Software text-to-speech

This paper describes the port to software of a mature text-to-speech synthesis technology that has been sold as a series of hardware products for over ten years. Originally developed as an alternative to a character cell terminal and for telephony applications, today it is also used to provide people with visually disabilities access to information. The quality of text-to-speech is extremely high in both intelligibility and naturalness and uses a digital formant synthesizer to simulate the human vocal tract. Prior to very high speed processors, the computational demands of this synthesizer placed an extreme load on a workstation. This study used a Digital Equipment AlphaModel 600 workstation to simultaneously convert many text streams to speech. The power of modern RISC processors allows applications to freely use speech for output. This capability has prompted the need for a text-to-speech application programming interface (API). The API that we have developed for TTS software is supported on multiple platforms and multiple operating systems. This paper presents a description of the TTS software architecture. The API is also specified. Finally, our experience in porting the TTS code base from the previous hardware platforms is described. © 1997 Kluwer Academic Publishers.