It doesn't deal with images at all, but depending on what you're doing it might be useful in cleaning up the OCR text, e.g. asking LLMs to fix errors and then f

It doesn't deal with images at all, but depending on what you're doing it might be useful in cleaning up the OCR text, e.g. asking LLMs to fix errors and then filtering the diffs.