Microsoft’s new AI can simulate anyone’s voice with 3 seconds of audio

Text-to-speech model can preserve speaker’s emotional tone and acoustic environment.
Read more at Ars Technica…