In this paper we present a unistroke keyboard based on computer vision. The keyboard can be made of paper containing an image of the keyboard which has an upside down U-shape. Each character is represented by a nonoverlapping rectangular region. The user enters a character to the computer by covering the character region with a stylus. The actions of the user are captured by a camera and the covered key is recognized. During the text entry process the user need not have to raise the stylus from the keyboard and this leads to faster data entry rates. In a companion system the user imitates writing on a surface using a pointer or a stylus. In this case the trace of the pointer is analyzed and the characters are recognized. The character set of the continuous hand writing system is based on the Graffiti alphabet to achieve very high recognition rates.