Hi all, newbie in video enhance, im curious about the best way to deinterlace, for example dvd 720x480 generally bottom field first…better flowframes, hybrid (very slow or placebo) and after put in topaz or make all the deinterlace in topaz?
And if double frame rate (from 29.97 to 59.94) sometimes there are glitch…but seem the smoother in unifab make a better job…
Depends on the source. If it’s a film that was interlaced, you need to drop it to 23.976fps or 24fps after deinterlacing it. There are a great many other types of sources that will need different considerations.
Best deinterlacer is QTGMC (probably Hybrid uses QTGMC).
have a look at this
VDub is just for visual check it is topfield or bottom field interlaced. After QTGMC deinterlacing in Hybrid check the 720x480 when you have this size, should it really displayed as aspect ratio 1.5? Maybe it was made for 16:9 or 4:3
When it should be played as 16:9 resize 720x480 to 864x486, which is codec friendly gives no floating points, gives equal numbers and is exactly 16:9. When it should be 4:3 resize 720x480 to 720x540, then import this into TVAI
Anyone know a Hybrid (or simply QTGMC) version for MacOS?
Or at least for Win on ARM?
hybrid on mac was discontinued in 2021, but try the last mac version Hybrid_mac_2021.04.12.1.zip from here maybe QTGMC was already implemented
muksun with virtualdub keeps more details than qtgmc 
proteus v4 is powerful enough to fix muksun’s few artifact flaws
QTGMC will certainly preserve more details than Proteus4…
I think he means a Vdub plugin named “muksun”, I have found one with this name on a china website, but I have not tryed out, I probably won’t do it either. I visually check several deinterlacer and for me qtgmc is best of all (setting very slow/placebo) near perfect. I can’t see loss of detail so far, I’m very happy with it.
I guarantee you that qtgmc loses details even with the medium preset… sure it deinterlaces very well by removing the alliasing effect but if you set it to placeboo, the final temporal smoothing at 3 removes a lot of details on the movements, even if it is set to 1, it lacks details on the movements, so believe me, muksun is the best, even if some alliasing is still present, proteus v4 is powerful enough to correct the alliasing and keep the maximum details. if for you qtgmc in placebo is good, then keep it, but that means that you are not picky about preserving details in the movements 
ok thanks maybe I give it a chance, the things is, I can’t see problems QTGMC lowering sharpnes on movements, I have visually checked. Maybe it was my content, or the problem existed on ealier qtgmc versions.
I have noticed that for some of my videos, QTGMC oversharpens, and bwdif comes out the clear winner.
For others it’s the best. It really does depend on the video.
so maybe it depends on the videos, but I really think it’s because you’re not picky about preserving details, and you see more the quality that qtgmc provides rather than the flaws. because with placeboo quality, it removes the maximum of artifacts while removing alliasing etc, in fact it denoises the image, so necessarily on movements, it’s logical that it’s in this situation where you really lose details. I’ve been on my videos for 9 - 10 years, juggling between hybrid with qtgmc, virtualdub, vegas pro, and video ai. so without bragging, I understand how qtgmc works, and it’s not magic, it’s not ai, it doesn’t repair anything but it smooths the image to put it simply. I work on family videos, 8 mm digital 8 and hi8 digitized cassettes. so going from sd to hd without artifacts on this kind of videos, it’s obviously much more complicated than going from full hd to 4k, in full hd there are already more details, and i’ve never tried qtgmc with a full hd video, because my other videos are progressive, i could but i haven’t done it. each video has different optimal settings anyway. for example with a go pro video filmed in 5.3k at 30 fps, i didn’t spend a lot of time on it but i found that stabilization with gyroflow and exporting in 4k in prores followed by a pass with proteus gives very good results. it was an example but i’ve been focusing for years on my family videos which are 1000 times more important to me, and so the slightest detail that will disappear, i don’t like it haha
yes when I noticed this at first, I used ez keep grain to restore the details, but I also used the denoiser fft3d filter, which seemed better to me, except that knlmeans retains more details anyway but increasing this denoiser value to 2 or 3 or more also restores artifacts, so the best is to skip ezkeepgrain and ezdenoise which also smoothes too much… I think the good compromise is to leave the preset on fast or very fast and leave final temporal smoothing on 1 but no more. it’s enough to remove a lot of artifacts, on the other hand the video once improved with vai, will look slightly false anyway… from the moment it lacks details in the movements, I think it’s quite weird, but there you go it’s still sd upscale videos in hd in the end. the quality remains very good, and someone who has never seen the original videos will necessarily find that they are incredible, without artifacts, without noise etc… this person will never notice that there are missing details for a video filmed between 96 and 2002 (for my part), but I have worked so much on my videos that I notice it in fact, that’s my problem, I am perhaps too picky in fact and suddenly it really slows me down… almost 10 years that I tell my family “don’t worry, now I found good settings, I will be able to do them” but in fact… no… haha
QTGMC always outperforms bwdif. If QTGMC oversharpens, set “sharpness=0”.
The recipe is simple: deinterlace with QTGMC, denoise with Neat Video, NO UPSCALING
Yep, it’s set to that. Still too sharp, but again, only on some videos.
On my SD Neat Video videos, it doesn’t work well, too many artifacts and distorted noise. Especially in SD, even with a good profile at 100%, it removes too much detail, especially on movements. I asked Chat GPT, and it said that the Muksun and MSU deinterlacers are better than QTGMC for preserving details. MSU is better overall. Muksun preserves details even better in the dark parts. He advised me to try deinterlacing with MSU, and do a pass with QTGMC in progressive and then Proteus v4 in HD. It’s true that I’ve never tried QTGMC in progressive, so I’ll give it a try
Hmm interesting, what does QTGMC in progressiv then? NEAT does removes to much noise I also noticed…so I’m not a fan of denoising in general because it always lowering details, but sure sometimes you have to deal with it when it’s to noisy, when Topaz models does bad things with your noise video
ChatGPT is not a good source of information.
In QTGMC you can set different presets.
The preservation of details depends on the degree of denoising, which can be set in QTGMC (for example via a preset).
No one who uses the Proteus filter (which destroys details like nothing else in the world) should complain
