Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Books AI

Should Audiobooks Be Narrated by AI? (publishersweekly.com) 113

"Proponents of AI audiobook narration tout its much lower production costs (compared to a traditional recording of a human narrator) as a way to improve profitability of audiobooks," reports Publisher's Weekly, "as well as allowing publishers to publish more audiobooks that have limited audiences.

"But according to actor and narrator Emily Lawrence, cofounder of the Professional Audiobook Narrators Association and president of its board of directors, 'It's very easy to reduce this issue to dollars and cents, but it's very complicated and nuanced.'" If AI narration proliferates, "it's not just narrators who will lose their jobs," Lawrence said. "There's an entire ecosystem of people who rely on audiobooks for their livelihood. People who direct audiobooks, people who edit audiobooks, people who check audiobook narration for word-for-word perfection against the manuscript.... Similarly, in audiobook narrator Hillary Huber's view, the negatives of AI outweigh any positives. She places "loss of livelihoods, loss of integrity in storytelling, and loss of personal connection" high on her list of concerns. "The only pros I see are financial," she said. "And it's the other team that benefits, not the narrators nor the listeners. Do you really think [AI company] Speechki is going to pass their savings on to the listener? No. Listeners make choices about what to spend money on, and they have a right to demand clear labeling of robot voices, as do authors....

[Anthony Goff, who until this month was senior v-p and publisher of Hachette Audio] noted at Hachette... his team is looking at using AI for some titles that have never been produced in audio before — a move that would help ensure that "the largest possible number of Hachette's titles are always accessible in audio format," he said. "Interest in previously unrecorded content would help us make decisions about what would make sense to bring to market as fully produced audiobook editions moving forward, created by a professional narrator and our dedicated production staff." Goff's experimentation lines up with the key point that those who champion AI narration raise: AI can provide publishers with a cost-effective way to produce more audiobooks to help meet burgeoning consumer demand. Industry statistics illustrate the gulf between the number of audiobooks that get to market and the number that could potentially be recorded. According to the most recent data from the Audio Publishers Association, more than 71,000 audiobooks were published in 2020. Though that number marks an industry high, it's still only a fraction of the number of print books published in 2020....

Despite AI narration's potential to help grow the audiobook sector, its emergence is "creating an existential crisis for our narrator community," Lawrence said. "It is not only threatening to take away our jobs and completely remove us from the equation but — and this is my main concern — it's threatening the art that we love. And as a community, we fully believe that what we do is art. Whether I'm out of a job or not, I would be devastated that the art that I care so deeply about is so horribly compromised."

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

Should Audiobooks Be Narrated by AI?

Comments Filter:
  • Let it play out (Score:5, Insightful)

    by timeOday ( 582209 ) on Sunday February 06, 2022 @10:43AM (#62242853)
    I consume plenty of audiobooks, and the narrators - the best ones anyways - are very, very good at their craft.

    It really helps me follow books with numerous characters when they create a persona for each one.

    I love it when they put an unexpected inflection on something, and I think, 'are they making that up?' and then in the following sentence, the prose states that's how the character said it.

    It's fun to get an appropriate accent for a particular work, like Sissy Spacek's reading of To Kill a Mockingbird.

    It's going to be a very, very long time until AI can match all this.

    • {{{ -- It's going to be a very, very long time until AI can match all this. -- }}} --- I agree. However, the wildcard is the cost of the voice actors vs the cost of AI. Ya know, profits.
      • by gweihir ( 88907 )

        {{{ -- It's going to be a very, very long time until AI can match all this. -- }}} --- I agree. However, the wildcard is the cost of the voice actors vs the cost of AI. Ya know, profits.

        Save a penny - lose a million. If they make audiobooks crap by bad voice acting (and that is the only one AI can do), they may just kill the market. Shortsighted greed at work.

        • {{{ -- Shortsighted greed at work. -- }}} --- Greed = profits. There may be a loss of the more discerning listeners. But will that loss be made up by a wider audience of listeners who really don't care that much about, or even would notice, better reading capability?
          • by gweihir ( 88907 )

            {{{ -- Shortsighted greed at work. -- }}} --- Greed = profits. There may be a loss of the more discerning listeners. But will that loss be made up by a wider audience of listeners who really don't care that much about, or even would notice, better reading capability?

