Speect: a multilingual text-to-speech system

Abstract

This paper introduces a new multilingual text-to-speech system, which we call Speect (Speech synthesis with extensible architecture), aiming to address the shortcomings of using Festival as a research sytem and Flite as a deployment system in a multilingual development environment. Speect is implemented in C with a modular object oriented approach and a plugin architecture, aiming to separate the linguistic and acoustic dependencies from the run-time environment. A scripting language interface is provided for research and rapid development of new languages and voices. This paper discusses the motivation for a new text-to-speech system as well as the design architecture and implementation of the system. We also discuss what is still required in the development to make the new system a viable alternative to the Festical-Flite tool-chain.