"Our findings demonstrate a dampening effect on perceptual, emotional, and evaluative processing of presumed deepfake smiles, but not angry expressions, adding new specificity to the debate on the societal impact of AI-generated content."
So some of you might remember this post (and the subsequent demonstration on national news) of using a voice cloning tool (AI, Audio Deep Fake) by @racheltobac
(If you haven't seen it, go watch it. Rachel is amazing.)
I'd never needed to do a similar attack before, but! I was just tasked yesterday with researching it.
Asked some friends for a turn-key solution to clone voices. Got pointed to a website. Signed up for $1 a month (first month... then it goes to $5 a month thereafter).
Pulled some audio of my mark down from a youtube interview (a podcast works great too).
Only needed a minute's worth of audio.
Uploaded it to the website for cloning.
Typed out a quick script for the voice to read.
30 seconds later, I had my cloned audio.
It was so good, that it even included natural voice inflections AND!!! verbal pauses like umm's and uhh's that matched the mark's original presentation. I can't tell the difference between the cloned voice and the original person.
Y'all... voice cloning and audio deep fakes are well past the ease of "script-kiddy" level. Anyone can do it.