[g-a-devel]gnome-speech Update

[Marc Mulcahy <marc mulcahy sun com>]

Hi All,

Just a couple quick notes on gnome-speech:

Thanks to all the reports about callback.c and callback.h, they're added 
now.  Sorry for the inconvenience.
The FreeTTS driver now can be activated in the standard way (after some 
hackery to ensure that the arguments passed to it by bonobo-activation got 
passed back to bonobo from Java).  Not pretty, but it appears to work.
To get the FreTTS driver to build, you'll obviously need JDK 1.4.1 and the 
FreeTTS sources/jar files.  The java binaries (namely java, javac, and 
idlj) must be in your path.  FreeTTS is available at:
http://freetts.sourceforge.net

To configure gnome-speech to build the FreeTTS driver, you'll need to 
specify the freetts-dir option:
./autogen.sh --freetts-dir=/home/me/FreeTTS

Note that the freetts-dir option expects the FreeTTS main directory, not 
the one containing the .jar files.
The FreeTTS driver has the nasty problem of crashing irrepairably after 
three or four speakers are created.  I'm investigating this now with the 
authors of FreeTTS-- since I'm new to Java, it's probably something I've 
done wrong in the driver code.
Note that it seems a bit sluggish on my RedHat 8 system and I'm told this 
has to do with the JMF sound implementation on Linux.  Not sure, but it 
takes awhile to stop seech in some cases.
Also note that creating speakers takes a *long* time (a couple seconds on 
my p1.7).  I'm not sure there's a solution to this (this problem is the 
reason that all drivers now create separate instances of the engine for 
each speaker so you only have to incur that huge time hit once at program 
startup).  there may be some more driver optimizations we could do.  Maybe 
the driver should hold an instance of each "voice" and then should mediate 
the speakers use of those voices.  Not sure.
Regards,

Marc