Better sounding Text to Speech RRS feed

  • Question


    Does anyone know of any techniques for getting better sounding text to speech. What I mean is somthing that doesn't soubd so Robotic. Is there any notation you can use to do this?
    Thursday, September 6, 2007 5:03 PM


  • Hi,


    There are a number of things you can do to make your app sound better.


    1) In most cases the bulk of your application is formed of fairly predictable phrases. You can get the best quality for those by recording a live human saying those phrases, and storing those in a prompt database. We will then use that instead of TTS for those utterances.


    2) There is a markup language called 'SSML' which is used to help the TTS engine work better. For example you can use it to say 'this number is a currency value' or other things to help disambiguation how things should be pronounced.You can find more details in the documentation.


    3) For specific words which are mispronounced, you can use SSML again to provide alternate pronunciations.


    4) You could try an alternate voice (for some languages we provide both male and female voices).





    Thursday, September 6, 2007 6:24 PM