Can't understand this frame rate

[ZetaStax]

Hello everyone. Could you help me figure out the original frame rate of this clip please? https://mega.nz/file/Hx5FVYwC#trnidgXi_paWhZEIBag8P21Z3K77_1NcyH3o0MRjoBI

The pattern is 1 progressive for 2 interlaced, from a 25i source. So how to work with something like that, is it supposed to be ivtc to 12.5 fps? or partly deinterlaced to 16.67 fps? I can't make head or tails of this format. I'm not even sure how to apply such exotic ivtc/deinterlacing...

[poisondeathray]

It's 25p "worth" of content, but there are many orphaned fields .

You can either A) double rate deinterlace and decimate, or B) field match even/odd, interleave then decimate. The difference is more original frames are kept with the 2nd method. Eitherway, you need to deinterlace a signifcant number due to orphan fields/combing . Something like QTGMC will smooth over , instead of deinterlacing every ~2nd frame leaving a type of flicker

For some reason, specifying cycler, cycle leaves duplicates ,even with a larger window (50,100) - but mode 7 works ok. You might be able to tweak some of the other settings to get other modes to work, I dont' have time to fiddle with it

A - this is the one I would use, you can tweak the qtgmc settings to retain more detail, but the "denoising" helps in this source. But this one has no duplicates

orig=MPEG2Source("2020-08-12.d2v")

orig
#yadif(1,1)
qtgmc(preset="faster", sharpness=0.5)
tdecimate(mode=7, rate=25)
#tdecimate(cycler=25, cycle=50) #leaves duplicates
#tdecimate(cycler=50, cycle=100) #leaves duplicates
prefetch(6)

B, interleave method, leaves more original frames matched, but leaves a few duplicates regardless of decimation

orig=MPEG2Source("2020-08-12.d2v")

deinta = orig.assumetff().qtgmc(preset="faster", sharpness=0.5).selecteven()
deintb = orig.assumetff().qtgmc(preset="faster", sharpness=0.5).selectodd()

a=orig.assumetff().tfm(clip2=deinta, display=true)
b=orig.assumebff().tfm(clip2=deintb, display=true)

interleave(a,b)
#tdecimate(cycler=25, cycle=50) #leaves duplicates
#tdecimate(cycler=50, cycle=100) #leaves duplicates
tdecimate(mode=7, rate=25) #leaves duplicates
prefetch(6)

c) "Regular" TFM(display=true) leaves a few duplicates (around scene changes)

[ZetaStax]

Hmm. I'm not sure... First method indeed leaves no duplicates but motion is jerky (I myself was using QTGMC with FPSDivisor=2 which achieves the same result). This is why I was under the impression that restoring all 3 frames would be better. But I would need to select half the original frames and follow each frame with 2 frames from the deinterlaced source.

On a side note you talk about denoising but you only enable sharpness, any reason for that? I myself was thinking of using EZDenoise=2.0 for this source.

[poisondeathray]

Hmm. I'm not sure... First method indeed leaves no duplicates but motion is jerky (I myself was using QTGMC with FPSDivisor=2 which achieves the same result). This is why I was under the impression that restoring all 3 frames would be better. But I would need to select half the original frames and follow each frame with 2 frames from the deinterlaced source.

Restore from what ? Did you mean interpolate extra frames ? You only have 25p worth of information if you check with separatefields

On a side note you talk about denoising but you only enable sharpness, any reason for that? I myself was thinking of using EZDenoise=2.0 for this source.

I just left it at default, which denoises too. I think QTGMC's default sharpness is too high, I almost always turn it down. If I denoise I usually use a dedicated denoiser

[ZetaStax]

Oh yeah I must've made a mistake. It is 25 pure frames, but the motion is jerky as I said. It feels like it's missing one full progressive frame, as if the film was shot 30i.

[poisondeathray]

There could be drops, but the pattern is not clear in this clip. This clip has intentional moments of acceleration , deceleration (you can tell from the variation in the magnitude of motion blur in frames).

You can look at other clips from the same source , ones with steady motion and see if you can figure out the pattern and then try to apply it here (but it might not be applicable to this section, because this was shot and edited in a specific way). If you know where to insert frames you could try to add frames and interpolate, but that would only be smooth if they are in the right spots, otherwise you introduce some jerkiness. Also I doubt this type of material will interpolate cleanly due to the motion characteristics, and motion blurred frames rarely interpolate cleanly