At the AI conference re:Mars, Amazon presented a new Alexa function. The language assistant Alexa can imitate grandmother.
Alexa can mimic individual voices!
Alexa learns to imitate the voice using short audio recordings. The function was presented by the head of development Rohit Prasad at a keynote at Amazon’s AI conference. He made Alexa imitate the voice of a deceased grandmother. I don’t know whether to find this fascinating or creepy.
Rohit emphasized empathy is important for the language assistant. All the more important at the moment because many people have lost friends and relatives to the pandemic. AI can’t compensate for the pain of those losses, but it can still keep the memory of those who died alive, Prasad said.
Prasad showed that the voice was generated with less than a minute of recorded audio data. This is a breakthrough, as it used to take hours of professional and purposeful recording in a studio to create the voice. You could also use your own voice.
Prasad did not mention how far the function for Alexa has been developed. So it may be a while before Alexa users are allowed to experiment with it.
I personally think it would be cool if the Rock spoke to me. How do you find that? Is it scary or science fiction? Write it below in the comments.