(1.1.2), we have greatly expanded support for and integration of automated
text to speech synthesized voice systems. Using "flite" based text to
speech synthesis engines, initially developed through Carnegie Mellon
University, and commercial derivatives of "flite", such as Cepstral's
Theta, it is now possible to have fully automated high quality voice
synthesis for GNU Bayonne based telephony applications.
With this new release, we have also expanded upon application scripting
support to enable the creation of applications that can either use
synthesized voices, or pre-recorded samples, based on the GNU Bayonne
server runtime and build configuration. This allows one to create
standard GNU Bayonne telephony applications which will operate whether
voice synthesis is available or not, as well as simplify the development
and maintenance of such applications.
To improve synthesis performance, GNU Bayonne supports options for
automatic caching of non-dynamic synthesized speech content. GNU Bayonne
also supports active selection of different synthesized voices within
scripting. In the future we will tie together synthesis with GNU
Bayonne's existing support for XML parsing to provide a voice web browsing
portal over the public telephone network.
While we have initially worked with integrating "flite" based text to
speech synthesis systems with GNU Bayonne, we are interested in supporting
other voice synthesis systems as well. Our long term goal is for GNU
Bayonne to provide pervasent application driven interactive voice response
in every voice communication medium possible. As part of this goal we are
also interested in integrating automated speech recognition systems.