Originally published at: https://boingboing.net/2020/03/06/audio-deepfake-jfk-tells-anak.html
…
“We choose to go to the Forest Moon of Endor in this decade and do the other things, not because they are easy, but because they are hard…”
And here I always thought NIXON’s Oval Office recordings were peculiar.
FDR: “So, first of all, let me assert my firm belief that the only thing we have to fear is fear itself, because fear is the mind-killer. Fear is the little-death that brings total obliteration. We will face our fear. We will permit it to pass over us and through us. And when it has gone past we will turn the inner eye to see its path. Where the fear has gone there will be nothing. Only we will remain.”
Why do they keep moving the microphone? Perhaps a bit of normalization would have been in order.
Microsoft’s Azure Speech Service has been able to do this sort of thing for a while. A couple years ago, while I was a contractor at Microsoft, I made a “Hoover Bot” that spoke using a custom voice based on that of J. Edgar Hoover.
It is clever, but there’s no phonetic averaging across the model prior to output. The voice accuracy to JFK comes and goes… there are some dead-on parts and some oddly “off” portions that might as well be the voice of a different person entirely. (Disclaimer: I’m originally from New England, so I assume I’m more familiar with some of the local accent variations than someone who grew up elsewhere in the country might be? This might make it easier to hear some of the finer differences?)
And then there are the audio shifts that make it sound like microphone placement is changing.
It’s impressive, but still not quite at the level that a good human impersonator might achieve with focus. I fear we’re getting closer to the day when those differences will no longer be detectable, other than by using software (which could be coded to misrepresent its findings). Brave new world, eh?
Forget Oswald. Han shot first.
This topic was automatically closed after 5 days. New replies are no longer allowed.