NPR logo

New Software Can Mimic Anyone's Voice

  • Download
  • <iframe src="https://www.npr.org/player/embed/527013820/527355070" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
New Software Can Mimic Anyone's Voice

Technology

New Software Can Mimic Anyone's Voice

New Software Can Mimic Anyone's Voice

  • Download
  • <iframe src="https://www.npr.org/player/embed/527013820/527355070" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

A company in Canada has come up with away to recreate anyone's voice and get it to say almost anything. It opens up dangerous possibilities, however.

DAVID GREENE, HOST:

OK. So we've talked a lot about fake news. Now we have fake voices. A Canadian company called Lyrebird has come up with software that it says mimics anyone's voice. Here's part of a fake conversation the company created using computer-generated voices of Barack Obama, Donald Trump and Hillary Clinton.

(SOUNDBITE OF ARCHIVED RECORDING)

COMPUTERIZED VOICE: (As Barack Obama) Hey Donald, have you heard of this new technology?

(As Donald Trump) Are you speaking about this new algorithm to copy voices?

(As Barack Obama) Yes. It is developed by a startup called Lyrebird.

(As Hillary Clinton) Hey, guys. I think that they used deep learning and artificial neural networks.

(As Barack Obama) Hillary is right.

RACHEL MARTIN, HOST:

That's creepy. Artificial voices like Siri and Alexa are pretty good, but, let's be honest, they still sound like computer voices. Lyrebird actually samples a person's voice and captures the nuance of the original speaker. Here's fake Donald Trump saying the same sentence three different ways.

(SOUNDBITE OF ARCHIVED RECORDING)

COMPUTERIZED VOICE: (As Donald Trump) I am not a robot. My intonation is always different.

(As Donald Trump) I am not a robot. My intonation is always different.

(As Donald Trump) I am not a robot. My intonation is always different.

GREENE: Oh, my goodness. Yeah. Not good enough to fool anybody yet. But as the technology gets better, the company says that the voices are going to get more natural.

MARTIN: But sampling someone's voice and making them say something they never said obviously raises ethical questions. In a statement on its website, Lyrebird acknowledges the software, quote, "could potentially have dangerous consequences." It talks about the legal and political implications of copying someone's voice.

GREENE: Although, Rachel, it does not address the potential impact on public radio hosts.

COMPUTERIZED VOICE: (As Rachel Martin) This is MORNING EDITION from NPR News. I'm Rachel Martin.

(As David Greene) And I'm David Greene.

(SOUNDBITE OF DAFT PUNK SONG, "ROBOT ROCK")

GREENE: What was - was that supposed to be us?

MARTIN: Who is that?

GREENE: Yeah, not...

MARTIN: (Laughter).

GREENE: There's a lot of work to be done here - not us.

(SOUNDBITE OF SONG, "ROBOT ROCK")

DAFT PUNK: (Synthesized voices) Rock, robot, rock.

Copyright © 2017 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.