Text-To-Speech Formant Synthesis For French

Abstract

This chapter presents a formant synthesizer for French. The first steps of text-to-speech synthesizers are independent of the synthesizer’s type. Letter-to-sound conversion, phoneme duration and intonation calculation can be done in different ways regardless of the signal synthesizer. The main difference resides in the use or not of a large dictionary to get information on each word. Although there are other approaches to design a text-to-speech synthesizer from the acoustic point of view, formant synthesizers still have the advantages of being small in computer memory size and easy to modify to get new voice characteristics. Many fundamental voice parameters can be controlled. Consequently, the formant synthesizer development is time-consuming. Each detail of speech theory has to be understood and mimicked in the synthesizer. Singing can be achieved with the synthesizer by controlling duration and intonation for each phoneme of the song.