ML-FreeTTS

Multilanguage Java voice synthesizer

(English and Italian voices)

 

 

ML-FreeTTS (Multilanguage FreeTTS) is a portable Java library dedicated to text-to-speech (TTS) synthesis.


The library is based on Java FreeTTS library (
http://freetts.sourceforge.net/docs/index.php/), developed by the Speech Integration Group (Willie Walker, Paul Lamere and Philip Kwok) of Sun Microsystems Laboratories.

 

Sun FreeTTS library, in its turn, is based on FLite version 1.2 (http://www.speech.cs.cmu.edu/flite/), a library for C language originally developed by Alan Black and Kevin Lenzo at Carnegie Mellon University (CMU).

.

The original version of FreeTTS supports only the english language. For the other languages, like italian, the original library cannot be used.


Some months after the release of the version 1.2 of the CMU FLite library, Piero Cosi and Fabio Tesser, two researchers of the Institute of Cognitive Sciences and Technologies (CNR-ICST Padova) have released a module for italian voice in Festival and a C program called Italian-FLite.

 

Italian-Flite is based on the the italian version of Festival. Italian-Flite and the italian version of Festival are the result of a collaboration between CNR-ISTC and the Istituto Trentino di Cultura - Centre for Scientific and Technological Research (ITC-IRST now Bruno Kessler Foundation). It is only a program for C language, so it cannot be used in network or browser-based applications.


Filippo Battaglia, a researcher of Visilab Research Center of the University of Messina, after releasing the C library ML-FLite (hosted in this website too), has experienced a new approach that has allowed
to extract the descriptive informations relative to the italian language directly from ML-FLite source code.

 

These informations have been ported into the source code of the Sun Java FreeTTS library. Furthermore, the core of the original Java library has been modified in way to add the support for LTS functions (required for the italian lexicon) and to make easier the use of the library in browser applications (using  Java web start (JWB), a technology provided by Sun in Java Runtime Environment (JRE)).

 

In this way, our Java source code contains both the LPC diphone database/lexicon provided by ICST-CNR institute for the italian language, and the diphone database/lexicon provided by CMU/Speech Integration Group for the english language.

 


The result is ML-FreeTTS, the first Java library dedicated to text-to-speech synthesis, able to use both italian and english voices without differences.

 

Sponsors

 

ML-FreeTTS has been developed by CECUM (Centro di calcolo Università di Messina) in collaboration with VISILAB (Computer Vision and Human-Computer Interaction Lab, University of Messina) with the support of the governmental project "Innovazione nell'area dello stretto", sponsored by CNIPA (Centro Nazionale per l'Informatica nella Pubblica Amministrazione - Italy).


Features

  • Supports english voices (KAL, KAL16) and an italian voice (IT_PC);

  • The system is based on LPC (Linear Predictive Coding);

  • The required databases are fully contained in JAR archives: all you need can be integrated in a single JAR/JNLP Java application;

  • Fully optimized;

  • The engine is compatible with LTS rules and LTS functions;

  • English and Italian lexicons provided;

  • Expandable with other languages;

  • Compatible with Sun Java Development Kit 1.6 (JDK 6);

  • Integrates a new API that makes easier the use of the synthesizer at high level;

  • Supports Java Web Start (JWS) for remote applications performed via web browser.

Supported platforms

 

ML-FreeTTS can be used with all operating systems that support a Java Virtual Machine (JVM) compatible with Java Development Kit 1.6 (JDK 6). We advice to use the IDE NetBeans, because of its native support for JNLP format (Java Web Start applications).

 

The browsers must use the Java Runtime Environment 1.6 (JRE 6). We have tested the compatibility of the applications with Windows XP, Windows 7 and Ubuntu Linux.

 

Note: OpenJDK is not supported by ML-FreeTTS.

 

Demos

 

 Listen FreeTTS in italian

 

 Listen FreeTTS in english

 

 Use FreeTTS with your text

 

 

 

Note: Java Web Start (javaws) is required. The Linux users must check that their browsers are configured for opening JNLP applications using javaws and not OpenJDK.

 

Downloads

 

Download MLFreeTTS library (version 1.1)

License and terms of use

ML-FreeTTS is free software. You can redistribute it under the terms of GNU Lesser General Public License (LGPL).  See the file README.PDF in the package for further details.

 

This project is mantained by Filippo Battaglia. If you have some questions about it, you can contact me at the mail address: filbattaglia@libero.it

 

Other archives

These packages are for the developers of the library code only.

A normal ML-FreeTTS user, not affiliated with Visilab, doesn't need these archives.

 

Download ML-FreeTTS auxiliary package (developers only)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Università degli

Studi di Messina

 

Visilab

Computer Vision Laboratory

 

 

 

 Back