At the present time, one of the most popular application of the text-to-speech (TTS) synthesis technique in the telecommunication area is the "text-to-voice" service allowing E-mail users to access their mailbox through an ordinary touch-tone telephone for receiving a voice output of their messages (BERNEY, 1985). We present here the specific problems encountered in French, and the adopted solutions for adapting the CNET's synthesis system to this specific application.