Recent update: 24/07/2016
Tesseract is powerful open-source text recognition engine, currently being developed by Google (yes, that almost-evil, spying and annoying corporation).
It uses neural networks and already is trained for many fonts, etc.
It is used for example in OpenCV framework. But there are cases when you don't want to take whole bloated up OpenCV with you — then you can take only whole bloated Tesseract :3 (it needs at least ~20mb database of neural networks data... nothing to do with that).
The following bindings are from code I've used in some project, so they are incompleted — just doing "what is needed", mostly a starting point.
Anyway ready-to-use engine binaries included (Windows x86 .dll + .lib to use it), as well as other required libraries and simple "helloworld" example.