Installing Festival and its dependencies on CentOS is straightforward. Simply use yum to install the festival package:

$ sudo yum install festival

To install Festival and its dependencies on Ubuntu, simply use apt-get to install the festival package:

$ sudo apt-get install festival

With Festival installed, we need to modify the festival.scm file in order to enable Asterisk to connect to the Festival server. On both CentOS and Ubuntu, the file is located in /usr/share/festival/. Open the file and place the following text just above the last line, (provide 'festival):

(define (tts_textasterisk string mode)
"(tts_textasterisk STRING MODE)
Apply tts to STRING. This function is specifically designed for
use in server mode so a single function call may synthesize the string.
This function name may be added to the server safe functions."
(let ((wholeutt (utt.synth (eval (list 'Utterance 'Text string)))))
(utt.wave.resample wholeutt 8000)
(utt.wave.rescale wholeutt 5)
(utt.send.wave.client wholeutt)))

After adding that, you need to start the Festival server:

$ sudo festival_server 2>&1 > /dev/null &

Using menuselect from your Asterisk source directory, verify that the app_festival application has been selected under the Applications heading. If it was not already selected, be sure to run make install after selecting it to install the Festival() dialplan application.

Before you can use the Festival() application, you need to tell Asterisk how to connect to the Festival server. The festival.conf file is used for controlling how Asterisk connects to and interacts with the Festival server. The sample festival.conf file located in the Asterisk source directory is a good place to start, so copy festival.conf.sample from the configs/ subdirectory of your Asterisk source to the /etc/asterisk/ configuration directory now:

$ cp ~/asterisk-complete/asterisk/1.8/configs/festival.conf.sample \
/etc/asterisk/festival.conf

The default configuration is typically enough to connect to the Festival server running on the local machine, but you can optionally configure parameters such as the host where the Festival server is running (if remote), the port to connect to, whether to enable caching of files (defaults to no), the location of the cache directory (defaults to /tmp), and the command Asterisk passes to the Festival server.

You can verify that the Festival() dialplan application is accessible by running core show application festival from the Asterisk console:

*CLI> core show application festival

If you don’t get output, you may need to load the app_festival.so module:

*CLI> module load app_festival.so

Verify that the app_festival.so module exists in /usr/lib/asterisk/modules/ if you’re still having issues with loading the module.

After loading the Festival() application into Asterisk, you need to create a test dialplan extension to verify that Festival() is working:

[LocalSets]
exten => 203,1,Verbose(2,This is a Festival test)
   same => n,Answer()
   same => n,Playback(silence/1)
   same => n,Festival(Hello World)
   same => n,Hangup()

Reload the dialplan with the dialplan reload command from the Asterisk console, and test out the connection to Festival by dialing extension 203.

Alternatively, if you’re having issues with the Festival server, you could use the following method to generate files with the text2wave application supplied with the festival package:

exten => 202,1,Verbose(2,Trying out Festival)
   same => n,Answer()

; *** This line should not have any line breaks
   same => n,System(echo "This is a test of Festival" 
| /usr/bin/text2wave -scale 1.5 -F 8000 -o /tmp/festival.wav)

   same => n,Playback(/tmp/festival)
   same => n,System(rm -f /tmp/festival.wav)
   same => n,Hangup()

You should now have enough to get started with generating text-to-speech audio for your Asterisk system. The audio quality is not brilliant, and the speech generated is not clear enough to be easy to understand over a telephone, but for development and testing purposes Festival is an application that can fill the gap until you’re ready for a more professional-sounding text-to-speech generator such as Cepstral.