Fifth ISCA ITRW on Speech Synthesis

June 14-16, 2004
Pittsburgh, PA, USA

Overview of Voice Building

Alan W. Black

Language Technologies Institute, Carnegie Mellon University Pittsburgh, PA, USA

The work involved in building synthetic voices has substantially simplified over the past few years, partly due to support through the FestVox project, a free set of scripts, tools, and documentation for building voices in new and already supported languages (http://www.festvox.org). However, the building of reliable high quality synthetic voices still requires a substantial amount of skill. In this tutorial I will outline the basic processes required in building new synthetic voices, including phoneset definitions, data collection, labeling, lexicon construction, evaluation, etc. I will also discuss aspects of voice building when considering minority languages where standard resources are not readily available. Finally, I will discuss some of the challenges in making the building of synthetic voices more robust and such that it requires less knowledge of speech technology to be successful.

Bibliographic reference.  Black, Alan W. (2004): "Overview of voice building", In SSW5-2004 (abstract).