Design suggestion: The server just for synthesis

[Michael Pozhidaev]

Hello, everybody!

I would like to discuss one suggestion. There were a lot of ideas of
TTS calls unification and what do you think about creation an new
separated server dedicated only to speech synthesis? Main features
and properties are:

1. The server has no any audio output.
2. It can accept connection, receive portion of text and send audio data
back to the client.
3. Connections are completely independent.
4. SABLE(SSML) markup emulation for TTS without their support.
5. Clear configuration format and plug-in structure.

With UNIX domain sockets performance overhead must be minimal. Speechd
or VoiceMan may just forward all speech synthesis tasks to new server
and care only for audio output  and screen readers dispatching if user
has more than one .

Consequences:

1. Simplification of servers we currently have and unification of
configuration methods. Now server has plug-ins for both TTS and audio output,
but will have plug-ins only for audio output, TTS calls will be moved
into separated application.
2. Network speech synthesis support for using single speech server
machine in the LAN.

I can participate in this work so much as I have time  free from my job, it
interesting to me and seems can be very useful. This can be new step to
merge speechd and voiceman, not right now, but a real step in any case. 

The language to use can be C or C++. I prefer C++ because of its
development and maintaining simplicity but I also can easy use and C, if
other developers has other preferences.

Thanks!
-- 
Michael Pozhidaev. Tomsk, Russia. E-mail: msp at altlinux.ru
Russian info page: http://www.marigostra.ru/