Tricky NTSC source, field blending, slow motion action

[mwesten]

I'm working with an NTSC DVD source that's somewhat unconventional.
There's some manner of field blending as well as moments of slow-motion sequences interspersed throughout.
Wondering if anyone with a keen eye can help make sense of this.

The best simple solution I can think of is to deinterlace with QTGMC, leaving the output at 29.976fps with no further alterations.
Is there perhaps something better I could be doing with this?
Attached are before and after clips, the source as well as the QTGMC results (in Selur's Hybrid).

[davexnet]

qtgmc and srestore to detect the blends gives 25 fps:

qtgmc(preset="slow")
srestore()

[mwesten]

qtgmc and srestore to detect the blends gives 25 fps:

qtgmc(preset="slow")
srestore()

That had crossed my mind. I'll try "restoring" it to 25fps with sRestore and see how it looks.

[Selur]

QTGMC+sRestore(23.976) seems to be a good solution.
Depending on your preferences, you could also add try adding de-blurring through BasicVSR++.

[mwesten]

QTGMC+sRestore(23.976) seems to be a good solution.
Depending on your preferences, you could also add try adding de-blurring through BasicVSR++.

>23.976
I think this drops too many frames though? At least on the sample I ran there appeared to be some missing frames.
davexnet says 25fps is true framerate, which may be right, but I don't know how to recreate his method of discovering this.
The instructions are too vague for me since I'm a novice with no mastery over the software or the synth scripting.

[Selur]

Look at the frames after applying QTGMC and count the unique or duplicate frames.

[jagabo]

23.976 looks right to me. 25 fps leaves a duplicate about every 25 frames.

[manono]

Yes, I'm with Selur and jagabo. 23.976fps is correct.

[jagabo]

Note that the clip has several short shots (the woman's spin kicks) that were slowed down with duplicate frames. Those are intended to be there.

[mwesten]

There are a lot visual effects like this all throughout

[Selur]

Depending on what your target and your preferences are, you could bob deinterlace and replace duplicates with interpolations and end up with 59.94fps output,...

[mwesten]

It appears there's several correct ways to approach this.
I'll try each of these and inspect the results.
Thank you Selur, jagabo, manono and davexnet