            I don't know. I do not think the primary audience for audiobooks are illiterate people or people that just do not want to read themselves. I think it is people that like being told a story instead of reading it themselves and to them the quality of the telling will matter very much.

            But time will tell.

            • by tlhIngan ( 30335 )

              Depends really.

              There are many markets where a computer is good enough.

              E.g., would be disposable reading - think news articles and such. Sure you could listen to the radio, but sometimes long form journalism is more appropriate. Given the amount, this sort of this can be offloaded to a computer to produce an audio version of a newspaper.

              Then there is content created for the hard of seeing - in which case they may just want access to the content of books. For things like non-fiction and such, sure a computer

        • by AmiMoJo ( 196126 )

          AI can already do a pretty good job with inflection and emphasis just based on the predictive text models we have.

          It would also be much cheaper to have AI read a book and make a few mistakes, which could then easily be fixed with some manual hinting.

    • by splutty ( 43475 )

      I have yet to hear a good audiobook from an AI/Speech Synth/whatever.

      They're all, without exception, horrible.

      So unless they're doing something wildly progressive here (which I seriously doubt), it's going to be the same sort of emotionless crap as it's always been.

      • We use Amazon Polly [amazon.com] at my work for training CBTs. Most of the voices are OK, some are not very good. None of them are ready to replace a real human in reading a book. They are good enough to make quick training that is reading off FAA or Boeing information. But that stuff is supposed to "emotionless", so...yeah.
    • It's going to be a very, very long time until AI can match all this.

      Vocal synthesis seems to be coming along quite nicely. It's not all there but I can imagine myself listening to AI generated narration by historical figures. As an example here's six U.S. presidents reciting the lyrics to N.W.A - Fuck Tha Police https://www.youtube.com/watch?... [youtube.com]

      • As an example...

        There's no emotion or nuance in their robotic "speech".
        Not to mention the dumb, machine-like, copy/pasting of noise that the AI confuses for speech - FDR sounds like a pull-string talking doll.

        A proper audiobook narrator ACTS with his or her voice, switching between characters, genders, emotions, speech patterns and impediments... all of that on the fly, sentence after sentence.
        It is a voice acting job, not a simple reading of text.

        Just to understand which character goes with which voice an AI would need to

      • That is much better than anything I have heard before.
    • "I consume plenty of audiobooks, and the narrators - the best ones anyways - are very, very good at their craft."

      Indeed, I doubt that any AI is going to match them any time soon, but perhaps they can replace the bad ones? They do need replacing.

      "It really helps me follow books with numerous characters when they create a persona for each one. "

      Exactly!

      • by jwhyche ( 6192 )

        Omg, no. There are so many YouTube channels now that do their narration by AI voice. I always turn them off and block them. The voice quality is excellent, and they get most if not all the common pronunciations correct now, even the cheap ones. The problem is they read at the same monotone voice and speed. They read the words, but they don't have the inflection that a human reader would bring to the story.

    • I agree. Youtube has a number of channels where the audio is done by AI. The best of them are sort of okay. Most of it, however, is dreadful with a very artificial voice that frequently mispronounce words and does weird things with abbreviations, acronyms, and numbers. Not to mention the whole monotone thing that even the "good" ones still have. If the future of audio books sounds anything like that, their business will die a painful death and deserve it.

    • It's going to be a very, very long time until AI can match all this.

      I can't count the number of times people have said
      something like that, and been proved wrong.
      Still, for fiction I deliberately choose audiobooks by
      my regard for an excellent actor.

      For example, listen to "The Love Song Of J. Alfred Prufrock"
      by T.S. Eliot, and read by Anthony Hopkins.
      It's available on YouTube.

  • and they want [something something] back.

    • The switchboard operators called and they got stuck in an infinite loop in the IVR. Pressing 0 for an operator just hangs up.
  • by WoodstockJeff ( 568111 ) on Sunday February 06, 2022 @10:46AM (#62242861) Homepage

    In my opinion, an audio book should be a "dramatic reading" of the text.

    I've heard "AI produced" audio "renderings", and they sounded just like what they were - text to speech with occasional reaction to punctuation. No nuance.

    • In my opinion, an audio book should be a book in audio form. If I want to pay extra for drama I'd like to have the option.

