Betacam Avisynth Script - Levels and Denoising

[ENunn]

I've been doing some Betacam transfers for about 6+ months now. I've been enjoying the experience and I'm very impressed with the quality. I have been uploading some of these to YouTube, but for my newer captures I was wanting to do some cleanup before scaling to 4K (for less YT compression).

My capture setup:
Sony PVW-2800 > Component Out > Open Source Scan Converter - Component IN - Passthrough (480i) mode > HDMI Out > Avermedia Live Gamer 4K > Virtualdub > MagicYUV 4:4:4

The reason why I record in 4:4:4, is because I get a bit more information in the reds. It's unnoticeable to 99% of people without zooming in a lot, but I wanted to record as much information as possible.

First off, I usually adjust levels for all my tape transfers, VHS or Betacam. I lower the brightness until the blackest part of the tape is at 16, then I raise the contrast until the brightest point of the tape clips at 235, then I bring it down until it isn't. But a dilemma I've been having is that there's some instances with white text or an all white background. Even though they're white, some pixels clip.

RAW:

converttoyuv444(matrix="rec601", interlaced=true).convertbits(10)
Levels(55, 1, 905, 0, 1020, coring=true,dither=true).tweak(bright=0, cont=1.00, hue=-0, sat=1.0, coring=false, dither=true).convertbits(8)

Should I still adjust until it's not clipping or is it OK for it to clip since it's an all white background?

And another thing, I would like to do some denoising, as my captures are full of noise, especially after sharpening. It might be normal, but I do want to reduce the noise. I'm pleased with the results of this one command, but it's unfortunately unusably slow. When encoding, it runs at less than a half a frame per second, and I'm using hevc_nvenc. So in the span of 8+ hours, it only rendered five seconds of it. When previewing in AvsPmod, it used to take maybe a minute for it to show up, but now it takes well over 15 minutes for some reason. I don't know why.

No Denoise+Sharpening, Aspect Ratio Corrected, Cropped, Upscaled to 4K

pre=fluxsmootht(3).removegrain(11)
SMDegrain(tr=6,thSAD=500,contrasharp=40,prefilter=pre,str=1.2,refinemotion=true)
converttorgb24(matrix="rec601", interlaced=false)
unsharp()
cropresize(indar=15.0/11.0, resizewo=true)
selectodd()
Crop(6, 0, -8, -2)
SuperRes(matrixin="rec601", 2, .43, 0, """nnedi3_rpow2(4, nns=4, cshift="Spline16Resize")""", matrixout="rec709")
x = float(width) / float(height)
spline36resize(round(x * 2164),2164)

I was wondering if there was a way to speed this up a little. Is my Avisynth install broken? Is there a way I can optimize my script? I have some scripts attached. I can't seem to attach this video sample, so here's a download.

[dellsam34]

I believe if you can take HDMI out of the equation you will get better results, I also see dot crawl, this could be baked in the tape or the deck is causing it, The Y level should be legal 16-235 range, it should not be clipped at all.

My capture setup for Betacam is:
Sony J3-SDI > SDI Out > BM UltraStudio SDI > USB 3.0 Out > BM MediaExpress > Lossless AVI YUV 4:2:2 > AvsPmod QTGMC de-interlacing/cropping/resizing to 1440x1080/encoding HD h.264.

I don't do any other processing so I can't help you on that but there are members here who are passionate about video restoration and I'm pretty sure they can help.

[ENunn]

I believe if you can take HDMI out of the equation you will get better results

Maybe, but the OSSC is one of the best A2D converters out there. Quality is really good. There's that, and I don't have any good component capture cards, hence why I'm using HDMI right now.

I also see dot crawl, this could be baked in the tape or the deck is causing it

It's baked in the tape.

The Y level should be legal 16-235 range, it should not be clipped at all.

So I guess I should keep it lower, got it. It doesn't look too bad either way and it doesn't affect the main content, just the text and any white background where only a few pixels clip.

My capture setup for Betacam is:
Sony J3-SDI > SDI Out > BM UltraStudio SDI > USB 3.0 Out > BM MediaExpress > Lossless AVI YUV 4:2:2 > AvsPmod QTGMC de-interlacing/cropping/resizing to 1440x1080/encoding HD h.264.

I want your setup so bad lol. You put out some great stuff on YT! But J3's are so expensive and I can't find any Blackmagic SDI USB3 capture cards anywhere, only thunderbolt which I don't have any devices that support it.

[Sharc]

@ENunn: You need not worry too much about occasional short luma excursions beyond 235 or below 16. It may be caused by noise and sharpening halos at high contrast edges.

Your script is quite "heavy" I think. You may want to identify the "time guzzlers" using AvsMeter which is included in AvsPmod (Video->Tools....), and/or by enabling/disabling individual filters.

[dellsam34]

Maybe, but the OSSC is one of the best A2D converters out there. Quality is really good. There's that, and I don't have any good component capture cards, hence why I'm using HDMI right now.

You could try some cheap options for component capture, I would start with this low risque Roxio one. Betacam component is very forgiving.

[Bwaak]

Too much NR for my taste.The texture is lost. The edge of his right cheek - left on the picture - shows strong aliasing compared to the soft edge of the original.

[Alwyn]

Add, the very end of a script:

Prefetch(XX)

Where XX is the number of Logical Processors you have (Task Manager>Performance Tab>CPU)

That gives me a 400% boost in encoding speed compared to no Prefetch (I have 24 logical processors)

I have also put, at the start:

SetFilterMTMode ("QTGMC", 2)

You're not running QTGMC so I will have defer to an AVISynth expert for the exact code in the brackets for you.

That gives me a total of 500% better than no Prefetch.

[ENunn]

@ENunn: You need not worry too much about occasional short luma excursions beyond 235 or below 16. It may be caused by noise and sharpening halos at high contrast edges.

Gotcha.

Your script is quite "heavy" I think. You may want to identify the "time guzzlers" using AvsMeter which is included in AvsPmod (Video->Tools....), and/or by enabling/disabling individual filters.

AvsMeter is barely running. It's been sitting at this for more than five minutes. I have done nothing to my script other than change prefetch. I did an encode overnight and I was getting 2-3fps. Now I can't even get that much.

Add, the very end of a script:

Prefetch(XX)

Where XX is the number of Logical Processors you have (Task Manager>Performance Tab>CPU)

That gives me a 400% boost in encoding speed compared to no Prefetch (I have 24 logical processors)

I always have Prefetch at the end of my script. I use 6 instead of 12. When I have it set to 12, Avisynth hangs.

I have also put, at the start:

SetFilterMTMode ("QTGMC", 2)

You're not running QTGMC so I will have defer to an AVISynth expert for the exact code in the brackets for you.

I am running QTGMC. I have an avsi file in my plugins folder called mtmodes.avsi. I remember getting it off of some site, it automatically added MT modes for pretty much everything. But to be fair, I dunno if that helped or made it worse.

I added it, but now my script is running at less than half a frame per second. I just did an encode overnight, so I dunno if that's the reason why or not.

I think Prefetch is causing the slowdown. It shouldn't, but every time I try to add Prefetch, Avisynth either hangs or takes ages to process the preview. I really don't know what's going on.