Data Recovery Plans

AI May Be Voicing the Next Foreign Film You Watch – Tech News Briefing

This transcript was prepared by a transcription service. This version may not be in its final form and may be updated.

Zoe Thomas: This is your Tech News Briefing for Friday, October 8th. I’m Zoe Thomas for the Wall Street Journal. Streaming services have made content from around the world more accessible to new audiences. That’s increased the demand for dubbing where an actor’s voice is replaced with another’s in a different language. But the next time you watch a dubbed film, it might not be an actor’s voice you hear at all. It might be a computer-generated one. An artificial intelligence that’s automating movie dubs can do a lot more than just that. On today’s show, our reporter, Ellen Gamerman who covers arts and entertainment joins us to discuss how this AI works, what it’s can capable of, and some of the controversy around it. That’s after these headlines.
Tesla plans to move its headquarters from Silicon Valley to Austin, Texas, where it’s building a new factory. CEO Elon Musk made the announcement during the company’s shareholder meeting yesterday. Musk said the cost of living in the bay area, the distance many workers had to travel, and the company’s growth led him to the decision.

Elon Musk: So just to be clear though, we will be continuing to expand our activities in California. So this is not a matter of Tesla leaving California. As I said, our intention is to actually increase output from Fremont and from Giga Nevada by 50%. We’re just hitting the sides of the bowl. If you go to our Fremont factory, it is jammed. I mean, it is jammed. It’s like, whoa.

Zoe Thomas: Comcast says it’s built a smart TV that’ll be powered by broadband and run on its own operating system. The TV called Sky Glass is part of the company’s efforts to stand out in the streaming wars. The product will be rolled out in the UK later this month. UK customers of Comcast European pay TV unit Sky will be able to use Sky Glass to receive content without a satellite dish. Comcast said the smart TVs will be in other European markets next year. And Twitter is selling the mobile ad firm MoPub for more than $1 billion in cash. MoPub helps app publishers make money by selling ad slots on their apps. Twitter bought MoPub in 2013 for just 350 million and said the sale would allow it to focus more on its own platform. The buyer AppLovin builds tools for app developers to improve marketing and revenue. Its CEO said MoPub would allow it to generate more demand and competition for ad views.
Okay coming up, spoiler alert, that actor is older than his voice sounds. And that reporter doesn’t really speak English. We’ll discuss how AI technology is a voice dub in Hollywood and beyond. That’s after the break. Computer generated imagery has become a regular feature of film and TV, but the use of computer generated sound and specifically voices is becoming more common thanks to advances in artificial intelligence. AI is being trained to dub movies in foreign languages or even change an actor’s voice to make them sound older or younger. It’s also being used in some controversial ways. In the 2021 documentary, Roadrunner about the celebrity chef Anthony Bourdain, an AI generated replica of his voice was used in one scene.

Speaker 3: You are successful and I am successful. And I’m wondering, are you happy?

Zoe Thomas: Those words were written by Bourdain, but the sound was created after his death. Joining us to discuss the future of AI voiceovers and some of the ethical questions it’s raising is our reporter, Ellen Gamerman. Hi, Ellen.

Ellen Gamerman: Hi. How are you?

Zoe Thomas: I’m doing great. Ellen, can you describe to me a bit how this technology works?

Ellen Gamerman: Sure. A lot of it is based at the very beginning on an actual human, but the companies that do this voice cloning as they call it, they don’t even really need that much in some cases and they can take five minutes of your voice and build out an artificially enhanced and new version of your voice so you can’t tell the difference between your real voice and your fake created in a lab voice.

Zoe Thomas: Okay. So we already heard Anthony Bourdain’s voice recreated after his death. What are some of the other things that this technology can do?

Ellen Gamerman: You can use this technology to make a voice sound younger, to make a voice sound healthier if an actor has strained their voice. Deepdub, an AI voice company based in Tel Aviv has a video on their YouTube page that shows a journalist speaking Hebrew, but then seconds later he’s speaking English and what sounds like his own voice.

Speaker 5: (foreign language). And now, let’s see how Deepdub is making me speak in French or Spanish, and without even an accent.

Ellen Gamerman: Another company, Respeecher, a voice cloning company has a video on their website that shows a man’s voice changing into a woman’s voice while singing.

Speaker 6: Another neat feature is singing conversion. Check this out.

Zoe Thomas: Ellen, it sounds like there’s a lot this technology can do, but I’m curious how much TV and movie dubbing does the industry really need?

Ellen Gamerman: Well, the industry will say the sky’s the limit because streaming has exploded so much with foreign content and you can see how some shows from across the globe really take off and even go so far as to win Oscars. So there’s a lot of interest in making these experiences feel more organic so that you’re watching something in your native language and you don’t even really notice that it’s been dubbed. And there’s also technology that’s moving along very quickly that can make the mouth look like it’s actually saying those words of the actor. So the actor could be speaking Spanish, but it will look and sound like he’s speaking English. And if you know that actor’s voice, it will actually sound like that actor, but speaking English without an accent, or sometimes speaking English with a little bit of an accent for realism.

