I wanted to say thank you to everyone that responded on the previous post. Whether intentional or not, some of the comments gave me some ideas for elaboration and improvement of my pipeline.
I discovered that the TINK 4k has the ability to do 480i pass through via custom modelines, which I enabled and updated my capture script to take advantage of. The biggest downside to using the TINK is that it converts color to RGB, which results in slightly-less-than-faithful raw captures.
This leads into the new VapourSynth pipeline I put together. When strung all together, we're now performing:
- Lossless RGB NTSC capture, converting to YUV 4:4:4 for processing until the final 4:2:0 output
- QTGMC source-matched deinterlacing to 59.94p
- Motion-compensated denoise and CCD chroma cleanup
- Dehaloing, contrasharpening, and debanding
- Lanczos upscale to 4:3 1080p
I'm currently punting on any neural upscaling, because I really don't want to introduce any semblance of hallucination.
You also may have noticed I added waveform and vectorscopes to the monitor stream, to get a quick and dirty readout of the brightness and color levels of a given tape.
I understand that this project is now probably blurring the lines of appropriateness for this particular community, so it will likely be my last post on the matter here. I just wanted to give an update, and say thanks!
