Comment by lauretas

How does this work? Where does the audio come from? Does it use some database of recorded IPA sounds, or does it convert from text to speech on the fly?