what happens if I made a dataset of downscaled anime girls in 64x64 sprites from 1024x1024 anime art?

Can an illustrious lora learn to do something like pix2pix from the low res to the high res?