Could use some Avisynth advice

[HansensUniverse]

Hi,

Been experimenting a bit with Avisynth the past days, I've initiated a project where i will be digitizing all of my VHS-C tapes, i have enough time with Avisynth to the point where i can operate it and know a few settings and what they do and have so far been able to derive some what i would consider really good results. However i thought i would share my current settings on Avisynth where i use QTGMC for deinterlacing which has done a fantastic job so far, you will find the parameters below, if there is anything i should change, add or remove let me know, feel free to butcher it if needed, after all all i want is to optimize my results to the best of my capability with the equipment i have. Spite feeling satisfied i feel like i am potentially missing out on some great features there is just so much to choose from, so some advice from the experts would be nice.

Recording software used: Bandicam configured with Lagarith lossless, YUY2 (720, 576) 25fps PCM audio. Pal


Avisynth parameters: SetfilterMTMode ("QTGMC", 2)
FFmpegSource2("Puerto-1.avi", atrack=-1)
AssumeTFF()
QTGMC(preset="Slower", EdiThreads=2, SourceMatch=3, LossLess=2, Sharpness=1.0, TR0=2, TR1=2, TR2=3)
BicubicResize(1280, 960)
Prefetch(6)

[manono]

I'd use AviSource. But I don't know the consensus on that. I'd use nnedi3_rpow2 for upscaling.

[HansensUniverse]

For whatever reason i couldn't get it to work using AVISource, but FFmpeg has worked well thus far, i will try out the upscaler you suggested, i can drop a sample so people can check out my results.

[lollo]

QTGMC(SourceMatch=3, LossLess=2)

I use QTGMC "lossless" configuration only when I want to interlace back the video at the end of the filtering.

Otherwise I think that a not lossles setting is more effective for pure deinterlaced output at double frame rate.

[HansensUniverse]

QTGMC(SourceMatch=3, LossLess=2)

I use QTGMC "lossless" configuration only when I want to interlace back the video at the end of the filtering.

Otherwise I think that a not lossles setting is more effective for pure deinterlaced output at double frame rate.

Interesting, i will try and change the parameters and do a few runs without, would be nice if i could save my self some processing time, it's worth mentioning that the playback deck i use is the Sony SLV D900 which has been an incredibly good reliable unit so far, i believe it's a mid range deck, i cannot find any details regarding an inbuilt TBC so i assume it doesn't have one but it's hard to believe otherwise because it manages to keep even the most worn tapes i have stable, never had more than 1 dropped frames, that was from a 31 min video, very happy with this deck and can recommend if anyone is looking for one for digitization without breaking the bank.

[Sharc]

I prefer NOT do deinterlace my VHS captures. I process the grouped fields individually and don't alter the original interlaced structure. But that's perhaps just me ....
QTGMC is an excellent bobber, no doubt, but I got better results without it (even with lossless=2), especially for somewhat noisy VHS sources.

[lollo]

It is not just you, I do the same!

In general, I do not interlace for the final version too; I deinterlace lossless with nnedi3(field=-2) or QTGMC lossless just to apply the spatial/temporal filtering in a more effective way, and then interlace back. And I also keep HuffYUV file to watch, no loss compression with whatever codec!

However, for youtube videos I want to share with some friends, I deinterlace with QTGMC (not lossless), and then YT compresses to whatever it uses...

I got better results without it

Well, that means that the deinterlacer you use to watch the video is better (altough difficult to believe)

[emcodem]

BicubicResize(1280, 960)

Wouldn't spline64 deliver better results here? nnedi as mentioned above mabye even better i guess.

I process the grouped fields individually and don't alter the original interlaced structure. But that's perhaps just me ....

When rescaling, this techique usually ends up perfect motion but spatial interlaced problems, especially in the middle of the picture, at least if you generate interlaced in the end. (e.g. upscale interlaced SD to interlaced HD)
But when the final result is double framerate and progressive, i guess it would be a very good choice