Biometric systems, such as face tracking and recognition, are increasingly being used as a means of security in many areas. The usability of these systems depend not only on how accurate they are in terms of detection and recognition but also on how well they withstand attacks. In this paper we developed a text-driven face-video signal from the XM2VTS database. The synthesized video can be used as a means of playback attack for face detection and recognition systems. We use Hidden Markov Model to recognize the speech of a person and use the transcription file for reshuffling the image sequences as per the prompted text. The discontinuities in the new video are significantly minimized by using a pyramid based multi-resolution frame interpolation technique. The playback can also be used to test liveness detection systems that rely on lip-motion to speech synchronization and motion of the head while posing/speaking. Finally we suggest possible approaches to enable biometric systems to stand against this kind of attacks. Other uses of our results include web-based video communication for electronic commerce.