we have a IVR that works with Linux/Asterisk/Dialogic phone card.
The logics is:
1. a user dials the given phone number
2. a greeting message invites him to dial his (numeric) roll
3. Asterisk calls a small external procedure.
4. the external procedure runs a query into a database
and returns a text string
5. the text string is converted into sound by Cepstral TTS
6. the user is invited to hang
We are not satisfied with the current TTS.
As the words to translate are 99% the same, we would like to prepare a library of prerecorded tokens.
An application will scan the text starting from the first token to the last one in a list.
The text will be composed by concatenating all possible prerecorded tokens .
The TTS will be invoked only for the minimal parts of the text that change everyday and are not important (usually the name of person or the name of a ship).
The application will be coded in python or php.
The pseudocode of the wanted application is attached.
The selected coder will install his application directly into our IVR (where he will have access via SSH) until complete satisfaction.
The precorded sound format is to be agreed on.