NPR logo

Computer Scientists Demonstrate The Potential For Faking Video

  • Download
  • <iframe src="https://www.npr.org/player/embed/537154304/537174852" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Computer Scientists Demonstrate The Potential For Faking Video

Computer Scientists Demonstrate The Potential For Faking Video

  • Download
  • <iframe src="https://www.npr.org/player/embed/537154304/537174852" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

DAVID GREENE, HOST:

We have an update now from the wild, Wild West of fake news technology. This is a little scary. A team of computer scientists figured out how to make words come out of the mouth of former President Barack Obama on video by using artificial intelligence. NPR's Aarti Shahani explains.

AARTI SHAHANI, BYLINE: If you've been on the internet at any point in the last year, there's a good chance you've come across fake news articles. Well, soon enough we may see a wellspring of fake news videos. As a team out of the University of Washington explains in a new paper, they've made several fake videos of President Obama.

(SOUNDBITE OF YOUTUBE VIDEO, "SYNTHESIZING OBAMA: LEARNING LIP SYNC FROM AUDIO")

BARACK OBAMA: The investigation is ongoing. But we know that the killer was an angry and disturbed individual who took in extremist information and propaganda over the internet.

SHAHANI: Obama did, in fact, say these words about the Pulse nightclub shooter. But the computer scientists could make it look like he said them in places different from where he actually did - could be a room at the White House or during a visit to the Vatican.

It could also look like he said them at a different time - say as a much younger Obama.

(SOUNDBITE OF YOUTUBE VIDEO, "SYNTHESIZING OBAMA: LEARNING LIP SYNC FROM AUDIO")

OBAMA: We are and we will keep doing everything in our power to stop these kinds of attacks.

SHAHANI: The researchers did this not by cutting and pasting his body into different scenes, but by having a neural net, a computer system, study hours and hours of video footage to see how Obama's mouth moves so that computers could then manipulate his lips, contort them to have any word come out. In this experiment, they didn't make up things for him to say. But they did manipulate his mouth to seamlessly cut out entire sentences from speeches he'd given on camera and even to change the order of sentences.

(SOUNDBITE OF YOUTUBE VIDEO, "SYNTHESIZING OBAMA: LEARNING LIP SYNC FROM AUDIO")

OBAMA: Enrollment is open right now but only until January 31. Go to healthcare.gov.

SHAHANI: In that speech as Obama originally gave it, those two sentences were in reverse order, and there were many other words in between. Researchers chose Obama because there's so much high-definition video available of him to study. In principle, the artificial intelligence that studied him could be used to make others lip sync, too. Aarti Shahani, NPR News.

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.