Zoe Thomas: Right. So you’re really sort of seeing and hearing and being immersed in that experience. Is there an example that listeners might be familiar with, any big examples from TV or movies?

Ellen Gamerman: Sure. The Mandalorian, the season finale, it got a lot of attention because it brought back Luke Skywalker who took Baby Yoda with him off. It’s the only person we could trust Baby Yoda with. Luke Skywalker, who’s played by Mark Hamill, and the movie came out decades ago, obviously. So he’s aged quite a bit since then. So everyone knew this is Luke Skywalker, they must have done something with CGI. He’s acting. It’s new material. Wow. How did they make him look that young?

Luke Skywalker: He is strong with the force, but talent without training is nothing. I will give my life to protect the child.

Ellen Gamerman: What people didn’t even really process in their brains was that he sounded young. It wasn’t really his voice. He would say it is his voice because it started with that kernel of his actual voice, but it’s created and they had to go back 40 years, find samples of his voice from audio books and extra material that he had recorded, stuff from Star Wars that wasn’t actually in the movie, just samples that they could easily take from and they built it out.

Zoe Thomas: Okay. We’ve been talking a lot about TV and movies, but are there other places that this kind of technology could be used?

Ellen Gamerman: Yes, there are. One big use of it is video games. When video game developers are making their new products, it can take years and they have stages of demos and demo after demo after demo. And they could bring in actors and put them in a recording booth and do a series of recordings, but that costs money and it takes time. And this is a very quick way to voice something that wouldn’t be out there for huge public consumption.

Zoe Thomas: So let’s talk about some of the controversy around this, because there are some risks to using a fake voice for anybody.

Ellen Gamerman: Right. One of the controversies around this was Anthony Bourdain, as you mentioned. His ex-wife says that she hadn’t given the full permission to use her late husband’s voice. And that gets into basically a concern about the ethics over deep fakes. There is another example of a Canadian voiceover artist whose voice she believes and says she has evidence to prove was pilfered on TikTok and has been used thousands of times to say things she never would’ve said.

Zoe Thomas: Actually, we were able to find several videos scrolling through TikTok that had that voice. Here’s one of them.

Speaker 10: This filter is supposed to show you which celebrity you look like.

Ellen Gamerman: It’s text-to-speech. So you’re typing and the voice says it. And she, the voiceover artist believes that she had done a very big job of hours and hours of reading into a microphone for a translation service that she believes possibly could have been leaked and that’s the raw materials that could have fueled the AI rip off of her voice. And she considers her voice a product and she wasn’t paid for any of the uses of that voice. And she didn’t give her permission. TikTok and its parent ByteDance didn’t comment on the substance of that complaint by the voiceover artist.

Zoe Thomas: So what’s the future for this technology and how widespread is it going to be?

Ellen Gamerman: That’s a really good question. I don’t think anyone really knows. It’s got a ton of potential. And once you start falling down the rabbit hole of these demos and looking on the websites of these AI companies, it’s fascinating to see. For a while they were really obsessed, by the way, with Morgan Freeman. Everybody had a fake version of Morgan Freeman, either publicly or privately to show to people to show off the technology without Morgan Freeman’s permission, by the way. But in any case, the future, I think it could be used in very small ways that we don’t even really notice. You just need one line of dialogue in the recording booth and the movie’s almost going to be released and you just need one little tweak. But I did speak with an entertainment lawyer who was very hesitant about this technology. And just because you can do it doesn’t mean you necessarily will or it’ll take the form that we’re imagining it could.
He had clients over the last year who were asked to give permission in contracts for their voices to be used, AI versions of their voices to be used. And in every case, the client thought it was creepy. They agreed in some cases because they felt they had explicit permission in their contract to reject it if they didn’t like it.

Zoe Thomas: So we’ve been told for years not to trust our eyes. Now maybe we can’t trust our ears either.

Ellen Gamerman: I don’t think you can. Are you really talking to me or are you talking to an AI version of me? How will you know?

Zoe Thomas: Such a good question.

Ellen Gamerman: Or am I really talking to you? Is this really you? I don’t know. It could be. I guess we all have to take our word for it.

Zoe Thomas: Fair point. That was our reporter, Ellen Gamerman. Thanks for joining us, Ellen.

Ellen Gamerman: Thanks for having me.

Zoe Thomas: All right. That’s it for Tech News Briefing this week. Our producer is Julie Chang. Our supervising producer is Chris Zinsli. Our executive producer is Kateri Jochum. And I’m your host, Zoe Thomas. Thanks for listening and have a great weekend.