With its stilted phrasing and flat tone, computer-generated speech is still pretty robotic.
But now the geniuses at Google have developed WaveNet, a program that uses artificial intelligence to produce a natural, human-sounding voice that can even synthesize other audio like music.
RELATED: A.I. Slam Dunks Impossible Game of 'Go'
The technique brings us closer to a world where computers not only understand everything we say, but are able to communicate as naturally and smoothly as a real person.
Some of the best computer-generated speech comes from a method called concatenative text-to-speech (TTS for short), which knits together short, recorded phrases that were spoken by a human and saved to a database.
But although each individual phrase sounds humanlike, the overall dialogue sounds unnatural because the combined phrases can't be varied to represent inflection or emotion.
WATCH VIDEO: Where the Future of AI Is Headed