Second ESCA/IEEE Workshop on Speech Synthesis

September 12-15, 1994
Mohonk Mountain House, New Paltz, NY, USA

A Modular Architecture for Multi-Lingual Text-To-Speech

Richard Sproat, Joseph Olive

AT&T, Murray Hill, NJ, USA

We describe the architecture of the AT&T Bell Laboratories New text-to-speech (TTS) system. The modular pipeline design of NewTTS enhances capabilities for research on TTS since it facilitates revision and replacement of modules of the system, and it makes it easy to insert into the pipeline tools that modify various TTS parameters. While we illustrate most of these points with reference to the English version of NewTTS, we will also sketch how systems for other languages are being developed within the NewTTS framework.


Full Paper

Bibliographic reference.  Sproat, Richard / Olive, Joseph (1994): "A modular architecture for multi-lingual text-to-speech", In SSW2-1994, 187-190.