|
|
|
|
LLSTI uses Festival and Festvox. These are mature, open source tools which have already been utilised by many people from around the world.
However, LLSTI partners have been extending the tools in various ways to allow
even better TTS systems to be produced.
| | | |
Over the last two years or so, LLSTI partners Centre for Speech Technology Research (CSTR) at Edinburgh University have been developing a unit-selection based synthesizier, MultiSyn, for Festival. The early availability of MultiSyn has enabled us to produce a much higher quality of speech synthesis than would otherwise have been possible. MultiSyn has now been incorporated into the latest release of Festival.
| | | |
LLSTI partners Language Technology Research Centre (LTRC) at IIIT Hyderabad have produced FLAN, a generic shell for Morphological Decomposition (MD) and chunking. It can be used as a stand-alone system or as a Festival module. FLAN uses a very powerful FST formalism which allows MD in (we believe) almost all languages.
Data for each language needs to be supplied in the specified format, but no
programming is required. We are keen for anyone interested to review the
framework and tell us if it cannot support a language, and better still suggest
an extension which would allow it to. We are aware that it does not support
Arabic - we think this language is a special case which would best have its own
specific MD module.
| | | |
See the downloads tools page for other Festival modules & tools produced by LLSTI partners.
| | |
|
Top
|
|
