Hi Sergio,
We got it to work, not with pyttsx python library but espeak. After quite some trials and errors, we figured it's too much a trouble to install libportaudio0(which is a dependency of libespeak, and libespeak is used by pyttsx), which is used for output generated audio directly to sound card. But it is still possible to synthesize the voice with espeak binary and save it as a .wav file in the file system. So I wrote a simple script to feed strings to the espeak binary, save as a file, and play the file right after. Quite a hacky solution but works.