Play.ht is a new AI that can simulate voices and dialogues with real accuracy to what it sounds in real life. There is still a lot of robotic voices but the AI is making progress on simulate how humans talk.
I'm kinda surprised about the rise of AI in the last months, a year ago this technology wasn't even talked about but now is everywhere and more and more people are experimenting with it, from images to audio to videos to text.
Now because I'm a an idiot and have a childish humor I fucked around a little and made some samples, hope you enjoy
https://play.ht if you want to go there and fuck around yourself
Years ago I watched some video I think at an Adobe conference where they showed off some voice cloning software that was damn impressive, but the public never got it. Someone higher up probably thought it would be a danger to our democracy(tm) in the plebs hands. Well, you can't stop the endless march of development that way. Later, the average internet joe did get access to a form of this with the neural net generated 15.ai, and though decent it struggled with emulating real human speech. Pretty good at copying the mannerism of fictional exaggerated characters voices, though still you could tell most of the time it was a robot. This sofware here is where the cat is officially out of the bag. If you tinkered you could scramble together something like this if not better with a lot of work and technical knowledge, we've already had some especially good scammers do this. Some guy thought he was speaking to his boss and lost thousands of dollars. But the average person really had no access. Just from the generations op made you can tell this will be trouble and I'm here for it. With a bit of cutting and smoothing those cuts in an editor you can make this shit believable with at least 70 - 80% efficiency while with 15.ai it was ok at doing Glados 30% of the time.
With a bit of cutting and smoothing those cuts in an editor you can make this shit believable with at least 70 - 80% efficiency while with 15.ai it was ok at doing Glados 30% of the time.
I could use Audacity to change the tone and shift (maybe even a little of background sound like in bus or a street) and yeah, you can confuse the fuck out of people. But that would require a lot of time and effort for even a 2 or 3 second clip.