    • by Potor ( 658520 )

      There's a place for that, but many books are not dramatic. I am a writer, and one of my books was recently turned into an audio book. But it was on the concept of "nothing" in Kant. Drama would have been quite out of place.

      I am not quite sure what is the precise market for such an audio book; I imagine that if you were really listening to my argument while driving, you 'd end up crashing. But I do know a few see-impaired scholars who love them.

      However, to this point AI readings remain too uncanny, if not

      • I am not quite sure what is the precise market for such an audio book; I imagine that if you were really listening to my argument while driving, you'd end up crashing.

        I think I understand. There's so much more time in a day that can be spent listening. All the driving, all the monotonous work an average person does every day. You hit pause when your environment demands more attention, double/triple click your ear piece to rewind a bit. When I'm done I'll wash my ears out with a good sci-fi story, then I'll go over your book a second time and finish it all within the length of time it might have taken me to read your book once.

        For me, I had (pre-COVID) forty reliable

      • There's a place for that, but many books are not dramatic.

        And a good narrator understands that and doesn't try to inject drama where it doesn't belong, but instead understands the text and projects the right tone and nuance for it. There there are some books that are best rendered as flat and as emotionlessly as possible, but not most. Even in non-fiction, it's often valuable to project a sense of interest, or excitement, or intensity.

        Obviously, it's always possible that the narrator's rendering of the text doesn't match what the author intended, or what the rea

    • by AmiMoJo ( 196126 )

      Try asking Google Assistant for a joke. What you get will be corny, but at least in British English the text to speech dose a good job telling it. I wouldn't underestimate what is possible just because there are bad examples out there.

  • I listen to audio books and prefer human over computerized readers/narrators. I lot more than just the words can be communicated by a human narrator; I've never heard that with computerized and AI narrations. The few AI narrations that I have heard wouldn't make a 8 to 15 hour book enjoyable. But some bean counter will make the decision and the variety and nuances that human narrators will become the province of the wealthy and the bane of the middle class. I suppose, as with other things, we will get u
    • Any publishers who try to sell me that crap will get an earful from me, and many others, I suspect. I mean, the same argument could be made that they could save money by printing books on newspaper-quality paper as well, or by eschewing the use of an editor, or cover artist. Obviously, there's a minimum standard that people will accept, quality-wise. I think publishers will quickly find that to be true of audiobooks as well. Don't underestimate the power of consumers' wishes and market competition. It

  • "...people who check audiobook narration for word-for-word perfection against the manuscript..."

    Yeah! We need to fight to maintain the jobs of our finest artisana..wait, what?

    Uh, not trying to sound like a critic, but...well enunciation is not exactly a challenge for AI here, so this seems annoyingly tedious for someone to perfor..I mean do humans actually enjoy that work?

    • by IdanceNmyCar ( 7335658 ) on Sunday February 06, 2022 @11:06AM (#62242911)

      Many humans do. You know there is a whole aspect to historical texts called textural criticism. A single word can have significant meaning. One of the best examples might be in the Bible itself where the oldest manuscripts on a particular event say Jesus was moved by anger to perform a miracle but later manuscripts say it was compassion. The difference between an act out of compassion and anger is significant. Just bring up this issue of manuscripts with most Christians and see why maybe one monkey decided to do this little "favor". Especially with older text these changes can be huge. The original ending of the book of Mark is basically the zombie apocalypse but if you go look in most Bibles, you won't get the true abrupt and gathering horrifying ending of the original text.

      So yes, in terms of artistic nature and historical accuracy, this shit matters and people do enjoy nitpicking at it...

      • PS... fuck auto correct.

        • by splutty ( 43475 )

          Wait. We're not gathering endings? Well damn!

          • Alternative endings can be elegant but we should always question original intent of the author...

            This becomes the significant issue with religious textual criticism. Sense the Bibles "perfect" nature is the product of God's will, we cannot even really ask well without being heretic why Mark felt it perhaps necessary to leave such a dramatic ending. For me, it's literally some of the most poetic of Christian writings because it suggests just how little people believed or understood what was being said by thi

        • PS... fuck auto correct.

          Very apropos for a discussion on AI interpreting what someone wrote. The crude spellcheck interpreted what you were typing and "fixed" it.

