[SOLVED] QTGMC - Weird ghost artifacts

[AOA]

Hello.
I am trying to deinterlace Haruhi's DVD.
In the past I've already used QTGMC and didn't really have any problems with it but with this source I suddenly get some weird artifacts that seem to stem from lines in the past frame. I'm not fully experienced with interlaced material but also couldn't find other posts about a similar problem.



clip = core.lsmas.LWLibavSource("VTS_01_1.VOB")
qtgmc = hav.QTGMC(Input=clip, Preset='Slower', EZDenoise=1, NoisePreset="Slow", Sharpness=1, TFF=True, FPSDivisor=2, InputType=0,
NoiseProcess=2, GrainRestore=0.4, NoiseRestore=0.2, Sigma=1.8, NoiseDeint="Generate", StabilizeNoise=True)
rescale = insaneAA.rescale(qtgmc, dx=854, dy=480)
rescale.set_output()


Example output: https://slow.pics/c/UnhBOGf1

Is it possible to get rid of these "ghostings" or am I out of luck?

[Selur]

Source probably isn't interlaced, but a mix of content. From just a single frame, no idea whether those artifacts can be avoided.

[AOA]

Source probably isn't interlaced, but a mix of content. From just a single frame, no idea whether those artifacts can be avoided.

Hm ok. Should I upload more screenshots, or should I upload a small clip of that scene to mega, ...?
Mixed content sounds very complex.
Do you have any tips for me, or should I just give up?

Edit: Waiting for moderator approval.

[kedautinh12]

Upload small clip will easy to findout solution

Edit: if you update to mega will faster than Waiting for moderator approval :D

[AOA]

Upload small clip will easy to findout solution

Edit: if you update to mega will faster than Waiting for moderator approval :D

Ok thanks for the tip.
Are 200 frames enough?
https://mega.nz/file/JRd3GZCB#cVK6qP8HA7Gw81q6nV2SugXoMUVbG8JoGUpMBtJB-qA

This scene is much shorter with only 50 frames but the artifacts are a lot more noticeable:
https://mega.nz/file/BFFnUQAT#JPifgnoe31Ftyre8lcg-jZQ_RI68qDu8XxfEPFZBSq4

Should I add the unprocessed clips?

[Reel.Deel]

Should I add the unprocessed clips?

Yes, it's always best to supply the original clip.

[AOA]

Yes, it's always best to supply the original clip.

Oh.
Here is the src for the second clip:
https://mega.nz/file/dUEx0bJb#CcyaMbMmPkCpH3La24zBZcSGiryDYCSQ41-zuZaB__8
Greetings from the inability to cut a video without encoding. Sorry

[mastrboy]

Oh.
Here is the src for the second clip:
https://mega.nz/file/dUEx0bJb#CcyaMbMmPkCpH3La24zBZcSGiryDYCSQ41-zuZaB__8
Greetings from the inability to cut a video without encoding. Sorry

You need IVTC, not deinterlacing for this source. (It's 3:2 pulldown, aka telecined)
Though the clip is too short to determine if its CFR or VFR, TFM/Tdecimate can deal with both scenarios: http://avisynth.nl/index.php/TIVTC

(You can use QTGMC when fieldmatching fails in TFM with the clip2 parameter, but most sources don't need it.)

[AOA]

You need IVTC, not deinterlacing for this source. (It's 3:2 pulldown, aka telecined)
Though the clip is too short to determine if its CFR or VFR, TFM/Tdecimate can deal with both scenarios: http://avisynth.nl/index.php/TIVTC

(You can use QTGMC when fieldmatching fails in TFM with the clip2 parameter, but most sources don't need it.)

Thank you so much. Found some great stuff in core.vivtc, which seems to be the solution.
No more artifacts and something new I can learn, YAY!!!
THANK YOU!!