Posts: 785
Threads: 16
Joined: Mar 2020
Yep, probably make sense to send a report to developer.
I also think it is more like different logic in different filters:
vivtc arrange frames somehow in it's own way and as result produce interlaced (dropped) frame sometimes.
qtgmc combines and deinterlaces everything that it feels may be deinterlaced and as result instead of single interlaced (dropped) frame it produce some multiple combined frames with slightly jagged lines.
As i told before - i don't know too much about detelecine and it's variations and settings. it is all total guess and plain side by side tests.
Well without a doubt if i can reverse telecine a telecined material.... the quality is much, much better.
I had a basic understanding before, but seeing those animation stills opened my eyes a bit.
I dont think the jaggies on the qtgmc image is a hybrid problem, but a known issue with qtgmc. I remember seeing someone post pictures of diagonal ropes on a boat being distorted. Again would love to know if the author of qtgmc is actively trying to fix this, or if qtgmc hasnt been updated in years and this is the best it will ever be.
When using qtgmc as a deinterlacer it has a ‘lossless‘ mode and a ‘match to source’. I don’t know if those options are available or needed when used under the vivtc filter. Also i don’t know what is the slowest speed setting for qtgmc under the vivtc filter.
I tend to do the placebo settings for all filters despite any real benefit over very slow. The jaggies themselves maybe unavoidable and are part of how the math in the algorithm work. I know the vivtc has lower threshold settings and ability to make the detecting box for combing smaller and smaller. I know for the x/y i couldn’t go below 16/16 on the old hybrid 2018. without it crashing. So i may have to play with those settings once i finally update to hybrid 2020.
I will be running a few animations thru topaz after i reverse telecine. This guy i’ve been talking to who only works on avisynth was the one who suggested needi3. So if they both produce the same jaggies and qtgmc is better, then it ismprobably what i will use. i will check them both out obviously, and likely choose qtgmc.
Posts: 785
Threads: 16
Joined: Mar 2020
needi3 produce more visible and thick jaggies than qtgmc very slow.
but needi3 is about 5x faster than qtgmc very slow
Also note that animation example i provide is rerally fast and scenes change in every frame very randomly See here
https://youtu.be/3pT7gadYPso?t=1261
Posts: 785
Threads: 16
Joined: Mar 2020
Yep. From that thread sounds like QTGMC need special adjustment for VIVTC.
Problem: the source is a clip with mixed progressive/interlaced frames.
There are purely progressive frames, purely interlaced frames and frames contain both p and i area, like this example.
Problem fix: The jaggies are not caused by QTGMC itself, they are caused by TDeint.
Simply using your QTGMC call alone, without TDeint, results in a jaggie-less video.
The problem is you are wrapping QTGMC in TDeint. I suspect it is caused by the fact that by default QTGMC does not leave the original fields unaltered. TDeint does, so the interpolated fields by QTGMC do not match the original ones that TDeint injects, resulting in jaggies.
---
I tested normal QTGMC (non VIVTC) with same source and it produce same jaggies effect. So seems both use same script source.
Posts: 10.976
Threads: 56
Joined: May 2017
jagged lines, can happen with same rate deinterlacing -> use AA
Posts: 785
Threads: 16
Joined: Mar 2020
Those are totally different sort of jagged lines. They impossible to filter with AntiAliasig filters
Posts: 10.976
Threads: 56
Joined: May 2017
03.09.2020, 05:29
(This post was last modified: 03.09.2020, 05:37 by Selur.)
If it's mixed content I like to QTGMC +sRestore the content,.. but there is always that content that you can't simply filter properly by automatic methods,..
And there a tools like
https://github.com/dubhater/Wobbly which can be used to filter a video per scene, and/or to improve upon VFM and VDecimate's decisions, but no clue if they could be used on MacOS,...
+ there's always the possibility to write you own script
Posts: 785
Threads: 16
Joined: Mar 2020
QTGMC + sRestore also produce jagged lines
