A diphone speech synthesis approach applicable to different languages

An unrestricted text-to-speech diphone based synthesis system, easily adaptable to different languages and to different voice styles is presented. In order to obtain such a flexibility, the system has been conceived around language-transparent modules, the language specific elements being well isolated and essentially introduced as interchangeable rules or specific lexicons. In particular, prosody treatment, one of the most language dependent part, has been realized with an approach easy to transpose from a language to another.