Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Technology

Radio Reporter Who Lost Voice Returns To Air Using App Built From Archived Audio (ajc.com) 45

McGruber writes: Jamie Dupree had been a radio reporter from 1983 until the Spring of 2016, when he lost his voice. His official diagnosis is a rare neurological condition known as "Tongue Protrusion Dystonia" -- for some unknown reason when he tries to talk, his tongue pushes forward out of his mouth, and his throat clenches, leading to a voice that is strangled and strained, as it is a struggle to string together more than a few words at a time. Dupree's plight attracted the attention of Rep. Ileana Ros-Lehtinen (R-FL), who went to the floor of the House and delivered a speech that publicized Dupree's voice troubles and the lack of answers. Other reporters wrote stories about Dupree and people inside his company, Cox Media Group, tried to find a high tech solution to get him back on the air. They eventually found a Scottish company named CereProc which agreed to sift through years of Dupree's archived audio to build a voice -- which, when paired with a text-to-speech application -- would sound like Dupree and get him back on the radio.

Dupree writes that the app works and will allow him to "talk" on the radio again. Starting next week, he will again provide stories to news-talk radio stations and be back on the air in hourly newscasts.

This discussion has been archived. No new comments can be posted.

Radio Reporter Who Lost Voice Returns To Air Using App Built From Archived Audio

Comments Filter:
  • Technology that gives people some form of their voice back is great - and something I hope can continue to be refined, improved (maybe eventually develope a device that is small enough to put in the body that does the work that external devices currently do?). ALS Ice Bucket Challenge co-founder Pat Quinn lost his voice last year due to ALS, and worked with a group called Project Revoice - who used what limited audio they could find to make it so he sounded less like a "Stephen Hawking on steroids" and more like his old self, and I have to say, it is amazing - not perfect, but perfect enough where he can communicate not sounding robotic, but like himself. Hopefully, these organizations can keep going, as they provide a great service to these people, IMO.
    • by Jason Levine ( 196982 ) on Monday June 11, 2018 @02:53PM (#56767330) Homepage

      My wife's grandmother recently passed away due to ALS. It attacked her throat and severely impaired her ability to eat or drink. Towards the end, she was getting all of her "nutrients" via a sponge squeezed in her mouth, hoping that enough went down her throat and not out of her mouth or into her lungs. On the speech front, her voice went from a slight slur (which might have made you think she had a stroke) to full fledged "mumble talking." If she was trying to tell you "It's on the top shelf", all that would come out of her mouth would be "iiiii ooooo eeeee fffffff." Needless to say, this was very stressful for both her and us. She knew was she was trying to say but couldn't express it and we couldn't understand her. She would use a writing pad to write down what she wanted, but it wasn't perfect either.

      ALS is a horrible disease. You slowly lose control over your body while your mind continues to function. Eventually, you become trapped in your own body. I wouldn't wish that on my worst enemy.

      • Why would they bother with something as imprecise (and frankly, dangerous) as a nutrient-soaked sponge instead of just implanting a gastric feeding tube ("G-tube") with inflatable-cuff feeding port? Attach the syringe to the port, squeeze in the liquefied food over the span of 10-20 minutes, done. Repeat for each meal.

        For some reason, feeding tubes freak people out and seem scary, but they're really no big deal. If someone can't swallow food, a G-tube is vastly better than almost any conceivable alternative

    • by sconeu ( 64226 )

      They've been doing this for years.

      My wife passed away in 2013 from ALS. We were going to "bank" her voice for her eyegaze text-to-speech computer. Unfortunately, her condition deteriorated rapidly, and we were unable to bank her speech before she lost the capability.

      But it's nothing new.

  • does his family get royalties from his voice when he's long gone?
    • does his family get royalties from his voice when he's long gone?

      Hm. Excellent question. I'd say yes ... unless there's something in the app EULA that says "all your words are belong to us." If the latter, I'll get some popcorn and watch what the courts do.

    • Based on archived audio which has an automatic copyright. The voice model should be considered a derivative work under copyright law. So I assume the family would.

  • Your tongue gets in front of your eye teeth and you can't see what you're saying!

  • Is there any good text to read or corpus of sounds to use to make a recording of ones own voice for such a use in the future? I would be curious what one should record of themselves as a basis for ensuring this could be done for them should something tragic like this happen in the future.
  • I bought Cereproc's "William" voice for my (Android) phone years ago, and still use it as my primary TTS engine.

    I couldn't STAND Android's only (at the time) included TTS engine, which (to me, at least) sounded like a stoned, prepubescent elf of indeterminate gender. TTS with a posh British accent was a million times better.

    Personally, I'd *love* to see Cereproc do a pair of voices (one male, one female) that sound like someone who grew up in Britain with a vaguely-posh BBC-like accent, but have nevertheles

    • as well as absorbed the nuances of American highway-number names (eg, "I-595" = "I five-ninety-five", "405 Freeway" = "Four-oh-five Freeway", "State road 836" = "state road eight-thirty-six",

      Just curious, in Britain would it be "I-four hundred and five"?

      Coming from a family of mathematicians and engineers (of which I am not yet pretend to understand the math jokes they make), I understand the annoyance of hearing O instead of zero. Would "405 Freeway" be "four-zero-five freeway" or ""four hundred and five freeway"?

  • Have you heard that new WaveNet TTS generator from Google? It is damned impressive, and significantly less robotic than the voice that was presented as the results of CereProc attempting to transform his voice into a TTS engine. Even if he uses CereProc for day-to-day speech and dubs over it with WaveNet before the clip airs, it would be significantly more impressive.

  • which, when paired with a text-to-speech application -- would sound like Dupree and get him back on the radio

    Soon they may have to treat his Carpal tunnel syndrome.

  • In the last few years of Roger Ebert's life, CereProc, a Scottish company, assembled a synthetic version of Ebert's voice from his DVD commentaries. Although Ebert did all those TV shows, they couldn't use the voice tracks from those because there was always too much (unmatchable) background noise. CereProc said at the time they were hoping to be able to provide an inexpensive web service that could create a good-enough synthesizer for others who had lost their voices.

    From Scientific American, in 2011: http [scientificamerican.com]

  • Can we do this to resurrect the voice of Anthony Bourdain...? Just curious...

  • Now that we have the technology to create edited continuous speech from previous voice recordings (not that it hasn't already been done before) how long before fake, incriminating audio recordings become more common? The FBI can whip one of those up and then interview a subject and charge them with perjury because their answers don't jibe with the concocted phone recording they have. Combine that with the current practice of the NSA sharing their surveillance with law enforcement and cops using that informa

Enzymes are things invented by biologists that explain things which otherwise require harder thinking. -- Jerome Lettvin

Working...