I'm not quite sure to explain my issue fully, but I have a music video, in 29.97, if I IVTC it, its nearly perfect, however, it has issues on crossfades, where it cant decimate the dupes properly because the dupes are in different places on each layer of the crossfade.
See the de-telecined snippet attached below. Go frame by frame, and you'll see what I'm talking about.
I also included the untouched m2v for you to experiment with.
+ Reply to Thread
Results 1 to 3 of 3
I would use manual overrides and/or replacement frames for that section (or if you want better results, combine some areas with masks) .
Replacement frames can come from deinterlaced versions such as QTMGC, yadifmod2/nnedi3 . Some areas you might not want to use QTGMC because it can blend more.
You can mix/match frame versions using remapframes in avisynth or a NLE (or parts of frames using masks/roto)
If you use TFM(pp=0).TDecimate() - that is your "base" decimation. You can see the decimation is correct, but there are combed frames (because pp=0) . So use that as your timing guide (baseclip in remapframes, or base layer in a NLE), and repair the combed areas
(and some other issues like other minor combing and rainbows you might want to address)
eg. using TFM(pp=0).TDecimate() as the "base" and QTMGC as the sourceclip for replacement frames
MPEG2Source("snippet.d2v") tfm(pp=0) tdecimate() t=last MPEG2Source("snippet.d2v") AssumeTFF() QTGMC(sharpness=0.5) q=last remf(t, mappings=" 4 10 6 15 7 17 8 20 10 25 12 30 14 35 16 40 18 45 19 47 20 50 22 55 23 57 24 59 ", sourceClip=q) trim(2,0)
Timing wise it's correct and some people might just leave it like that. Personally, I would go a step farther and add back details /blurring lost by QTGMC , or combine different deinterlacers. For example, that hat "shady" text is slightly blurred on some of the replacement frames. You can mask it out and bring back the original version with some minor roto work