Dużo nie drążyłem tematu, ale jak tylko znajdziesz jakiś czytacz IVONA obsługiwany z linii poleceń, to w Pythonie uruchomisz czytanie bez problemu przez popen.
http://forums.debian.net/viewtopic.php?p=162132:
ivonacl - a command line tool and voice server for Ivona TTS System.
IVO Software Sp. z o. o. Copyright (c) 2006-2007. All Rights Reserved.
http://www.ivosoftware.com
Usage:
ivonacl voice [options] wave_file - run in command line mode
ivonacl voice -Dd [options] - run as a text-to-speech server
Command line mode:
--text, -t text - use text given as an argument
--file, -f filename - read text from file, "-" means stdin
The speech is written into wave_file in PCM wave format, wave_file can
be "-", which means stdout and RAW format without header.
Text-to-speech server mode:
-d - run in server mode
-D - run in server mode as a daemon
--port port - set server port number instead of default 9123
Common options:
--load, -l - load voice from specyfied voice file
--otype fmt - output sample type: riff, raw
--version
Voice and speech engine options:
--encoding - input text encoding: iso8859-2, cp1250
--use_devel_enc { 0, 1 }
--dur [ 1, 99 ] - set duration stretch
--vol [ 1, 99 ] - set volume
--read_full_numbers { 0, 1 }
--spell_capitals { 0, 1 }
--read_dates { 0, 1 }
--spell_mode { 0, 1 }
--lts_spec_char { 0, 1 }
--super_spelling { 0, 1 }
--pause_break - set length of pause breaks: 1 - regular, 2 - long
--dictionary - use dictionary file
Tutaj masz małą receptę na Microsoft Speech: http://code.activestate.com/recipes/93025-speech-recognition-in-windows-using-the-ms-speech-/
O ile dobrze pamiętam, to IVONA się integruje jakoś w MS Speech... nie jestem pewien ale wydaje mi się, że można IVONĘ ustawić jako lektora systemowego...