VoiceCraft: Open-source neural codec text-to-speech model with voice cloning

Once in a while you come across a paper or example that just leaves you astounded, this is that. I cannot tell the edited voice, it is seamless and whats shocking is that it doesn't care if there is music playing in the background.

There's just so many ways to abuse this but also lot of hilarious/good uses too. Scary times ahead.