Jeremy this is good timing, or so I think anyway. Thank you for taking this nightmare on.
I've been on the stump for the past two to three weeks, calling for improvements for text to speech software in Linux. Being a KDE user, I was extremely dismayed with both the quality the voices and the heavy coat of dust that's accumulated over Jovie and Kmouth. An effort is being put together to try and make TTS in Linux a usable option for those of us who are speechless. I lost my voice by choice. Pre-cancerous growth on my larynx prompted me to pre-empt this monster and take away its playground. But for now we are looking at MaryTTS and hoping we can do something with that app. Maybe get a deb or rpm package made so that getting it installed isn't such a nightmare
The biggest problems users face when trying to get TTS to work in Linux is the convoluted and maddening process it takes to add more human-sounding voices. And even then if you hit the "I screwed up and did something right" lottery, those voices supplied by Mbrola aren't all that good. And that's being polite.
You can get an idea of what we are doing and why at fossforce. I hope we can collaborate sometime soon. Kmouth was promising at first blush but as you mentioned, there's a lot to do to bring Kmouth up to current standard. If you have any ideas, especially if you've found a secret stockpile of actual human sounding voices that aren't guarded by hell hounds (patents and skyrocketing licenses) let me know.
If interested, here's an idea of what we are doing and the direction we hope to go. This is more fluff than substance...I can say that because I wrote it. =:^)) I hope we get some time to talk and to compare plans of direction and scope. http://tinyurl.com/k89q3xc