search

mush42 / sonata-nvda

This add-on implements a speech synthesizer driver for NVDA using neural TTS models. It supports Piper
GNU General Public License v2.0
37 stars 8 forks source link

Pronunciation Issues #28

Open Turkrosoft opened 7 months ago

Turkrosoft commented 7 months ago
  1. Individual words/characters, like digits and letters or sentences consisting of one single word, are pronounced in an odd way.
  2. Strange pauses occur in the middle of phrases that are supposed to be read out together, at once.
  3. In Turkish, and possibly other languages as well, certain vowels are long and certain consonants are soft/palatal in certain words. It shouldn't be so difficult to have the AI read out these words as it should.
  4. We should have more neural voices with different emotions and styles, and child voices. It should be easier to find data to train the AI to get neural voices particularly in English. It would also be great to have more information on how we may contribute to the creation of new voices, as well as what certain settings like length scale mean.