Mail Notification Festival

Mail Notification Festival

mail-notification-3.0-festival.tar.gz 2 kB

This small patch adds support for the Festival speech synthesizer to Mail Notification 3.0, http://www.nongnu.org/mailnotify/. Whenever is detects new mail, the origin and the subject of the message will be uttered as follows:

New e-mail from: <name + e-mail address>. The subject is: <subject> You can apply the patch as follows from within the directory extracted from Mail Notification's tar ball:

	patch -p0 < mail-notification-3.0-festival.diff

You can then compile and install Mail Notification as described in the INSTALL file. In addition to Mail Notification's requirements, Festival needs to be installed and running. It is usually included in most Linux distributions or available in precompiled packages, but if it is not your case, it can be downloaded and installed manually:

http://www.cstr.ed.ac.uk/projects/festival/download.html

The 'festival_client' program is used to communicate with a running Festival server. So, 'festival_client' needs to be in the search path, and Festival needs to be running in server mode whenever you want to hear those notifications:

	festival --server

Tips

The Festival server might not accept connections from localhost. If this is the case, replace the line

	(defvar server_access_list '(localhost)

with

	(defvar server_access_list '("127.0.0.1" "localhost" "localhost\\.localdomain")

in the festival.scm file (/usr/share/festival/festival.scm under Fedora).

The best sounding voice in my opinion is "cmu_us_slt_arctic_hts". To set it as default, instead of 'kal_diphone', place it first in the list starting with

	(defvar default-voice-priority-list

in the voices.scm file (/usr/share/festival/festival.scm under Fedora).

Festival uses the old /dev/dsp OSS interface. In order to share the sound card, start it with aoss, esddsp, or artsdsp, depending on which tool you are using to share your sound card:

	aoss festival --server
	esddsp festival --server
	artsdsp festival --server

GNOME Speech

Why not use GNOME Speech? Well, I did try to use GNOME Speech, and the API worked quite well within Mail Notification, but unfortunately, there are two major issues that made it less interesting to use:

It works quite well with the default ked_diphone and kal_diphone voices, but does not work so well with other voices such as cmu_us_slt_arctic_hts. This is most likely due to bugs in Festival, since GNOME Speech uses a lot of functions in addition to the SayText function in order to interrupt or wait for speech.
It supports other back-ends, in particular FreeTTS, but FreeTTS does not have as good a support for voices such as cmu_us_slt_arctic_hts, and it is quite hard to get GNOME Speech working well with FreeTTS because the GNOME-Java bridge has some issues as well.

Nonetheless, when these issues are resolved, GNOME Speech should be used since it abstracts the API of the back-ends, and in this manner lets the user choose his preferred back-end. It also takes care of the synchronization issues among multiple clients.

Last Modified: June 15, 2006 11:11:27 EDT.