PHILADELPHIA (CBS) — From Siri and Google Assistant in phones to Alexa at home, many of us are never far away from a computerized voice that sounds pretty natural, all things considered. Now comes a startup that says its innovation can simulate any human voice and make it say anything you want. But the technology raises new and critical questions.
“Sick people, but they love me!”
John Di Domenico does a mean Donald Trump.
This impersonation is nowhere near as polished…
(Lyrebird’s Trump:) “This is huge. They can make us say anything. No, really — anything.”
But consider it’s generated solely by algorithm — and it’s only the beginning.
(Lyrebird’s Trump:) “This is huge. How does their technology work?”
With a minute of recorded speech, Canadian startup Lyrebird says it can learn the ‘DNA’ of a person’s voice and use that to generate any words, even control the emotions of the resulting audio.
(Lyrebird’s Obama:) “The good news is, they will offer the technology to anyone.”
Developers of personal assistants, video games, and text-to-speech solutions for people with disabilities may be interested in this. But so may fraudsters and identity thieves — and worse. Lyrebird itself notes the potential for “dangerous consequences” and says that’s why it’s making the emerging tech public and starting the conversation.