VoiceGlue started as a project to obtain a less expensive VoiceXML capability for a commercial application. This commercial application was written in VoiceXML, but was served by expensive proprietary VoiceXML servers. The application used only DTMF input and pre-recorded output. Having no text-to-speech or voice recognition requirements, it was just about the easiest kind of VoiceXML application to port. We searched for free software or open source software that would implement VoiceXML files, but found none. However, we were intrigued by the existence of both Asterisk and OpenVXI, and were curious if it was feasible to integrate the two. Our curiosity got the better of us, and we started in.

By Summer 2006, we had achieved an integration of OpenVXI and Asterisk that could run the commercial application. The result was quite stable. Encouraged by this initial success, we began to consider if such an integration could be of general usefulness. Having received much benefit from the open source and free software communities over the years, we felt it was time to give back to the community.

During the winter of 2006 and through the Spring and summer, we continued to develop and test a general purpose release. We refactored our initial work, to integrate with OpenVXI more cleanly. We added Flite’s TTS. We wrote from scratch SRGS grammar parser, so that the project could handle standard VXML apps. We added caching of prompts, grammars, and TTS results.

We’re delighted to release this to the community, and hope that people find it useful.