An Urdu OCR has been developed for Nastaliq font. It is currently in the initial development cycle. It utilizes a technique of separating the main body of ligatures from secondary components. A right-to-left sliding window captures projection, concavity and curvature information. The data from the window is fed to the model for training which has a recognition rate of 98% for the widely used ligatures. The system has direct application in translation software.