This document provides documentation for the processes and issues in building a new voice in the Festival Speech Synthesis System. This covers the necessary stages in building a voice either in an already supported language or in a completely new language.
Although the task of building very high quality generalized text to speech voices is still very difficult with many open research questions, we believe the building of reasonable quality voices, acceptable for many tasks, can be done with the information provided within this document. Already Festival has been used to implement a number of different languages and voices, including: English (US and UK), Spanish (Castillean and Mexican), German, Polish, Greek, Welsh Gaelic and Basque. These voices were often built in very short times (a few weeks) by people with only minimal knowledge of speech synthesis (e.g. Master's students). Although the quality varies, the results all produced text to speech synthesizers of a quality capable of reading text, like online newspapers, at a level that native speakers can easily follow.
This document, and related scripts and examples, is often updated. You should check the latest status at http://www.festvox.org.
This document specifically offers
Note this document is not a manual for the Festival Speech Synthesis System itself and assumes that the user has access to the Festival system and the Edinburgh Speech Tools. Except where explicitly mentioned constructing voices can be done using these tools alone.
The latest details and a full software distribution of the Festival Speech Synthesis System, and related documents and resources are available through its home page which may be found at http://www.cstr.ed.ac.uk/projects/festival.html
Go to the first, previous, next, last section, table of contents.