There is a new maintenance version on github. This version deals with API changes to the Google speech recognizer. Please see the speech project for the Jaivox library and the gui project for the JAG tool (August 12, 2014).

Download the maintenance version Jaivox 0.7a.

Download Jaivox 0.7, the last released version of the library.

For a new graphical tool for generating applications please see Jaivox Application Generator.

Download the Jaivox Application Generator

This includes the source code for the tool, and the Jaivox library as a jar file.

There is a maintenance version of the gui tool using the updated Jaivox library.
Download the maintenance version: new version, zipped including libraries.

Download here: zipped including libraries.

You can also download
from SourceForge.

For the latest version of this tool see tool Git repository.

For the latest version of the Jaivox library see Jaivox Github repository.. You can also find all the older versions on Github.

For specific Jaivox applications please see Application notes downloads

Download the Jaivox source code

The Jaivox code is mostly in Java. There is a small agent implementation in C++ that is generated for each example.

Download zipped

The current version is 0.7 (March 2014). The main difference from 0.6 is that this version contains an error correction method for increasing recognition accuracy. This method is currently implemented only for English. There are some new wave file routines that can be used for aligning audio with text, as in generating subtitles.

You can also download from Source Forge. Note that older versions are on Github (see link at top of this page.)

After downloading unpack to a directory and follow the installation instructions. Note these versions are for Linux, it does not use any specific features of individual versions of Linux.

Please see installation instructions.

You can also review the API or browse the source code.


If you are using the web-based Google recognizer, you do not need any other software to make things work. If you are using Google's web-based text to speech, you need jLayer from Javazoon.

The Jaivox software needs a few things to make it work with Sphinx. You may have some of these components installed already (for example festival is usually installed on Linux computers.)

  1. Java Development Kit for example from OpenJDK or from the Oracle Java SE download site Based on our understanding, the recently announced Java security problems do not apply to Java applications, only to web plug-ins and only for the Oracle Java 7.
  2. Ant from Apache
  3. sphinxbase from Carnegie Mellon University. After installing sphinxbase at a particular location (usually the libraries are in /usr/local/lib) make sure that LD_LIBRARY_PATH includes this location.
  4. Language Modeling Toolkit from Cambridge University and Carnegie Mellon University..
  5. Sphinx 4 from Carnegie Mellon University
  6. Sphinx requires JSAPI from various implementations of the JSAPI.
  7. Festival from the University of Edinburgh. You will need the development version, if you are installing it as a package, please install both festival and festival-dev.
  8. Instead of festival, you can use the Free TTS from sourceforge. See using freetts for information on installing and using this Java-based synthesizer. In some installations, festival may have conflicts with Java, thus freetts can be used in those situations.