Re: Making a speech synthesizer
I'm connected to a team here in South Africa who develops text to speech voices for our local languages, and the problem with a lot of these languages are that they're resource-scarce. For text to speech technologies like HTS, you need a lot of data to produce a good quality voice. Typically you have a bunch of text, and then you get a voice artist to read the text out loud, from which (along with a phonemic dictionary and/or some starting rules) you train your synthesis model. However, a lot of our local languages have very little text available to use for the training process. For some of the languages, the SA constitution and the government website are the only textual sources available.
As #2 said, the Espeak route is probably the easiest, although that depends on the phonemic structure of your language. I know for a fact that the Afrikaans Espeak voice was created by two totally blind people here in South Africa, and that voice is really good. However, Afrikaans, being part of the germanic family, has a phonemic structure not too different from other languages for which data could be obtained. Dutch for example, is for the most part mutually intelagable with Afrikaans.
If you have a language quite different from any of the existing Espeak languages, it might be much more difficult, although I'm sure someone connected to the Espeak project would be able to help out.
-- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
