When it comes to personal privacy and overall security, we often think of passwords, fingerprints, and even our own faces as being the keys that unlock our world, but what about your voice? If someone could perfectly mimic your voice, what kind of damage could they do? If they contacted people you know, could they lie their way into gaining private information about you? Unfortunately, we may soon live in a world where such a danger is real, thanks to extremely powerful — and admittedly very cool — deep learning technology that can mimic your voice using just 60 seconds worth of your speech.
Canadian tech startup Lyrebird’s new voice-copying tech was unveiled this week, and it sounds like something straight out of a futuristic thriller. With a minute of your own speech on file, the company’s complex algorithm can break down what makes your voice unique and the reverse engineer it into computerized vocals that sound just ›like you. That AI voice can then be used with text inputs to get “you” to say whatever the person controlling the software desires.
It sounds insane, and as Digital Trends points out it’s actually a concept that was used in the Mission Impossible films to great effect, but the applications for such technology range from amazing to terrifying. On the cool side, imagine being able to insert your own voice into a video game or VR experience, with a character that sounds just like you, or even having a digital Morgan Freeman clone narrate your own amateur documentary. Pretty neat, right?
Now imagine someone with bad intentions using your voice to call a loved one and ask for an emergency loan or, in a true nightmare scenario, a digitized “parent” calling their child and asking them to meet somewhere. As with any new, powerful technology, keeping Lyrebird’s remarkable AI out of the wrong hands could be the difference between making our entertainment more interesting, and making our lives a living hell.