In this case it might be better to first downsize the video (e.g. with handbrake, should be very fast) and only then upscale it in TVAI 2x with Iris or Proteus.
Otherwise you would need MUCH more time and get a really large file, because it would end up as 4k video but without being near to 4k from the amount of detail.
A bit as it is now: formally 1080p resolution but details only like a bad SD.
So I guess my workflow on this would be:
Handbrake with halving the resolution in both axes. Then load the resulting video into TVAI. There it could be good approach to start with the “Upscale to HD resolution” preset
I would only turn on stabilization and Deblur if really necessary. Upscaling 2x automatically enables “Enhancement” where you can then chose the used model:
The preset uses Proteus which isn’t bad at all - but I mostly prefer Iris.