Excellent.  I thought that would be the best way to accomplish this.  Do you think it is reasonable to allow URLs on the Sound= line?  I'm thinking that the speech server should be able to retrieve the sound file so that I'm not forced to manually copy the sounds to each speech server.  Also the text could be optional if all I wanted was a chime.  I know this might be more functionality than most speech servers will implement.  My client side will be user configurable so the user could configure a simple sound name if the user wanted to interoperate with speech servers that do not support URLs.

I guess an alternative is to only use a sound name in the xAP message and then implement a mapping function in the speech server.  Then a simple sound name like 'chime' might be mapped in the server to 'ftp://ha.example.org/sounds/chime.wav'.  Humm... the more I think about this the more I like your implementation where the messaging is generic and the server decides how to locate the actual sound media.  Nothing there prevents me from making a smarter speech server that can download its own sound files.  It makes configuration more complex but keeps the smarts in the speech server.  I'm a big fan of making each controller as smart as is practical.  OK, you talked me into it .  I'll implement simple messaging and a smart speech server.


--- In xAP_developer@yahoogroups.com, James <james.traynor@...> wrote:
>
> Hi,
>
> Actually this functionality is included in the latest xAPSpeech. It
> can play a wav file before speech which can be selected from the xAP
> message as well as a default for all messages. Looks like I
> completely missed updating the schema doc, sorry! I have added the
> extra item to the doc on mi4 and xAP wiki which is just a Sound=
> item in the tts.speak message. Only thing to note is the sound name
> should refer to the sound without the file extension
> As you say this kind of this is best handled in a single message and
> not two. A xAP message can contain multiple blocks/sections but
> will only ever be about 1 class.
>
> hth
>
> James
>
> On 30 Apr 2009, at 13:57, mstovenour wrote:
>
> >
> >
> > I have an alerting application where I need a TTS application to
> > play a wav file prior to speaking the text. There are multiple wav
> > files, from a simple chime to a siren, that accompany the text; this
> > rules out hard coding the wav file into the TTS application. Clearly
> > I'll need to build both a new client and TTS application. My
> > question is regarding the schema to use.
> >
> > Is there a single schema that can communicate both speech and wav
> > files? If not, do you think I should extend the tts.speak to include
> > attributes for intro audio files or should I create a totally new
> > schema for the combination (e.g. tts.alert)? I think it is a bad
> > idea to try sending separate messages because UDP does not guarantee
> > sequenced delivery and it would require a complicated client that
> > knows to wait for multiple messages from different schemas. As an
> > alternative, is it possible to encapsulate two different schemas in
> > the same message (e.g. tts.speak and Playlist.Track)?
> >
> >
> >
>