• SparroHawc@lemmy.zip
    link
    fedilink
    arrow-up
    1
    ·
    10 days ago

    There’s a few indicators. First, as others mentioned, there’s the smoke. Then there’s the way he moves; if you watch all of him at once instead of focusing on details, you’ll notice that he tends to start or stop moving all at once, or limbs move without it impacting the rest of him.

    Audio is easier; sometimes there’s a weird buzzy quality to his voice that is fairly distinctive when you notice it. Most voice cloning AIs have that. The occasionally stilted speech patterns are masked by the foreign accent unfortunately.

      • SparroHawc@lemmy.zip
        link
        fedilink
        arrow-up
        1
        ·
        4 days ago

        I think that one is real. The audio has that buzz which initially made me suspicious, but the video seems clean. Natural, subtle movements that affect his whole body, no stilted speech patterns, and the way he rambles about Beverly Hills is something an AI typically wouldn’t produce. The camera moves slightly, and you can see it jump slightly between cuts as well.

        • wowwoweowza@lemmy.world
          link
          fedilink
          arrow-up
          1
          ·
          4 days ago

          I went to the fuckai community and brought up the idea of a community dedicated to conversations like these. One participant suggested a community called slopornot— I love it. Would you be interested?

          As regards the Zappa video, I’m stuck on that his hair doesn’t move… it still could be real… I don’t know… likely is…

          • SparroHawc@lemmy.zip
            link
            fedilink
            arrow-up
            1
            ·
            2 days ago

            I’d be amused enough to join a community like that, but honestly I don’t use social media enough to actually be able to support it in any significant way.

      • SparroHawc@lemmy.zip
        link
        fedilink
        arrow-up
        1
        ·
        5 days ago

        Sora watermark at the top is the smoking gun; Sora is an AI video generator.

        No buttons on the guy’s shirt.

        Buzzy voice quality, audio is crisp and lacking background noise despite being outdoors with no visible mic.

        Surfaces on face and shirt are weirdly smooth, and lighting is too even.