          • It gets worse when you consider language. Using a Chinese phone with the Pint in dual English keyboard clearly seems to make it worse...I imagine a James Bond novel narrated by an American to likely fall short in a similar fashion. Culture cannot be ignored in a reading and yet I agree the cost of producing audiobooks is limiting.

            I agree AI readings should be labeled but it may be an interesting question to say how much "value" would be lost in doing so... I think if an audio book is popular enough with an

    • Enunciation isn't the problem. It's understanding the text. It's putting a slightly different voice to different characters. It's knowing when they are being sarcastic, or humorous. So far there's no AI that can interpret a story and pick up cues like that.

      • "...people who check audiobook narration for word-for-word perfection against the manuscript..."

        I was more referring to this specific job.

        I know enunciation isn't the problem. It's more my point. Most if not all audiobooks are recorded with people who have excellent enunciation, which is all the more reason AI would be excellent at double-checking their work, which is essentially all this (human?) job is doing.

  • It is not just a matter of telling the story but it is great for parent/kid bonding.

    • by gweihir ( 88907 )

      Also gives the parents practice reading out loud, which is a useful skill _and_ makes it clear to the kid that reading is something educated people can do.

      • I actually found that I read too fast and stumble when reading aloud. I found this study [nih.gov] about the brain's ability to see patterns, skip letters, etc to increase reading speeds. If your a fast reader, it becomes VERY obvious when reading aloud just what our brains are doing. I can easily get through 50-60+ pages in an hour, a normal 400-500 page book can get consumed in a weekend. I stopped buying actual books many years ago, as I can remember the details of most of them pretty easily.

        This ability REALL
        • by gweihir ( 88907 )

          Hey, no argument. But you can learn a second mode, the slow, dramatic one when doing a performance for kids. It is not either-or.

    • "It is not just a matter of telling the story but it is great for parent/kid bonding."

      My car radio reads to me and I need no bonding with it, it has Bluetooth.

    • Totally agree. Though I don't think this is about kids in general. Does feel like we are one generation away from, "Alexa, read Go Dog Go."

  • by rattaroaz ( 1491445 ) on Sunday February 06, 2022 @10:57AM (#62242889)
    I don't buy that the lost "art" is the major problem. Art changes with time, and that's just normal. Just because computers create art, doesn't mean it's not art. It just means that art is now controlled by computers. If people don't like it, then the demand will be for the human narrators. If they do like AI, then the cost of the human is not worth the value of the art. I do appreciate that their jobs are on the line, and that sucks. I would be very worried, and I get that part. That is a valid concern. The lost art argument seems silly to me.
    • by dfm3 ( 830843 )
      Check out a variety of content on Librivox [librivox.org] and you'll see what they mean about audiobook narration being an art. The range in quality between individual readers is drastic. Some of the best readers can really bring old works to life with their range of tone, inflection, voice acting, and timing. On the other end of the spectrum, some of their contributors are... painful to listen to.

      As an avid audiobook listener, I've come to appreciate a quality production by a talented reader. Eventually you even get to
      • I completely agree that there is a huge difference. And it may be possible that people will pay more to have humans read, rather than AI. In fact, I hope that is the case for now, for the sake of the narrators. I'm simply saying that the argument of the art being lost is silly. It won't be lost if it has value. If it has no value, it will be lost, but then again, is it really art anymore?
  • by msauve ( 701917 ) on Sunday February 06, 2022 @10:59AM (#62242891)
    Automobiles shouldn't be allowed, there's a whole buggy whip industry which will be killed off.
  • Understanding (Score:5, Interesting)

    by VeryFluffyBunny ( 5037285 ) on Sunday February 06, 2022 @10:59AM (#62242895)
    AI doesn't understand what it's reading. It doesn't understand which parts of a text are more significant than others & to emphasise & inflect them appropriately & this is just in non-fiction. When you get to dramatic fiction, do you really want every sentence read with the same emphases & inflections, essentially flattening the story? This is the art & craft of voice-acting, it's difficult, takes a lot of training & practice, & not everyone can do it well. I think customers will notice avoid robo-voices as much as they can. There's a reason Morgan Freeman gets paid as much as he does for narration.
    • Re:Understanding (Score:4, Interesting)

