"Designed to be a state-of-the-art audio editing application VoCO can also generate new words using a speaker’s recorded voice. Apparently the software can understand the quality of a persons voice and replicate it from only 20min of recorded material."
Like that won't be abused to hell and back.
E1: Looks like it will feature detectable watermarking.
Here's a video of the demonstration:
as featured in this ARS article:
It's clearly not as "photoshop" as they'd like, but it's still an early version,