I haven't seen a lot of direct comparisons between different hardware deinterlacers and QTGMC. If anyone has links, I'd appreciate seeing some of those comparisons.
QTGMC or any software deinterlacer should be superior on paper in that it doesn't have to do it in real time and can look a few frames forward and backwards to determine the context of how to deinterlace and if certain things are more likely to be noise vs detail I would think.
Capturing interlaced lets you try a variety of software deinterlacers though including AI later which could improve down the road. Capturing as progressive kind of bakes in whatever the hardware deinterlacer did and you can't get back the starting material from it. The exception to that possibly being a line doubler like the retrotink 2x as long as it's captured at 59.94 fps.
That being said, if you plan to view this on something like a small cell phone screen or you plan to heavily compress it for final distribution and want to keep the processing time and number of steps down, I would speculate that there's less of a difference that's going to be visible. Where you're going to see bigger differences is if you're full screening it on a large TV or a computer monitor. Seems a lot of videos end up going to YouTube these days, and the compression it does can take away some of the QTGMC advantages.
Out of curiosity, which hardware deinterlacer were you considering using?
I've got a few devices that can deinterlace, but haven't actually compared them yet - but I will eventually post a comparison versus QTGMC vs Handbrake.
They include:
Teranex 2D
DVDO iScan HD+
AJA Ki Pro (can upscale as it captures)
AJA FS1
Snell and Wilcox Quasar Ph.c (phase/motion compensated upconverter) - requires SDI source
Snell and Wilcox CVR550
ADVC-G1
Retrotink 5x
Retrotink 2x
Probably some random Extron scalers as well
Cheap upscalers are apparently better (in some cases) than other usual (non-recommended) consumer conversion chains:
https://www.youtube.com/watch?v=9vyugJsLDpk
But again, that doesn't directly compare to recommended chains to see how much of a quality loss there is. Certainly would be a time and complexity savings if it was even close. Depends on your final format and the amount of quality/simplicity that makes sense to you. That being said, hardware upscalers in general aren't recommended on this site anyway including all of those that I have above. Doesn't mean I don't want to try it anyway to see what it looks like - for science!
