The Epos Speech Synthesis System

Epos is a new language independent rule-driven Text-to-Speech (TTS) system primarily designed to serve as a research tool. Epos is (or tries to be) independent of the language processed, linguistic description method, and computing environment.

Index

About Epos

Version 2.5.16 (August 2003) is now available for research, educational and individual use for free. It has the following features

Czech and Slovak text to speech synthesis configuration
LPC and time domain diphone speech synthesizers (only time domain synthesizers are freely available)
All languange dependencies are defined in the configuration files, as opposed to the source code
Documentation in English included in the distribution

Epos can hopefully compile using GNU C++ (preferably egcs), Microsoft Visual C++, Watcom C++ or Borland C++ on most UNIX and MS Windows NT systems. (Some of the ports are less maintained than others. Let us know if your OS/compiler/Epos version combination has problems; if you want to attempt another port; we may also be able to help you.)

Call for partnership

As Epos is still a very young and evolving system, we're very open to any comments and especially cooperation. In particular, diphone inventories and rule files for other languages than Czech and Slovak are relatively easy to add for some languages, and we will be happy to provide the necessary instruction and support for this. Also, Epos is currently unable to perform a complex syntactic analysis of a text or to apply a context dependent probabilistic model to distinguish between homographs in languages like English. These and other limitations offer wide room for experimentation and improvement, and the modular structure makes Epos an ideal environment for testing potential Text-to-Speech components without the need to reimplement the rest of the system. If you have a question about the Epos Speech System, you can send it to the Epos developers' mailing list, namely epos@braille.mff.cuni.cz.