ARIA Speech Recognition and Synthesis Packages

The ARIA speech packages provide software libraries which, in combination with a sound card and microphone/speakers will allow you to do speech synthesis and simple speech recognition.

The three speech packages are:

• ARIA-ArSpeechRec_Sphinx for speech recognition using Sphinx2
• ARIA-ArSpeechSynth_Cepstral for speech synthesis using Cepstral
• ARIA-ArSpeechSynth_Festival for speech synthesis using Festival

To download, shift-click on the version below that matches your PC's operating system. Make sure you have the same version as you have ARIA.

For Windows, the self-extracting installer will by default install into your Aria directory, with shortcuts in the Start menu for documentation, source code and projcet files. For all operating systems, the library will be installed in a subdirectory of the Aria directory (Aria/ArSpeechRec_Sphinx, Aria/ArSpeechSynth_Cepstral, or Aria/ArSpeechSynth_Festival) with header files, documentation, source code, and neccesary third-party libraries and header files. Please read documentation and README.txt files in each package for information about usage and rebuilding.

ArSpeechRec_Sphinx and ArSpeechSynth_Festival are available to all customers. ArSpeechSynth_Cepstral is only available to customers who purchased the optional Speech Synthesis and Recognition software package, or purchased a robot model with Speech Synthesis and Recognition as a standard feature, after February 1, 2005.

ARIA-ArSpeechRec_Sphinx

The ArSpeechRec_Sphinx library provides a C++ API to the Sphinx2 speech recognition system, open source software from Carnegie Mellon University. This API integrates with other Aria sound classes, and provides a simpler, C++ interface to the Sphinx2 C library.

Archived versions available here

ArSpeechRec_Sphinx requires ARIA.

ARIA-ArSpeechSynth_Cepstral

The ArSpeechSynth_Cepstral provides a C++ API to the high-quality Swift speech synthesis library, provided by Cepstral, LLC. ArSpeechSynth_Cepstral implements the common Aria ArSpeechSynth interface and integrates with other Aria sound classes, and provides a simpler C++ interface to the Swift C library. Two voices (David and Diane) are included. Swift produces higher quality speech than Festival, and additional voices are available from Cepstral.

Note, ArSpeechSynth_Cepstral is only available to customers who purchased a robot base with the optional Speech Synthesis and Recognition software package added, or purchased a robot model with Speech Synthesis and Recognition as a standard feature, after February 1, 2005.

Archived versions available here

ArSpeechSynth_Cepstral requires ARIA and libswift. (Contact ActivMedia support if you require a reinstallation or upgrade of libswift.)

ARIA-ArSpeechSynth_Festival

The ArSpeechSynth_Festival library provides a C++ API to the Festival speech synthesis system, open source software from The University of Edinburgh. ArSpeechSynth_Festival implements the common Aria ArSpeechSynth interface and integrates with other Aria sound classes, and provides a simpler C++ interface to the Festival C library. While Festival produces lesser-quality speech than Cepstral's Swift, the Festival library source code and voices are available for modification under an open-source license. Several English voices and one Spanish voice are included. Additional voices are available from the Festival and Festvox projects on the web, or from other researchers.

Archived versions available here

ArSpeechSynth_Festival requires ARIA.