Festvox simplifies the creation of artificial voices by offering a methodical and well-documented approach.
The Festvox project has recently introduced two new synthesis techniques, multisyn and HTS, with the aim of making the process of building synthetic voices more systematic and well-documented. This software has been designed in such a way that anyone can use it to create a new synthetic voice.
To utilize Festvox, a Unix machine is required; however, this does not necessarily mean that the software is exclusive to Unix, as there have been no attempts yet to port it to other platforms. Additionally, the software requires the use of speech tools programs and Festival itself at various stages in building voices, as well as for the final voices. Festival and the Edinburgh Speech Tools can be obtained from the official websites.
It is highly recommended that users compile their own versions of these tools, as the libraries and include files are necessary to build some programs in the Festvox package. The University of Macquarie's Speech Hearing and Language Research Centre provides labelling tools for speech databases, known as EMU Labeller. This software is utilized for viewing speech as spectrograms, F0s, phone labels, and more. Other labellers/viewers are also available, but Festvox includes support for EMU Labeller, as it meets their requirements and is freely available.
It is important to note that building a new voice requires a significant amount of patience and understanding, as it is a lengthy and tedious process that often involves repetition due to errors or issues. Even with great care, there is still the possibility that a newly created voice may not work as intended. By distributing this documentation, the creators hope to increase general knowledge about synthesis and encourage others to improve upon the software to make it easier and more reliable in the future.
The latest release of Festvox includes several new features, such as better clunits general voice support, support for CMU Sphinx and SphinxTrain to build acoustic models for labeling, the addition of a DOCBOOK version of the documentation with more general background information, initial support for Mac OS X, and configure support to match Edinburgh Speech Tools. Overall, Festvox is a powerful tool for creating synthetic voices and is readily accessible to anyone with a Unix machine.
To utilize Festvox, a Unix machine is required; however, this does not necessarily mean that the software is exclusive to Unix, as there have been no attempts yet to port it to other platforms. Additionally, the software requires the use of speech tools programs and Festival itself at various stages in building voices, as well as for the final voices. Festival and the Edinburgh Speech Tools can be obtained from the official websites.
It is highly recommended that users compile their own versions of these tools, as the libraries and include files are necessary to build some programs in the Festvox package. The University of Macquarie's Speech Hearing and Language Research Centre provides labelling tools for speech databases, known as EMU Labeller. This software is utilized for viewing speech as spectrograms, F0s, phone labels, and more. Other labellers/viewers are also available, but Festvox includes support for EMU Labeller, as it meets their requirements and is freely available.
It is important to note that building a new voice requires a significant amount of patience and understanding, as it is a lengthy and tedious process that often involves repetition due to errors or issues. Even with great care, there is still the possibility that a newly created voice may not work as intended. By distributing this documentation, the creators hope to increase general knowledge about synthesis and encourage others to improve upon the software to make it easier and more reliable in the future.
The latest release of Festvox includes several new features, such as better clunits general voice support, support for CMU Sphinx and SphinxTrain to build acoustic models for labeling, the addition of a DOCBOOK version of the documentation with more general background information, initial support for Mac OS X, and configure support to match Edinburgh Speech Tools. Overall, Festvox is a powerful tool for creating synthetic voices and is readily accessible to anyone with a Unix machine.
What's New
Version 2.0: N/A