OCR (optical character recognition) is pretty tricky to do at all, let alone well. There are a few open source libraries (tesseract) written in c, and wrappers written in python, but even this was difficult to get working well.
https://www.newocr.com doesn't look too bad, through the web interface. I have not tried automating the web interface, but it did not look too bad. It also has an rest api, which would make scripting easier, but you only get 200 conversions for free.
If you can host your own service on your Pc, http://www.openocr.net seems like a good option.
If you can claim to be a student, http://ocrsdk.com/for_students/ also offers an API.