17 comments

[ 3.4 ms ] story [ 52.6 ms ] thread
Oh thank goodness, now we'll always have Morgan Freeman.
Is this original Adobe research, or is there a paper about this somewhere?

  The project is currently in development as part of a
  collaboration between members of Adobe Research and Princeton
  University.
I hope they call it "wife".
(comment deleted)
This is good. So the general people found out how easy it is to 'modified' a speech.

So people will be more careful when making blind accusation based only on audio (speech). Unless what you're presenting is the truth.

Oh you know what I mean..

It doesn't work that way. Go ask someone about autotune or photoshop and you'll be met with a blank stare.

Yes, our little bubble will know, but the vast majority of folks are blissfully ignorant.

This will be used for all sorts, and will only further debt out confidence in reality.

Yes, people vast majority of people will not able to do Autotune or Photoshop.

But the idea of things (image/sounds) could be easily forge is getting widely known. That is good thing :)

We're getting closer the ability to create immortal copies of ourselves. Exciting times!
Actually sounds scary. This has tremendous potential of being misused!
It also has the potential to cause great amusement, just like with Photoshop.
Actually sounds awesome! This has the potential to turn all surveillance of voice communications irrelevant.
We will carry a device that will create hash of what we said using our private key and we will have to speak out that hash to make sentence valid. As time goes by people will learn/evolve to hash their speech in their heads.
I always thought this could be useful in a sort of advanced karaoke, where your nasty voice is processed to sound like the original singer.
ive been waiting for people to make completely new plots for animated movies. pick an actor you like, write a script, remix the movie audio with your new voices.
Does no one think that this could be used for nefarious purposes? Inserting words into a law enforcement interview of a suspect...? Inserting words into a video of a police shooting? Inserting words into historic videos to sanitize old wording for something more politically correct? Or changing what was said in historic videos to minimize the past crimes against humanity?

Imagine an audio-only recording surfacing of Hilter saying that his mother was a Jew.

There needs to be some sort of fingerprinting underlying the audio when created with this tool, akin to the photo forensic tools that can show where a picture has been 'shopped.