|
Speech Synthesis and Recognition in Linux
Festival speech synthesis
Software and guides
- Debian:
- festival festival-doc festival-freebsoft-utils gstreamer0.8-festival
- speech-dispatcher-festival speech-dispatcher cl-speech-dispatcher
- speech-tools speech-tools-doc speechd-el
speechd-el-doc-cs libgnome-speech3 libspeechd1
- konq-speaker gnopernicus (screen readers)
- gok gok-doc (optical character recognition)
- dasher (typing by mouse)
- Guide: http://festvox.org/festtut/notes/festtut_3.html
- KDE Text-to Speech system (KTTS)
- TuxTalk (the GNU/Linux operating system for the blind)
- PerlBox
- Speak to me, Linux (Jan 2005 article)
To convert a document to an mp3 file:
- Produce a clean ascii text version of the document (save in OOWriter as txt)
- text2wave < document.txt > document.wav
- lame -h document.wav document.mp3
Note that clitunno can perform all three steps -- especially the second is highly CPU-intensive.
Four modes possible (B, C and D require a sound card):
A. Convert an ascii text file to a wav file:
text2wav < textfile.txt > textfile.wav
B. Kmouth (part of KDE)
Copy and paste the text to be read into kmouth, which starts festival
C. Browser reads to you with konq-speaker (there are also keyboard shortcuts)
Mark text, select Tools | Speak Web Page
D. Command-line conversion to speech:
echo You are feeling very sleepy -- do exactly as I say | festival --tts
D. Festival shell (see man festival for instructions on conf file)
You can set up a configuration file (see /home/steen/festival/ex.scm):
- Start: festival ex.scm
(SayText "This is a pen")
- Set a voice:
(voice_kal_diphone) US Male -- this is the one I have
(voice_rab_diphone) Select voice (British Male)
(voice_ked_diphone) Select voice (American Male)
These voices are in /usr/share/festival/voices/english -- see if you can get the others.
You can also get Italian and German.
Dasher will use festival to speak what you write.
ViaVoice speech recognition
Guide at http://volker.orcon.net.nz/linux/viavoice.html
Start at http://www-3.ibm.com/software/speech/enterprise/te_3.html
I tried (and failed) to order ViaVoice directly from IBM today, from
their commerce
page, which contains links to the following:
Technical Support will be provided by participation in the Discussion
Group. To subscribe to the ViaVoice Dictation for Linux e-mail discussion
list, send an e-mail to join-vvdictator@laser.sparklist.com and follow
the directions.
Check the following website for the Java GUI code in our
latest SDK
Reviewed
in Linux Planet
Tcl/SMAPI
project
Reviews and resources
30 Day Money Back Guarantee
We are looking for people who have used Dictation for Linux for at
least 4 - 7 hours per week as a productivity application. If you want
to help improve the usability of the IBM voice recognition product on
Linux, just fill out our usability questionnaire. To get a copy of the
questionnaire, just send an e-mail note to Art Keller at kellera@us.ibm.com
.
<>11K8437 IBM ViaVoice Dictation for Linux U.S. English (includes Microphone
Headset): $39.95
|
