OCR of Japanese with furigana (superscripts and subscripts) in image-based subtitles

[jfoust2]

I've tried all four engine modes with Tesseract 5.2.0. All are yielding poor results, due to furigana. Is there a way to ignore the furigana?

Also, is there a way to retain the raw subtitle images after the OCR step, in order to better detect and correct the OCR errors?