for you try Yadif(mode=1) instead. Its deinterlacing isn't as good but it's much faster. When watched on a 60 Hz display it will have the same 3:2 judder that you would have seed on a CRT during the film parts, smooth video during the interviews, but without the flicker of a CRT.

[maudit]

hi, thank you both for the help, very appreciated.
I'm trying all those commands you pasted, just to see the differences.

I'm still deciding but I think I will follow the option:
- go for 59.94 fps and live with the duplicates in the movie scenes QTGMC()

Every commands above work, except this:
TFM().TDecimate().AssumeFrameBased().QTGMC(Input Type=2)
When I try it, I receive the error Script error: expected a , or )
but I don't know how to fix it.

[poisondeathray]

Every commands above work, except this:
TFM().TDecimate().AssumeFrameBased().QTGMC(Input Type=2)
When I try it, I receive the error Script error: expected a , or )
but I don't know how to fix it.

there is no space for InputType; it should be QTGMC(InputType=2)

[maudit]

there is no space for InputType; it should be QTGMC(InputType=2)

ah! ok I got it, it works, thanks.

[jagabo]

The interviews will be a little jerky at 24 fps but it's barely noticeable in talking head shots. The film portions may be smoother depending on the TV/player. For example a 120 Hz TV may display it with each frame for 5/120 seconds eliminating the judder you get with 60 Hz displays.

[SF01]

First of all, NEVER convert 720x480 , only the middle 704 samples contain the 4:3, or 16:9 active video. It's all written in ITU-R BT.470 and 601.
Secondly, to fix 29,970 the rest of the world would have to finally adopt the civilised 25 fps.

[Cornucopia]

Never, except the millions of 720x480 files that follow the full raster MPEG spec, or DV spec.

29.97 is not so complicated that a bit of simple math can't handle, made easier by using something known as computers. Not less, nor more, civilized than 25. And since so many sources are film-based, there is that "uncivilized" 24fps to contend with, too. No, we are going to have this triumvirate of framerates as options for a long time. Unless you want to join my bandwagon that combines them all into a single 600Fps rate?


Scott

[SF01]

Never, except the millions of 720x480 files that follow the full raster MPEG spec, or DV spec.

29.97 is not so complicated that a bit of simple math can't handle, made easier by using something known as computers. Not less, nor more, civilized than 25. And since so many sources are film-based, there is that "uncivilized" 24fps to contend with, too. No, we are going to have this triumvirate of framerates as options for a long time. Unless you want to join my bandwagon that combines them all into a single 600Fps rate?


Scott

DV spec follows the BT.601, just because it doesn't have analog blanking, doesn't mean it records 4:3 image on all 720 pixels, quite the contrary, it just has more image, than 4:3, and it's perfectly compatible with DVD, where you burn the video and it's still properly shown on TV as per BT.601. A simple test of shooting a circle and comparing the images is sufficient and I had this discussion before, either here, or digitalfaq and indeed the end verdict was that DV is in fact compliant with BT.601, especially that it is being used to digitize analog videos which are BT.470 and 601 is backwards compatible with 601.
As for 24, in order to convert them to 29,970, it needs to be slowed down to 23,976 first.
As for math, one frame in 25fps video is exactly 40 ms, how much is it in 29,970, or 23,976? A non-natural number. Same with audio samples per frame.
24 is OK, nothing against it, but fractional framerates are a mess. Created because Not the Smartest Choice couldn't handle colour at full 30 fps. The moment PAL was created it should have been adopted worldwide, as it was in majority of countries.

MPEG on the other hand is a file format, you can encode it however you want, be it 2:1 sampel aspect ratio, or any other funky number you can think of.
In fact, respectable NLEs have proper SAR values for SD videos, examples being Vegas, PRemiere, and of recent Resolve.