qtgmc() + srestore() not working on animation

[LaserBones]

The second season of a show I'm trying to encode doesn't fully work with qtgmc() + srestore() About 90% of the frames work but there is some color smear as well as some blending on some of the frames. How can I resolve this?

[hello_hello]

I've only looked at the first few percent of the video as it's only downloading at about 8 KB/s, but from the bit I've looked at it appears to have been blended from 24fps (or maybe 23.976). There's also some color smear and a ghosting effect baked into the fields near the beginning, but if QTGMC is making it worse, maybe try some other bob de-interlacer.


[Attachment 87640 - Click to enlarge]

I created the attached sample using the FixBlend function in my signature. It works great if the blend pattern doesn't change, but it will fail when it does. It's not automatic like SRestore. But anyway.... maybe try this (I remuxed the sample as a TS file so I could open it with DGIndex):

Mpeg2Source("clip.d2v")

TDeint(Mode=1, TryWeave=true)
# or QTGMC() or something else

Trim(0,699).FixBlendX(8, 10, 24000,1000) + \
Trim(700,0).FixBlendX(4, 10, 24000,1000)

# or SRestore(24) or SRestore(24000.0/1001.0)

The blending pattern did seem to change in the small part I looked at, and usually SRestore will leave a few blends behind until it locks into the pattern again. If it changes often though, it's a lot of work to fix it manually with FixBlend.

[Selur]

Aside from the Nickelodeon intro, the source seems to be 14000/1001 not 24000/1001.
So, maybe: QTGMC+CeDeblend+SelectEvery(cycle=4, offsets=[0]) might also be an option.
Here's a version where I also added some machine learning based filtering,..

[hello_hello]

I wondered about a lower frame rate, but there's a section roughly 34 seconds in with quite a few unique successive frames at 24fps.