(Aerth) (2024)

Image description: A woman is seated on a rocky surface, her eyes closed and her head bowed. The full moon illuminates the cloudy night sky with its soft, ethereal glow. She is wearing a flowing, light-colored dress that drapes elegantly over her form, capturing the serene and mystical ambiance of the moment. The moon’s radiant light casts a gentle illumination, accentuating the tranquil and introspective mood of the scene.

Full Generation Parameters:

Steps: 70, Seed: 8337459843, Model: juggernautXL_v8Rundiffusion, Version: v1.7.0, Sampler: DPM++ 2M Karras, CFG scale: 7, Model hash: aeb7e9e689, Style Selector Style: base, Style Selector Enabled: True, Style Selector Randomize: False

  • @LilDumpyEnglish
    5•10 months ago

    Wtf. This is the first AI generated art that made me stop and admire. Legit thought it was c/traditionalart.

    Kinda scary since I can appreciate it so much. Good job on the prompt and input to get this output

    • @[email protected]OPEnglish
      4•10 months ago

      This work isn’t mine, and the text part is just an image description I wrote for the visually impaired, the original artist is linked in the parenthesis. The actual prompt is under the “Full Generation Parameters:” spoiler tag, and it is noticeably missing any prompt at all. I’ll have to ask OP, but think this was made with a lot of inpainting which is why there isn’t a text prompt in the generation parameters.

      • @LilDumpyEnglish
        4•10 months ago

        Ahh, that’d make sense. Thanks for the context. I’m interested to hear what the OP says.

        • @[email protected]OPEnglish
          3•10 months ago

          OP Replied:

          Thanks a lot for the appreciation. There was no inpainting, I have only used once inpaiting lately and it was for a Cthulhu image because for some reason it kept growing buildings on his back.

          I use ComfyUI with a workflow I tailor for each image depending on the result I want. It’s part of my pleasure. Civitai has a bad time with complex ComfyUI workflows and doesn’t read well the metadata. Hence the lack of details… I am always happy to answer questions though and share when I can.

          This specific picture was made with a very simple ControlNet Canny, basically a circle where the moon should be, and a badly delimited shape on the left for the goddess, a blob really, sorry Selene. I wanted a specific set of colors, of the kind used in the neoclassic paintings of the XIX, a time with lot of Antique themed paintings (Jacques-Louis David, John William Godward etc…) so I usually do a color palette and I inject it with an IPAdapter to stir the AI to use them. FreeU is also used as it add some twicks I like. No Lora were used.

          Prompt: Painting of Goddess Selene, greek goddess of the white moon, she sits on a rock, thoughtful, foreboding colors, surreal background, extremely detailed, beautiful focus, great light, dynamic dramatic cinematic atmosphere, vivid vibrant color, highly detail, intricate, elegant, very inspirational, located deep clear

          No Negative prompt, SDXL doesn’t need them unless you want to remove explicitly something from a picture.

          If you are interested in the use of those technics, Scott Detweiler on Youtube has very interesting tutorials on how to use them.

          • @LilDumpyEnglish
            2•10 months ago

            Wow. Pretty interesting that there’s no unpainting.