This paper presents the development of Urdu text to speech system using Festival framework. Unit selection based, and HMM-based speech synthesizer are developed using 10 hours manually annotated speech data for Urdu text to speech synthesis. The architecture of the Urdu text to speech system is also discussed. The objective assessment of the synthesized speech is evaluated by using automatic speech recognition system. The results show that the speech produced by HMM-based synthesizer has the highest recognition accuracy.