Version 0.3 of Voiceglue is now available on the download page.

This version contains a one-line change to the voiceglue executable that handles newer versions of LibXML. These newer versions of LibXML give a different nodeName for text nodes than it used to. The symptom of this failing in the old version is that none of the TTS text gets heard.

Version 0.2 of Voiceglue has now been released.

It fixes a very nasty bug with cached grammars, that caused any cached grammar to terminate the script! Hard to get much done with that bug! Thanks to users on the mailing list for finding it.

Also, it removes install errors in cases where the special user accounts already exist.

The first release of Voiceglue is now available. We finished testing release 0.1 this afternoon, and it’s downloadable.

This first release supports DTMF input (no speech recognition yet), recording, and playback of TTS (via flite) and pre-recorded audio. We plan to integrate speech recognition soon using LumenVox’s engine and possibly Sphinx. Further details on features of this release are found here.

Lastly, to support the project and users, we’ve setup a mailing list.

Well, that last issue turned out to be a significant one. The base code was not properly processing SRGS grammars. We needed to fix the grammar parser to handle SRGS. We looked for an open source solution, couldn’t find one. So we had to write one from scratch. That work is now done. We simply need to do the final integration, some quick testing for obvious problems, and then package up the release. Alas, in the intervening time we had to shift resouces to a higher priority project. We are hoping to return to this last bit of effort in a couple weeks, in which case we should release by mid-May.

First release is now within days! (I’m pretty sure). We’ve been able to spend the extra time to get TTS and proper caching of both pre-recorded and TTS-generated audio working properly, shared across call threads. The last item before initial release is making sure the SRGS grammars for DTMF are getting processed correctly. As soon as that’s done, we’ll be posting the first code release.

We’ve been sidetracked with critical contract work, which is delaying us for the time being on finishing the release.

It’s getting close! With the official Aterisk 1.4 now out, we have a target platform against which to release. The initial release will be tested on Fedora Core 6 for both build and runtime. Digium’s voice recognition work that supports VoiceGlue is evidently done, but hasn’t yet been officially released.

The beta release didn’t make it in November, so we are now targeting December. It’s being built and tested against Asterisk 1.4, which still hasn’t come out. Also, Digium’s voice recognition work that supports VoiceGlue isn’t done yet. As with much work that is done for free, you have to be a little bit patient! Or, if you want to contribute, financially or otherwise,
it could push things along faster.

Our initial release is being pushed back to the end of November.This is due to all of the cleanup and testing that is going intothe first release.The good news is that this delay should turn the product froman alpha into a beta.