Is there a way to make Chinese (traditional) OCR more accurate?

[YamatokotobaEnjoyer]

Can't get google lens to work so have nothing to compare it with, but Paddle keeps messing up simple hanzi like:

於 sees it as 放
只 sees it as 尺
sometimes 刪 gets recognized as only 冊

Ther are some other examples, but those were the ones that kept popping up over and over again cuz those are very common words.

Somehow it gets the more complicated stuff right but breaks against the silliest things. Is there a way to calibrate this thing to be more accurate? I see there are settings that can be tweaked, does anyone have some good preset that would make stuff like this disappear?