Looking for speech-to-text tool (convert .wav to text)

I have the ability to get .wav files of voice mails emailed to me, but sometimes I'll be sitting in a meeting and I need to know the content of a message without playing it out loud.

Are there any good (and, preferably, free) tools for converting .wav files to text? I know Google Voice has this capability, but I can't determine if it'll work on a file-by-file basis.

I realize that this is a difficult research problem, but even an 80% solution might be workable.

David

Posted 2010-05-25T19:16:06.277

Reputation: 207

Should be tagged speech-to-text instead of text-to-speech. Just a heads up.

– Corey – 2010-05-25T19:41:10.057

Retagged as suggested. – sleske – 2010-05-26T11:08:07.000

Answers

I believe the best free speech recognition software is CMU Sphinx. It looks quite mature, though I have not used it. It is however more a research project than focussed on end users, so while it apparently works well, it's a bit of work setting it up; in particular, you will need to train it before using it.

Other software you might want to try:

http://en.wikipedia.org/wiki/List_of_speech_recognition_software

BTW, have you considered just listening to the messages using a small earplug if you're in a meeting. Might be the easiest way :-). Or, just have less meetings...

sleske

Posted 2010-05-25T19:16:06.277

Reputation: 19 887

Look at Julius

mkdir -p $HOME/tmp/ 
cd $HOME/tmp 
if [ ! -f  Julius-3.5.2-Quickstart-Linux_AcousticModel-2011-07-21.tgz ] ; then
    wget http://www.repository.voxforge1.org/downloads/Nightly_Builds/AcousticModel-2011-07-21/Julius-3.5.2-Quickstart-Linux_AcousticModel-2011-07-21.tgz  -O  Julius-3.5.2-Quickstart-Linux_AcousticModel-2011-07-21.tgz
fi 
tar xvpfz Julius-3.5.2-Quickstart-Linux_AcousticModel-2011-07-21.tgz 
echo "It might be installed."

patrick

Posted 2010-05-25T19:16:06.277

Reputation: 11