• onnekas@sopuli.xyz
    link
    fedilink
    arrow-up
    6
    ·
    2 days ago

    I think that there is AI “art” that goes beyond typing a few words into chat gpt and waiting for a result.

    I don’t know how popular this is today but about two years ago I watched lots of people go wild with stable diffusion workflows. It was a whole palette of tools: Control net, Inpainting, sketches with img2img for the composition, corrections in Photoshop and so on. It took hours or days of manual work until people “generated” the image that they initially imagined. I would say that this would count as art… Writing one prompt into your favourite llm and take what you get: not so much.

    One example for reference: https://youtu.be/K0ldxCh3cnI

    • SloganLessons@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      1 day ago

      It’s still used a lot. ComfyUI is the software of choice for the people you’re thinking about, and there’s some pretty advanced workflows that blow my mind how anyone even came up with that stuff. The end results are worth it though.

      AI slop that we are used to seeing are people just throwing prompts at chatGPT or Gemini, maybe ask it to change a detail or another if they are feeling less lazy than usual, and then share it on the web