      by freeze128 ( 544774 ) on Sunday February 06, 2022 @12:11PM (#62243099)
      Compromise - The humans that direct and edit audiobooks will still be able to direct and edit the AI, because it doesn't always emphasize words correctly. The fame from audiobooks will shift from the voice actor to the AI director. Voice actors can still get work doing silly voices for animated shows and commercials.
      • If you have to listen to a voice for 6-12 hours, I somehow think that certain special human quality becomes a lot more important. Next time you're watching a decent quality film, listen more closely to the quality of the actors' voices & how it contributes to the dramatic effect of the script & the scenes. There's real art & craft going on there. I very much doubt that AI will be able to match, no matter how much tweaking another human does & the additional workload that it would entail, aud
    • I could imagine the equivalent of a DAW but for speech being developed, where you can assign and automate parameters of the voice. It would enable people without good narrating voices to produce an audiobook, but likely would take _longer_ than having a voice artist record in a studio, much as it would take hours for a producer with a DAW to create a passable guitar part, whereas a proficient guitarist need only sit in front of a mike for a few minutes. Still, it will be interesting to see how AI voices dev

  • by awwshit ( 6214476 ) on Sunday February 06, 2022 @11:00AM (#62242899)

    There is a big difference between a cold reading of words and the kind of story telling a good voice actor can do. AI is not going to match human acting any time soon. Little differences in tone and inflection make a big difference and the how and when to apply them is not mechanical/programmatic.

    I've listened to a number of audio books and the person doing the reading makes a big difference.

  • by EvilSS ( 557649 ) on Sunday February 06, 2022 @11:15AM (#62242941)

    [Anthony Goff, who until this month was senior v-p and publisher of Hachette Audio ] noted at Hachette... his team is looking at using AI for some titles that have never been produced in audio before — a move that would help ensure that "the largest possible number of Hachette's titles are always accessible in audio format," he said. "Interest in previously unrecorded content would help us make decisions about what would make sense to bring to market as fully produced audiobook editions moving forward, created by a professional narrator and our dedicated production staff."

    Wonder if Anthony realized that if AI audiobooks become a thing, no one needs Hachette Audio or any other audio book producer. Traditional publishers, authors, or retailers (*cough* Amazon *cough*) can just do it themselves at that point.

    Personally, I think we are a long ways off from AI being able to replace human book narrators, particularly in fiction. It's more than just reading the text.

    • by gweihir ( 88907 )

      Personally, I think we are a long ways off from AI being able to replace human book narrators, particularly in fiction. It's more than just reading the text.

      Indeed. We may eventually get narrators that do not have the voice for it but everything else and have AI transform the voice. But that is basically it. It still requires understanding and insight and Artificial Ignorance cannot do this and there is actually no indication it may ever be able to.

    • I suspect that that problem is what the lawyers are for; indeed I'm pretty sure that it has already been postured over, if not actively litigated.

      If you aren't dealing with terribly exotic texts, or are OK with lousy output, fairly basic text-to-speech that not even obnoxious industry hype types don't call "AI" will take text and spit out a mostly comprehensible audiobook. Garbage compared to a proper narrator; but costs nothing except a pittance in compute power if you've already got an ebook version ha
      • by EvilSS ( 557649 )
        Today, sure. But those rights are not automatically assigned to the audiobook producers when the book is written. They involve deals with authors and/or publishers for the audiobook rights. If the publisher can suddenly do the audiobook themselves, well, why let a 3rd party company get the money when they can keep it? Today that involves hiring a VO worker, producing and editing the audiobook recording, etc. If it just involves feeding the book into a program, that's something they might not feel like they
  • by DrXym ( 126579 ) on Sunday February 06, 2022 @11:16AM (#62242945)
    I'm not a huge fan of audio books but when I do listen to one the person doing the reading imparts a lot to the experience. They'll emphasize words, change pace, impart character to speech etc. They are understanding the book and the meaning and imparting that in their reading performance. Or at least they should be. A computer isn't going to do that, or if it does it would be a disaster replete with errors.

    Now that isn't to say text to speech doesn't have its place. There are going to be a lot of books that have no audio version and are highly unlikely to ever get one. Books that are out of print, or that were never commercially successful. As such perhaps it is better than nothing but it certainly is no substitute for having somebody actually read the book for real. It might also be useful to condense / precis a book, or even annotate the text with additional information, or produce some kind of "Cliffs notes" version that can read out articles or background info related to the text.

