• Mothra@mander.xyz
    link
    fedilink
    arrow-up
    8
    arrow-down
    1
    ·
    1 year ago

    Serious question though, is it a render or AI? I’m inclined to think render because of the consistency for both front and rear views of the same model, usually AI slips up and makes things somewhat different. Thoughts?

    • stebo02@sopuli.xyz
      link
      fedilink
      arrow-up
      4
      arrow-down
      1
      ·
      1 year ago

      AI can do a lot of things but not this. If you ask an AI to make this it’ll be all kinds of fucked up

      • olicvb@lemmy.ca
        link
        fedilink
        English
        arrow-up
        8
        ·
        edit-2
        1 year ago

        Depends which, ik stable diffusion can. Either feed it a centaur openpose pic or use a Lora.

        Now it would take a good amount of tweaking to get right but it can

        • stebo02@sopuli.xyz
          link
          fedilink
          arrow-up
          3
          ·
          1 year ago

          if you know how to do this I’m actually quite interested to see the results

          (for scientific purposes of course)

          • olicvb@lemmy.ca
            link
            fedilink
            English
            arrow-up
            1
            ·
            edit-2
            1 year ago

            here:

            Made this quickly, you can see the openpose controlnet image and the result

            Prompt: “four legged female, centaur, clothed,”

      • Mothra@mander.xyz
        link
        fedilink
        arrow-up
        5
        ·
        1 year ago

        I can’t say I’ve used AI much. I keep seeing solid AI stuff lately, why would this in particular pose a challenge?

        • stebo02@sopuli.xyz
          link
          fedilink
          arrow-up
          3
          ·
          1 year ago

          idk I’m no expert but generating something that doesn’t actually exist with AI often leads to unexpected/unwanted results. How would you even make a prompt for this? “Woman with four legs”? Good luck with that.

          • Hexarei@programming.dev
            link
            fedilink
            arrow-up
            4
            ·
            edit-2
            1 year ago

            “Centaur but half woman, other half also woman”, maybe? BRB, gonna go consult stable diffusion

            edit: I come back empty-handed, for SD just kept trying to generate women standing near each other. oof.

              • SnipingNinja@slrpnk.net
                link
                fedilink
                arrow-up
                1
                ·
                1 year ago

                See if this prompt works:

                create an image of a human on all fours but the parts above the chest are replaced with another human in standing position
                

                Bing image creator just fails to create an image at all

                • stebo02@sopuli.xyz
                  link
                  fedilink
                  arrow-up
                  0
                  ·
                  1 year ago

                  Your prompt has got to be more descriptive. For example “woman centaur, human legs”. However this prompt is simply generating women on horseback for me. It’s hard for AI to imagine a horse with human legs.

                  • olicvb@lemmy.ca
                    link
                    fedilink
                    English
                    arrow-up
                    1
                    ·
                    1 year ago

                    The prompt will only work if you have a model thats trained/supports centaurs