Prerecorded sound snippets

[Hynek Hanke]

On 1.8.2010 21:24, Peter Grasch wrote:
> Because the main user of this
> application will be elder people we were thinking of using prerecorded sound
> snippets because "computer voices" are likely to put them off. Of course we
> still need a TTS system for the dynamic parts. Therefore, we'd need a
> combination of pre-recorded snippets and TTS.
>    

We have so far not investigated how to solve such a thing,
I can however propose some solutions:

1) Save the pre-recorded messages in the sound-icon library
and use them as sound-icons with names like time-announcement
etc. In Speech Dispatcher, you can mix these sound icons
with synthesized text in blocks of messages and the synthesizer
will play them one after the other.

This is a little hackish, but should be quite easy to setup. It is
necessary to know which synthesis engine do you plan to use.
Sound icons are supported in Festival and eSpeak.

2) Send the whole text as is to the synthesizer and rely on the
synthesizer to use pre-recorded messages where available for
the given parts of the text, if supported by the synthesizer. I think
Festival doesn't support it now but it would not be too hard to
add it.

3) A really nice and clean solution would be through the SSML
audio tag, but this would require additional support on the part
of the synthesizer:

http://www.w3.org/TR/speech-synthesis/#S3.3.1

I think Milan Zamazal, the developer of festival-freebsoft-utils,
will be able to add more.

Best regards,
Hynek Hanke