Lyrebird Can Listen and Copy Any Voice in One Minute

lyrebird-can-listen-and-copy-any-voice-in-one-minute photo 1

Of the many classic movies released that put technology to good use, one stand out example is Sneakers and the "My voice is my password, verify me" entry control system. Copying that voice to gain entry to a secure area required a concealed audio recorder and a fake date. Sneakers is now 25 years old, and speech synthesis technology called Lyrebird just rendered its voice passwords useless.

Lyrebird is a new generation of speech synthesis technology developed at the Montreal Institute for Learning Algorithms (MILA) lab at the University of Montreal. It's impressive claim is the ability to perfectly copy a voice given just one minute of spoken audio. After that, the artificial version can be made to say anything just like the original person.

As the examples below show, the artificial voice is not perfect. However, it's very close, and close enough that anyone listening to the speech will instantly know who it is meant to be.

Related

  • Amazon Offers Echo's Voice Recognition Tech to DevsAmazon Offers Echo's Voice Recognition Tech to Devs

So while it's not perfect, it's very good considering you are hearing the result of several months work and development is continuing. The Lyrebird team also claim they can generate 1,000 sentences in less than half a second using their GPU clusters, and that designing your own unique voice is already possible.

Ethics inevitably comes up, and the Lyrebird team acknowledges that their technology could be used for, "misleading diplomats, fraud and more generally any other problem caused by stealing the identity of someone else." That is why they have chosen to release Lyrebird publicly and made it available for anyone to use. By doing so, everyone is aware of it and that audio recordings will soon no longer be a reliable source of evidence.

Recommended stories

Relax Lyrebird Can Listen and Copy Any Voice in One Minute stories

More stories