• zephorah@lemm.ee
    link
    fedilink
    English
    arrow-up
    21
    ·
    13 hours ago

    So let’s hear it. Hand an AI Leviathan Wakes, the first book of The Expanse, and see how it does. My money is on it being garbage, but let’s hear it.

    Jefferson Mays is tough to beat as a human.

    Worse, hand an AI a Terry Pratchett book, see how that goes.

    • SGG@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      arrow-down
      3
      ·
      9 hours ago

      Hand the AI Twilight books. It’ll either delete itself or make them better?

    • Cort@lemmy.world
      link
      fedilink
      English
      arrow-up
      6
      ·
      edit-2
      12 hours ago

      I only read heard him read the last 3 books+ novellas after watching the show, and he REALLY did the accents well

      *Except Bobbie, not enough southern hemisphere OZ/NZ twang

      • zephorah@lemm.ee
        link
        fedilink
        English
        arrow-up
        13
        ·
        12 hours ago

        Audiobook narrators don’t “read”. They act. They vocally act the entire book. The ones who don’t generally get returned, unread, to either your audiobook platform choice or the library.

        Voice actors in games also don’t just read. They act. They vocally act their entire role.

        Jennifer Hale vs AI, who would win? Would any human other than Kate Mulgrew as Flemeth have made the character as compelling?

        • Uli@sopuli.xyz
          link
          fedilink
          English
          arrow-up
          2
          ·
          edit-2
          10 hours ago

          Technology will of course change, but I did complete an (unreleased) experiment where I made an animation with AI, using AI to provide voices. This was a few months ago.

          All of the models used to generate vocal lines out of nothing are very basic and robotic. But I had a lot of success recording the lines (and songs) myself and then using an AI tool to convert it into someone else’s voice. I blended two or three voices per character and for voices where the character was a different gender or age from me, it sounded like a real actor of that demographic giving the same performance I gave.

          So, context matters here. Is the tech ready to replace actors completely? Not at all. But could you have an actor record all the lines in different styles and then use licensed voice models to have it sound like a given voice actor? Absolutely. Actors should think very hard before agreeing to any licensing agreements using their voices. Because it might just result in a lot of the acting removed from their job role. And potentially worse quality dialogue in the end depending on who they hire to record lines in bulk. Not to mention that it’s only a matter of time before the fully AI models advance far enough to do the job completely.

          Hence, why I never released my animation. People wouldn’t be able to tell whose voices I used. But I would know. And I don’t want to be on the wrong side of history.