Synthetic voices go multilingual
Researchers are working with the BBC to generate high quality computer-generated speech in different languages.
Professor Simon King, Dr Junichi Yamagishi and Research Fellow Oliver Watts have received over £666k from the Engineering and Physical Sciences Research Council (EPSRC) to create high quality, controllable synthetic voices.
The project is called SCRIPT: Speech Synthesis for Spoken Content Production.
Professor King says:
“With new technology, BBC journalists could use synthetic voices to create news broadcasts in languages such as Swahili, Hausa, Somali and Amharic. This would enable them to reach a larger audience in more countries around the world.”
Although the researchers will work initially with the BBC and text-to-speech service provider ReadSpeaker, the team believe that the project has the potential to bring wider benefits.
“Modern media organisations like the BBC or the Financial Times must be able to react quickly to fast-changing world events. At the moment, the cost of maintaining enough journalists to produce up-to-date media content in multiple languages, for broadcast or online, is very high.
“Technology which automatically produces speech from text offers an attractive solution. We aim to produce broadcast quality computer-generated speech in many languages, and to make it cheap and easy to add more languages later.
“The news industry is not the only one that could benefit. If we are successful, commercial companies, charities, all kinds of organisations could find it a cost-effective way to tell more people about their products and services. In developing countries, the availability of affordable text-to-speech could have immense impact, in disseminating information about health or farming issues, for example.”