A complete text-to-speech system has been created by the authors, based on a tube resonance model of the vocal tract and a development of Carré’s “Distinctive Region Model”, which is in turn based on the formant-sensitivity findings of Fant and Pauli (1974), to control the tube. In order to achieve this goal, significant long-term linguistic research has been involved, including rhythm and intonation studies, as well as the development of low-level articulatory data and rules to drive the model, together with the necessary tools, parsers, dictionaries and so on. The tools and the current system are available under a General Public License, and are described here, with further references in the paper, including samples of the speech produced, and figures illustrating the system description.
