Donny Kavanagh has released an AGI file to allow text to speech from Festival to be used with Asterisk without patching Festival. He also asks for help with Sphinx (speech to text). Here's his post:

I'm not sure if this is a bug, or if I'm doing something wrong but I followed the instructions on the wiki for festival, I installed 1.4.3 applied two patches, one for it not compiling on fedora core2 (newer gcc) as well as one patch to make it work with cmd Festival. This seemed to work well however, it only works when called via the zaptel card. When I call via sip, I would see the Festival cmd execute on the cli, but I heard nothing, and the sip call hung up.

I ended up writing an agi myself which I believe will run on an unpatched copy of festival. It also works fine on both versions I have on my system 1.4.3 as well as 1.95beta.

As a side note, perhaps it would be a cool idea to be able to map agi's to be functions within extensions.
Note that this AGI scsript requires the Asterisk perl library. You do not need to start the Festival server for this to work!

Here's an example snippet for your extensions.conf:

exten => 1234,1,Answer
exten => 1234,2,AGI(festival-script.pl|Welcome to the wonderful world of Asterisk! Your phone number is ${CALLERIDNUM}.)
exten => 1234,3,Hangup

festival-script.pl (to be placed into /var/lib/asterisk/agi-bin/):

#!/usr/bin/perl
#make a tts dir inside your sounds dir.
#adjust the t2wp variable to point to your festival bin directory
use Asterisk::AGI;
use File::Basename;
use Digest::MD5 qw(md5_hex);

$AGI = new Asterisk::AGI;

my %input = $AGI->ReadParse();
my ($text)=@ARGV;
my $hash = md5_hex($text);
my $sounddir = "/var/lib/asterisk/sounds/tts";
my $wavefile = "$sounddir/"."tts-$hash.wav";
my $t2wp = "/root/www.cstr.ed.ac.uk/download/festival/1.95/festival/bin/";

unless (-f $wavefile) {
open(fileOUT, ">$sounddir"."/say-text-$hash.txt");
print fileOUT "$text";
close(fileOUT);
my $execf=$t2wp."text2wave $sounddir/say-text-$hash.txt -F 8000 -o $wavefile";
system($execf);
unlink($sounddir."/say-text-$hash.txt");
}

$AGI->stream_file('tts/'.basename($wavefile,".wav"));

--snip-

As for sphinx I've done some research via google, searched the mailing list and I've found very little information on this subject.

Looking at eagi-sphinx-test.c I think I see where the voice detection occurs, however I'm not sure if it just listens for sound, or if it will return a word it thinks its heard, or if we provide a word list of what we are looking for etc. I am just doing some preliminary research before I go ahead and install it.

Has anyone else installed sphinx? Do you have any working examples to post etc, if you know anything share it because there's a definite lack of information at the moment.

Thanks,

Donny

Comments (Click to post)

Name:
Subject:
Website:
Message: