Index of /misc/lojban-thing-1.0.3/local

	Name	Last modified	Size	Description
	Parent Directory		-
	0003_jbocradi_123005..>	2006-08-31 18:36	17M
	JCB_RAM_Alex.JPG	2006-08-31 18:38	9.3K
	KB_United_States.svg	2006-08-31 16:51	134K
	KB_United_States_Dvo..>	2006-08-31 16:51	133K
	L4B/	2011-09-28 08:29	-
	NORALUJV.txt	2005-01-10 08:27	790K
	akwavs.zip	2006-10-05 18:07	26M
	analyser.txt	2005-04-16 14:44	23K
	ascii/	2011-09-28 08:29	-
	banli_dizlo.html	2005-02-04 18:35	2.4K
	bnf.233	2006-10-06 00:19	7.9K
	bnf.300	2005-01-11 08:30	8.1K
	cmavo	2005-06-28 04:50	142K
	festvox/	2011-09-28 08:29	-
	generator.pl	2006-08-31 19:23	5.6K
	gismu	2005-06-28 04:50	360K
	grammar.233	2006-10-06 00:19	65K
	grammar.300	2005-01-11 08:30	69K
	hrefgram.zip	2005-06-24 04:20	570K
	hrefgram/	2011-09-28 08:29	-
	kal_diphone_jbo_us1.scm	2006-10-05 17:23	40K
	le_banli_dizlo.mp3	2006-08-31 19:44	3.5M
	lojban_parser_paper.txt	2006-10-07 14:03	24K
	lojbanbrochure.pdf	2006-09-11 19:25	82K
	lojbanistan.jpg	2001-08-11 16:07	115K
	lojbanrock.mp3	2006-10-07 19:37	1.5M
	lojbanrock.pdf	2006-10-07 19:37	26K
	parser.shar.gz	2005-04-16 14:46	75K
	rafsi	2005-06-28 04:50	30K
	zelda/	2011-09-28 08:29	-

Lojban Text-to-Speech using Festival

Intro

The goal is to create a voice for Festival that can speak any valid lojban text. (It should hopefully be able to handle other languages as well using language shifts and other Festival voices. That is the future.)

Requirements

• Festival
• Festvox
• Edinburgh Speech Tools (Debian package "speech-tools")
• a copy of the wavs
• this package, unpacked somewhere permenantish
• make
• perl
• sed
• other standard GNU stuff like rm

Installation

1. Make sure your copy of Festival is in working order.
2. Put this package somewhere permenantish.
3. Install (decompress) the wavs you downloaded into festvox/ljb_diphone/wav/
4. make sure your ESTDIR (Edinburgh Speech Tools Directory) is set properly (for Debian it should be set to /usr) and run make. This should build the timings from the Praat files and generate the pitchmark files from the wavs.
5. Edit ljb_diphone/festvox/lojban_diphone.scm so it reflects your installed path. XXX this should go away
6. power up festival and enter the following commands:
   

   
   (load "/path/you/installed/this/ljb_diphone/festvox/lojban_diphone.scm")
   (voice_lojban_diphone)
   (SayText "coi rodo")
   

7. If all went well, you should have a text to speech going. XXX eventually it should actually install stuff

Making of the Voice

The tools used in making the voice are as follows

Festival

The open-source text-to-speech engine that will eventually speak everything for us.

Festvox

Festival's voice development package. This is used to build various component of the voice database. See below for information on acquiring these two.

Praat

A wonderful tool for maniuplating, analyzing and generally poking at sound files in a phonetic nature.

Perl

used for converting Praat files to something Festival understands.

There's a great wealth of information on Festival's website, but the most important (or at least helpful to making this voice) have been:

• Tutorial on creating a voice, in this case a japanese one
• As we didn't have any EGG signal, this
• A more general tutorial on voice building:
• lojban doesn't need a lexicon, so this was used
• and of course, some info on the synthesizer we're using
• For more information on the project, see the Lojban Wiki

$Id: README.html,v 1.3 2005/05/08 16:37:10 xajbir Exp $
$Log: README.html,v $
Revision 1.3  2005/05/08 16:37:10  xajbir
Corrected incorrectly corrected non-typo, added clarifying note.

Revision 1.2  2005/05/08 16:04:33  xajbir
Fixed typo.

Revision 1.1  2003/03/20 17:32:44  xavier
added

Revision 1.2  2003/03/03 06:46:07  xavier
added more links