  • The people that think it is an art? Prove it. If it is an art than people will pay up for your work.

    If people refuse to buy your work, then you suck at your art. We should not pass laws to ensure you keep jobs that nobody wants!

    There will always be work. Work comes from two things:

    1) Poor inventors (who then often become rich - but they were never rich when they did the work that created the jobs - if they were, they would pay someone else to do the thing they dislike rather than invent a better way

  • by NeoMorphy ( 576507 ) on Sunday February 06, 2022 @11:23AM (#62242951)
    I listen to audio books while driving, raking leaves, shoveling snow etc. A good voice actor narrator is a requirement, otherwise it ruins even the best books. It's even better with at least two voice actors, one male and one female. It gets weird sometimes when they try to mimic the opposite sex. Using audible for an example, Hyperion was really well done and has a cast of 5 voice actors. The voice actor who did Martin Silenus was amazing!
    • "I listen to audio books while driving, raking leaves, shoveling snow etc. "

      Me too and I actually couldn't do without, these chores are boring.

  • by Hentes ( 2461350 ) on Sunday February 06, 2022 @11:24AM (#62242953)

    If I want to listen to an ebook read by software I can already use a text-to-speech program, in fact blind people do exactly that. Charging the same amount of money for something that I can do on my own machine and claiming it's the same as an audiobook read by an actual narrator is a ripoff.

    • The audiobook publisher can already hire a cheap no experience narrator and put no effort into the production, with no label at all on the finished product. If the book marketplace labels computer generated readings clearly and this opens up more audiobooks, I'm for it.

      Plus there's benefits to having someone at the publisher hinting to the AI about correct pronunciation of many words and phrases. There's plenty of room for quality improvements over general text to speech processing your device has.

      "Argh!"

      • The publishing industry should die. No one should stand between author and reader except their computers. (and those computers must be actually be owned by the ir respecive owners, not by malicious software vendors)..

        Book industry in current form is just a kind of censorship.

  • by RotateLeftByte ( 797477 ) on Sunday February 06, 2022 @11:27AM (#62242963)

    AI is not the solution to Life, the Universe and everything. The sooner those proposing AI as if it is, grok this, the better.

    • by Improv ( 2467 )

      It's not, but automation can reduce costs across the board, and with enough iteration can get to be "good enough" for most use.

      Human-read audiobooks will likely become a rare, premium service.

    • Just put a flag on it in the marketplace, that's more than what we get for low budget productions currently. I wonder how many complainers actually listen to audio books, good narrators aren't going away.

      Some obscure fifty year old book isn't going to get a good narrator or a good production. AI is a good option when the choices are no audio book, or your cousin Lenny takes a work for home gig he saw stapled to a telephone pole.

  • There is always somebody willing to make a product worse in order to earn a few cents more. Compared to writing a book, the effort of reading it out loud is negligible. Hence this is greed, plain and simple.

  • by Dr. Spork ( 142693 ) on Sunday February 06, 2022 @11:31AM (#62242973)

    I feel like I consume more than my fair share of audiobooks, and I have definitely run into human readers who were so douchey that their terrible performances ruined my immersion in the book. The worst is when you can tell by their intonation that they don't understand what they're reading. AI readers now are far worse than the worst smarmy douchebag voice actor, but AI improves fast.

    If some machine learning geniuses got a hold of the problem and used few million hours of transcribed, natural speech as a training set, they might achieve a lot. Another huge improvement would be to hire a human creative director with very high standards and give them easy-to-use tools by which to micromanage the diction of the AI on a granular level. I'm imagining something like a musical clef for the audiobook performance that would precisely determine each syllable's pitch, pace, and other elements of the performance. (The decisions of the human production director would themselves become training data for the AI's prediction engine.)

    Good production directors wouldn't need good voices to direct the AI how to read a certain hard passage. The AI could just extract the essence of the performance data from the voiced examples of the director - kind of like motion capture in CGI video. Companies could even give the authors the courtesy of reviewing the AI audiobook and fixing any remaining mistakes of intonation by simply recording those parts the right way and importing the essence of those recordings into the voice synthisizer.

    What I'm saying is that AI audiobooks will not automate everything all at once. If they're any good, lots of very competent human labor will be needed to guide the AI voice.

