How to improve OCR?

[mpek]

How can I improve OCR, because obviously some characters are being always recognized wrongly.
In my test, the language is set to german, because I want to work on german subtitle.
For example:
" h " recognized as " b "
" , " recognized as " ' "
Where I can accept the fail with " h " to be a " b ", it is soemwhat strange, that a "low" character appears "high"...
Do you have any recommendations for me?

[niksedk]

What OCR method do you use and what is your source?

nOCR is recommend for larger font sizes, while Tesseract/BinaryImageCompare are recommended for smaller fonts.

Use character inspect (double click on lines) to improve matches in nOCR/BinaryImageCompare. Also, it's often a good idea to start with a small % pct.

Read more here: https://www.nikse.dk/subtitleedit/help#importvobsub

[mpek]

I see, i was leaving it to default settings (binary image compare) and going to test the other methods. Thank you!