Comment by whodunser

It says they trained on 20 hours of a speech corpus subset. Will larger datasets influence the future of TTS?