Necromancer with a little Anime flare.

  • BallShapedMan@lemmy.worldOP
    link
    fedilink
    English
    arrow-up
    2
    ·
    10 months ago

    I cheat with my prompts. I find an image that’s close to what I want to do, dump it into Chat GPT and prompt it with: Describe this image in extraordinary detail at great length.

    I then alter what it gives me in the places I want it different and paste that back into Dall-E 3. I figure it’s going to know the phrasing it’ll best understand to get an effect.

    The style prompts at the beginning don’t come with the output. But those I got by asking it to describe different styles in the same detail so I can replicate it.

    Only the broken pixel style I use was made completely from scratch. All others started as an output from Chat GPT that I’ve alerted to get a more consistent result and a result that’s more in line with what I see in my head.

    • Mom Nom Mom@nom.mom
      link
      fedilink
      English
      arrow-up
      2
      ·
      10 months ago

      Hahaha - it’s not cheating, it’s using all the tools at your disposal! I simply haven’t learned how to do that yet :-P At any rate, there’s a talent there. Kudos

      We do have an LLM that is fed a prompt (built randomly), which it A) uses to build a better and more descriptive prompt and B) to create a “headline” or title for the image. It is entirely silly 😂

      ¯\_(ツ)_/¯