This is the isiZulu TTS system produced by CSIR in Pretoria. It consists of three tar files:

1. csir_isizulu_buhle_v1.0.tar is the CSIR Voice distribution v1.0

2. csir_isizulu_buhle_database.tar is the voice database (also available as 10 separate tars)

3. csir_festival_files.tar contains HPL Hindi Voice specific festival modules 

All tars should be untarred from the directory where speech_tools and festival are. Here are some notes on them:

1. Voice distribution
The voice will untar into its own isizulu directory at the top level, except for the main scheme file, which will be placed in festival/lib/voices-multisyn/isizulu/csir_isizulu_buhle_multisyn. The database tar(s) will untar into the database directory: isizulu/csir_isizulu_buhle_multisyn/db/lpc

2. Festival Modules
These are extended versions of some of the source files in festival/src/modules/Multisyn offering support for an alternative back_off procedure for diphone substitutions. If you have festival-1.95-beta, you can just untar these on top of the existing files. If you have a later festival version you should merge using cvs or a similar utility, or by hand.
Don't forget to remake festival.

There is also an alternative version of multisyn.scm - multisynCSIR.scm - untarred into  festival/lib/multisyn. This provides the interface to the alternative back_off procedure. The current zulu voice uses multisynCSIR.scm instead of the default multisyn.scm.

Running the CSIR isiZulu System
----------------------------
(a) run festival/bin/festival
(b) ignore the warnings about no voice. First load the isiZulu voice:
(voice_csir_isizulu_buhle_multisyn)
(c) once the voice has loaded you can synthesize by giving the command:
(SayText "text"), or any of the other festival synthesis commands
(d) remember you cannot use capitals like in English. If you do, you will get strange error messages.

Roger Tucker 
Ksenia Shalanova
Feb 2004