Having OCR capabilities via tesseract (https://github.com/tesseract-ocr/tesseract) would be useful to analyse screenshots.

Possible use cases of OCR currently include phish page detections (specific matches or generic such as Microsoft login page not on correct domain) or fake update style pages (https://www.proofpoint.com/uk/blog/threat-insight/are-you-sure-your-browser-date-current-landscape-fake-browser-updates) and likely others.

I am sure it used to be part of cuckoo-modified or main cuckoo sandbox branch as I remember it being useful for detecting phrases such as "enable macro" being present. I did look and I can't seem to find it but I know at one point it did exist.