This is the README file for generating SvxLink sound clips using the MARY
text to speech system (http://mary.dfki.de/). This instruction was written
for version 4.0.0 of the MARY speech synthesizer.

MARY is by far the best free text-to-voice system I have seen. You can
test it or generate sounds for SvxLink directly using the online demo:
http://mary.dfki.de/online-demos.

Start by downloading the latest installer jar file from the downloads
page: http://mary.dfki.de/Download. Install the application and the
wanted voices and languages. Start the installation by running the following
command:

	java -jar openmary-standalone-install-4.0.0.jar

A java JDK need to be installed first of course.

Start the MARY server application. Just go to the MARY/bin subdirectory
and type ./maryserver. It takes a while for it to start. Up to half a
minute on my old 1GHz laptop! The startup time depends on the number of
installed voices.

All sounds are generated from small text files. These are read and processed
by a Ruby script found in this directory. You may first have to install a Ruby
interpretator on your system. Then, to generate all sounds, type:

	./mary4_svxlink_sounds.rb */*.txt

To generate the sound clip for a single existing txt-file, just point out that
single file. Example:

	./mary4_svxlink_sounds.rb Core/online.txt

To generate a single _new_ sound, use the "one_file.sh" script. Example:

	cd Core
	../one_file.sh my_sound "this is my sound clip"

This will generate two files. The pure text go into my_sound.txt and the
generated audio will go into my_sound.wav. If you add new sound clips,
also make an addition to the subdir.cfg configuration file.

Almost all sound clips are generated as they should be but the 2X, 3X etc
sound clips in the Default subdirectory need some manual editing. Use a
program like "audacity" to only save the first part of these clips, the
"twenty", "thirty" etc part. These clips are used to say numbers like 24,
35 etc.

There is one more step before the new sounds can be used with SvxLink. The
filter_sounds.sh script must be used to maximize the amplitude and cut
away unnecessary silence plus gettig the sound clips to the right format
(sampling rate etc).

	../../filter_sounds.sh . /tmp/sounds-mary

Now you can put the sound clips in the right place and start using them.