  • by q_e_t ( 5104099 )
    Al Murray? All audio books? I mean, great for Al, but surely a bit tiring for the listener. I know he knows a bit of German and French, but does he also know Swahili, Urdu?
  • They should all just be read in the Uli's Moose (old Macintosh popup moose) voice.
  • Why not have both? Why do people have to make things so artificially complicated.

  • I've come across a bunch of AI-narrated Youtube video recently, and they're uniformly awful. MIspronunciations (and plainly no understanding of what they're reading), stilted, monotonic delivery. Ugh.
    And it's in the uncanny valley - it's obvious that it's not a human voice, but a computer trying to be human.

  • no problem with ai reading as long as the voice changes between characters, is pleasant to the ears and price for audiobooks goes down.

  • I used to buy all my technical books for Kindle that have the text-to-speech attribute enabled. This way I could listen on my way to work every day. That is two extra hours of productive time in my day and I went through many books on a very wide variety of subjects. The problem was the software could not handle mathematical equations, programming language source, and many words would simply be mispronounced. Adding AI to this process would allow the software to be actually trained to handle these difficult

  • It's not just Shakespeare either. The industry needs to realize that they're not just employing narrators. They're employing VOICE ACTORS. The best narrators don't just read the book, they bring the book alive. I currently go through 1-2 audiobooks a week during my commute. If they change to AI, I will change to only buying books narrated by humans. When that's no longer possible. I'll use my own software to narrate it much cheaper than the price of an audiobook. I'm not paying 3-6 times more for a voice to

  • Why not have an AI write the books, another read the books, and another AI listen to the books?

    All that's needed is for one AI to pay another AI for the product and to determine what kind of ads can reach the average AI effectively.

    The AIs are probably working on that.

  • I listen to a lot of audio books, it transforms my commute time from total waste to either entertainment or education. The value of that is very high to me - I'd rather drive an old wreck with an audio book playing than the world's best luxury car.

    A good reader makes a tremendous difference - whether its using different voices to indicate which character in a story is speakign, or in lecture, its knowing what parts of sentences to emphesize. Anyone who has had to listen to comptuter text to speech tr
  • I listen to a lot of audiobooks and podcasts, almost always while driving.

    Iâ(TM)ve also listened to a lot of text to speech for the same reason.

    Even after general AI exists, if that day ever comes, itâ(TM)ll be a while before audiobooks can be narrated by AIs are good enough to listen for tens of hours ...

  • We shouldn't let AI choose or write the stories on slashdot.
  • Either the AI will not ever be good enough, in which case what are you worried about?

    Or it will be good enough ... and "There's an entire ecosystem of people who rely on horse drawn carriages for their livelihood. People who raise horses, people who buy and sell horses, people who care for horses, people who clean up after horses."

  • Most commenters somehow assume it is going to be just soulless text to speech, but modern style transfer techniques, trained on a good body of examples, might be able to produce much more impressive quality.

    Some may argue that good quality requires full understanding of text, as well as professional voice actor training. I think that may remain the gold standard for high profile texts, but AI generated audiobooks might soon pass an equivalent of the "Turing test", get better than most of the cheap human n

  • A good audio book reader is like an actor. They are able to understand the story and put emphasis, tone, emotion and whatnot in the text that make the story come to life. This is a universe away from just reading the words. There was a time when AI voices were put under youtube videos a lot and they were *horrible*. Couldn't listen to them even if the subject interested me a lot.
    And let's be real here. An audiobook can be sold many times so the costs of hiring an actual professional to read the text can't
  • I mean... buy the book from Amazon and have a screen reader read the text to you? I know the processing power in a phone may not be enough just now to make it convincing but surely isn't this already solved? Seems like adding an AI-driven text-to-speech to the Kindle app or whatever would solve this problem.

    As for Audiobooks; leave them alone. Continue making Audiobooks the traditional way and we as consumers just purchase them separately. I already do; I own both versions of a number of books because a goo

  • If the AI was good enough, we'd do away with publishing the audio version and just publish the text.
    Pre-rendered AI audio is just silly. Why carry around something that could so easily be recreated from the (much smaller) source text?

  • Take jobs away from people so that CEOs and the ultrawealthy can make more money.

Life in the state of nature is solitary, poor, nasty, brutish, and short. - Thomas Hobbes, Leviathan

Working...