UTF8-characters coming as question marks while using generic driver

[Santhosh Thottingal]

hi,
I was trying to use the speech dispatcher generic driver for dhvani
Indic Text to speech system(dhvani.sourceforge.net).

In /etc/speech-dispatcher/modules/dhvani-generic.conf file I have
given  like this
GenericLanguage "ml" "malayalam" "UTF-8"

But somehow the input reached to the text to speech system is only
question marks for the unicode string.
For example the word ??????? (a Malayalam- ml_IN word) is reached as
just question marks in $DATA variable in configuration file.

Following are the lines from the log file.


 Sun Aug 24 17:46:49 2008 [387380]: speak()

 Sun Aug 24 17:46:49 2008 [387408]: Warning: Prefered charset not
specified, recoding to iso-8859-1
 Sun Aug 24 17:46:49 2008 [387516]: In stripping ssml: |??????? |
 Sun Aug 24 17:46:49 2008 [387543]: Requested data: |<speak>??????? </speak>|

 Sun Aug 24 17:46:49 2008 [387575]: Generic: leaving write() normaly
[...]
 Sun Aug 24 17:46:49 2008 [390317]: Sending buf to child:|??????? | 8
Sun Aug 24 17:46:49 2008 [390341]: going to write 8 bytes
 Sun Aug 24 17:46:49 2008 [390376]: read 8 bytes in child
 Sun Aug 24 17:46:49 2008 [390401]: text read is: |??????? |
 Sun Aug 24 17:46:49 2008 [390529]: child: escaped text is |??????? |
 Sun Aug 24 17:46:49 2008 [390563]: child: synth command = |echo
"??????? " > /tmp/dhvani-text && /usr/bin/dhvani /tmp/dhvani-text|
 Sun Aug 24 17:46:49 2008 [390588]: Speaking in child...

I could able to use the python speechd,  same problem is coming. This
is my program
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# Test program for dhvani generic speech dispatcher driver
import speechd
client = speechd.SSIPClient('test')
client.set_output_module('dhvani-generic')
client.set_language('ml')
client.speak("???????")
client.close()

Is there any special configuration I need to mention in the configuration file?

Thanks in advance
Santhosh Thottingal