[ma2rten]

I think this dataset is mainly for speech recognition and not text to speech. Speech recognition should be able to recognize as many different accents as possible.