SmartTextExtraction looks like exactly what I need! However, I have to be able to run offline.

How can I download it and run the algorithm locally?

Tesseract gave me pretty poor results, probably because I'm using photos of printed pages instead of flatbed scans.