A bunch of people sent me a link to this Adobe demo of a potential future feature in Audition that allows you to change the words in recorded audio so they say… just about anything.
It’s a whizzy demo, to be sure, but even in this (presumably) carefully chosen demo you can hear weird audio artifacts when the dialogue is changed. I’m not entirely convinced that this technology is quite as good as one might be led to believe by the staged marketing demo—shocker, I know.
Most of the reaction to this video has been to stir up fear that when audio can be altered like this, we won’t be able to trust anything we hear. Sort of how you can’t believe any pictures because of Photoshop techniques.
That’s true, but I’ve got a slightly different perspective: You can fake audio now. If you’ve got even a little bit of isolated audio from a speaker, you can get them to say all sorts of things. You can even change meaning entirely just by clipping words out of an existing recording. I have moved words around many times in order to ease transitions in podcasts that I’ve edited. Nobody notices.
So am I worried about Adobe’s technology? Not really. Yes, we’ll have a rush of people making people’s voices say all sorts of things that are dumb. But if the existence of this feature educates more people about how digital technology makes it easy to fake up just about any kind of digital document, perhaps it’s all for the best. You already can’t believe your eyes or ears—the more people who realize that, the better.
—Linked by Jason Snell