CHATR: A generic speech synthesis system

Abstract

This paper describes a generic speech synthesis system called CHATR which is being developed at ATR. CHATR is designed in a modular way so that module parameters and even which modules are actually used may be set and selected at runtime. Although some interdependencies exist between modules, CHATR offers a useful research tool in which functionally equivalent modules may be easily compared. It also acts as a simple system for those less interested in the internals of speech synthesis but just wish their computer to talk.