Prompt: tilt-shift photograph of a hellish landscape, demon army marching down a valley, flame and smoke --ar 16:9 --s 75

Up-scaled version

Edited: up-scaled version added

  • MarafonEnglish
    3·
    1 year ago

    This is badass! I’m curious what the last bit is communicating to the AI though… 16:9 I assume is the aspect ratio requested, but what does --ar, --s, and 75 mean?

    • empireOfLove@lemmy.oneEnglish
      4·
      1 year ago

      –ar = aspect ratio
      16:9 defines said aspect ratio of width:height (16:9 is standard computer monitor widescreen)

      –s is “stylize” which controls the realism passes.
      A lower value gives realistic images while higher values generates more cartoons, outlandish, “trippy” kinda images (75 is a very low value)

      • MarafonEnglish
        31·
        1 year ago

        Thanks for the info! Would these modifiers work with Bings image creator? Or only Midjourney?

        Also is there a resource available with other “commands” like these? I’m very interested in being able to give the AI precise instructions as you did with your prompt.

        • empireOfLove@lemmy.oneEnglish
          3·
          1 year ago

          Im not OP, but I’m 95% sure those commands are specific to the midjourney discord bot. I don’t know if Bing passes command args like the discord bot does or if it just ignores them. Worth doing some experimentation though

          • MarafonEnglish
            1·
            1 year ago

            Oh my mistake lol I appreciate your help though, I will definitely be looking into Midjourney soon.

        • M0oP0o@mander.xyzMEnglish
          3·
          1 year ago

          Good question, lets try “tilt-shift photograph of a hellish landscape, demon army marching down a valley, flame and smoke --ar 16:9 --s 75” in bing:

          I would say no, as this was not in 16:9

          • MarafonEnglish
            1·
            1 year ago

            Hey thanks for testing it out for me! I’m going to check out Midjourney then, since I’ve struggled to convey my output requirements to Bing well enough to get the images I desire.

            • M0oP0o@mander.xyzMEnglish
              2·
              1 year ago

              Let me know if you get it to work, when I tried it always told me the discord was to busy and better luck next time.

          • ThelsimOPMEnglish
            1·
            1 year ago

            Not sure about Bing, but you can tell DALL-E in the prompt to use an aspect ratio and it will often try to follow those instructions.
            Maybe something like: Create a tilt-shift photograph of a hellish landscape with a demon army marching down a valley. Flame and smoke fill the landscape. Use an aspect ratio of 16:9 ?

    • ThelsimOPMEnglish
      3·
      1 year ago

      Thank you!
      I see someone else already answered your question. But yes, those two values are Midjourney parameters. I’m pretty sure they’re specific for Midjourney though. The --ar 16:9 is for the aspect ratio (I wanted to see if I could make a wallpaper) and the --s 75 is the stylize parameter.
      A lower value for stylize makes it more true to the prompt and a higher value applies more of Midjourney’s style to the image. The default value is 100, so 75 is actually not a very low number :) I always like to use it to create a slightly more unique image and adjust if necessary if the image doesn’t come out right.

      • MarafonEnglish
        2·
        1 year ago

        Thank you so much for linking me to the documentation! I see that the cheapest plan is $10 a month but I would imagine the quality of the images generated is better than the free alternatives since it gives you much more control via the parameters.

        Me and some buddies like to generate TTRPG battle maps with AI and the bing one is hot or miss so I’ll see if they’re interested in checking Midjourney out.

        • ThelsimOPMEnglish
          2·
          1 year ago

          You’re welcome!
          You get a lot more control about certain things in Midjourney, but it’s not as precise with its prompts as DALL-E or Bing. Personally I don’t mind, I tend to find a compromise in what I wanted and what Midjourney insists it needs to be :) I guess you could try Stable Diffusion if you want total control, but the amount of options and models available there make my head spin. There’s a recent discussion over here about getting started with SD if you’re interested.

          Feel free to ask me anything if you have any Midjourney questions, I try to put my prompts in all of my posts (almost all of them are for this or other AI communities) so feel free to go through them if you want some examples. I’m always happy to help!