ibmtts output module and utf8

[Lukas Loehrer]

Olivier BERT writes ("ibmtts output module and utf8"):
> I'm using speech dispatc:er 0.6.2 with the ibmtts output module. 
> I am french so I use a french voice. Everything is ok with this 
> speech environment. 
> I recently tried to switch to the last CVS development version. Compilation
> and installation were ok but speech Dispatcher doesn't speak multibyte 
> utf8 characters correctly. Actually, two ISO-8859-1 chars are spoken instead
> of the appropriate accentuated char for example. 
> 
> I compared the ibmtts output module which is in the 0.6.2 version and the
> one in the last dev version. The only difference is that the last calls 
> module_strip_ssml whereas the 0.6.2 version does not.  

Actually, it is somewhat surprising that you were ever able to get
ibmtts to speak utf-8 characters beyond ascii, because as far as I know, ibmtts 
only supports
latin-1 as input encoding for western european languages. Thus, the
ibmtts output module would have to do the conversion from utf-8 to
latin-1. If the input contains unicode characters not encodable in
latin-1, you are out of luck anyway. 

Now, I am not sure why the new version strips out SSML. While I
believe ibmtts does not natively support SSML, it comes with an SSML
input filter (ssmlfilter.so) which seems to be enabled by default. At
least this is what my tests with pyibmtts indicate. Information about
input filters is in eci.ini.

Best regards, Lukas