FSR1 is pretty bad as it’s just upscaling the static image, I agree.
FSR2/3, XeSS and DLSS are temporal, meaning they use info from the previous frames to construct a higher resolution image that gives much better results. They also need to be implemented in the game engine, meaning not every game supports them.
Even without machine translation, stuff like that has been the bane of translating software for ages as they are almost always done with absolutely zero context whatsoever, just a list of words and strings.
Can’t see any reasons why that might be difficult.