Segfault on Ubuntu 10.10, 11.04

[mrmekon]

Situation:

import pyttsx
pyttsx.init()
Segmentation fault

Cause:
In pyttsx/drivers/espeak.py, in init(), self.getProperty('voice') results in a call to GetCurrentVoice() in _espeak.py, which calls espeak_GetCurrentVoice() in libespeak.so.1. libespeak is returning a valid result with an empty name. EspeakDriver._defaultVoice gets set to None. A subsequent call to self.setProperty() references the None value, and causes a segmentation fault.

Possible solution:
I know nothing of espeak or its API. Mine works, but does not return a 'current voice'. This solved the problem on my machine:

Replace line 37 of espeak.py:

EspeakDriver._defaultVoice = self.getProperty('voice')

With this block:

EspeakDriver._defaultVoice = self.getProperty('voice')
if EspeakDriver._defaultVoice == None:
    EspeakDriver._defaultVoice = "en"

This just defaults to english if espeak doesn't return a voice. Text-to-speech works as expected after this modification.

Version information:
Ubuntu Server 10.10
Python 2.7
pyttsx-1.0
libespeak1 version 1.44.05~really-1.44.04-0ubuntu1

[rodrigob]

Same problem in Ubuntu 11.04. Would be nice the have this patched and feed into pypi

[parente]

Forgot this was open. Will merge fix soon.

[rodrigob]

the posted fix did not work on my Ubuntu 11.04

[parente]

@rodrigob the fix I just committed seems to get pyttsx working for me on Natty. Let me know if it works for you.

I do get a large batch of ALSA/Jack errors when espeak initializes, but it happens whether I'm using pyttsx or the espeak command line program directly. I think it's a consequence of running in VMWare Fusion.

[rodrigob]

the test now runs fine on my Natty machine. Thanks for the fix !

pyttsx_git/tests/manual$ python run.py

[brousch]

Can you upload this fixed version to PyPi?

[juanantoniofm]

Same problem on debian 6.
espeak = version 1.46

[parente]

v1.1 with this fix pushed to pypi (finally).

[marcobaturan]

Hi!
I try use a module Chatterbot mixed with Pyttsx for do it speak the chaterbot, but the loop While cut de pronunciation of word chains, example: chatbot write: hello, computer only say "he". The example block code is:

# chatbot empiieza' while True: speak = raw_input('- ') response = chatbot.get_response(speak) # dice la respuesta print response engine.say(response) # mantiene el TTS enn marcha engine.runAndWait()`