I recommend turning the characters that Aozora Bunko represents using image files into the corresponding Unicode. IIRC, the file name contains the men, ku and t

I recommend turning the characters that Aozora Bunko represents using image files into the corresponding Unicode. IIRC, the file name contains the men, ku and ten parameters which you can use to derive the corresponding Shift JIS 2004 byte representation https://en.m.wikipedia.org/wiki/Shift_JIS#Shift_JISx0213_and_Shift_JIS-2004 https://en.m.wikipedia.org/wiki/Shift_JIS#Shift_JISx0213_and... and then turn into Unicode using standard transcoding utilities.