This paper discusses a Text-To-Speech (TTS) synthesis system embedded in a mobile. The TTS system used is unit selection based enative speech synthesizer, where a speech unit is selected from the database based on its phonetic and prosodic context. Speech unit considered in the synthesis is larger than a phone, diphone and syllable. Usually the unit is a word or a phrase. While the quality of the synthesized speech has improved significantly by using corpus-based TTS technology, there is a practical problem regarding the trade-off between database size and quality of synthetic speech, especially in mobile environment. Several speech compression schemes currently used in mobiles today are applied on the database. Speech is synthesized from the input text, using compressed speech in the database, The intelligibility and naturalness of the synthesized speech are studied. Mobiles contain a speech codec, one of the modules in the baseband processing. The idea of this paper is to propose a methodology to use the
already available speech codec in the mobile and read a SMS aloud to the listener, when TTS is embedded in a mobile. Experimental results show the clear possibility of our idea.