Hello, I am new here (TVAI user since November 2022), this is my first post. Please be lenient if it’s too “beginner” or not technical enough.
Generally, I am very glad I’ve started using TVAI and satisfied with the results I’m getting.
I am about to upscale a number of feature films ripped from DVDs (as .mkv files). Most of them are films from the 1990s and early 2000s, some may be from the 1980s or 1970s or even earlier decades. I will upscale them, cropping if necessary, from their native resolution to whatever is closest to FullHD (so, for example, 1920x816 if the film has the 2.35:1 pixel ratio, or 1440x1080 if it’s 4:3, and everything in-between), choosing the square pixels option.
The films I am referring too are interlaced, so I need to deinterlace them (I’ve tried upscaling them using Proteus without deinterlacing and the result looked horrible). My question is how to do it best.
I am using TVAI 3.15 on a system which allows me to process each film in no longer than 4-5 hours (a lot of them faster). Time is not an issue, I don’t mind waiting a couple of hours more, I’m mainly interested in the quality of the output.
I have so far used the Proteus model (for similar films which did not need deinterlacing), with results that I find satisfactory, so I would kind of like to stick to it. However, IIUC, I need to use the Dione model now to deinterlace the videos before upscaling (if I want to deinterlace within TVAI).
So if I understand the situation correctly (please correct me if I’m wrong!), I can go about it for example in one of the following ways:
Deinterlace the 720x480 video using Dione, either converting it to square pixels or not, and then, as a separate process, upscale it to FullHD (1920x816, or 1440x1080, or whatever is appropriate) using Proteus (converting it to square pixels if I didn’t do it in the first step).
Deinterlace, upscale and convert pixels to square ones using Dione.
Deinterlace, upscale and covert pixels to square ones using Dione and then enhance the upscaled video (without any more upscaling) using Proteus.
I’ve tried processing one of the films using methods 1 and 2 above and compared the results. I have a feeling the result of method No. 1 is slightly superior, but I’m not sure if I’m not biased, as I intuitively think method 1 should be better.
So my questions are:
Would anyone have any recommendations as to how to proceed? - based either on the theory behind the models, or your own experiences (or both)
From the theoretical point of view of how the models work - is any of the methods described above better than the rest? Or perhaps a different method still would be most recommendable?
Furthermore, which Dione model would be best for 1990s movies/2000s movies/1980s or previous decades movies? The info I’ve found on this forum is that "“Dione-DV” is trained mainly to deal with high quality interlaced digital source such as DV, DVCPRO, DVCAM, etc. “Dione-TV” is designed to handle analog TV/VHS/8mm sources . I would think 1990s movies (and any older ones) should be closer to “Dione-TV” - is this correct?
Any info, comments and suggestions would be appreciated .