Happy lunchtime everyone! It seems from my post last night that some people are interested in how I generate my images, so I'll be posting some images here and adding the prompts and what they do.
First up, is a line art style image. I do all of my AI generations through Midjourney, mainly because I like a lot of the styles they support and I have discord on all my devices so I can play with it anywhere I have access to the internet.
The downside to Midjourney is that they have a number of restricted words. You obviously can't do porn, especially since all their images are available on their site, but they also restrict certain levels of violence. Unfortunately, they are also fairly restrictive on anything that "approaches" sexual behavior. This includes wanting images of busty women (either realistic looking or various styles of illustrations). This is annoying because the same doesn't apply to men. I can get a 95% nude, ripped, barbarian man with rippling muscles, but if you want a woman with anything more than a fully covered B-cup, well good luck cause you probably won't get it (unless you know how to finagle the AI, which I kinda do, more on that for #nostrafterdark ). Unfortunately this means that some of the words will bleed into other definitions, like the word "Chest", so you couldn't get a treasure chest for a long time, now you can.
But I'll work through some posts and explain what they mean.
Prompt:
strawberry blonde d&d elf ranger, female, athletic muscular frame, black and white line art with soft shading, long pointed ears, HD, sharp lines, line detail, full body image --quality 2 --aspect 3:5 --v 5.1 --no weapons
1. Strawberry blonde D&D elf ranger, female, athletic muscular frame, full body image, long pointed ears.
These are going to describe the image I want. You'll notice this is a black and white image and I have "strawberry blonde", well, I was just copy/pasting a previous prompt and adjusting the styles. Sometimes describing a color in the prompt will "bleed" into a black and white image. So watch your prompts.
I specify D&D elf because that is a particular style that AI will look at. If you just put "elf" it may not come out as you planned. In this image, it ignored the prompts for elf and long pointed ears. That happens, but since the hair covers the ears, it can be easily ignored.
You can adjust the description of your image in almost any way you can think of. I've read that Midjourney puts higher weights on descriptions earlier in the prompt that later, or that repeating something can increase it's weight. I cannot confirm that for you. There is a weigh to control the "weight" of a part of the prompt and that is by using a double colon "::" and then puting a number after it. Positive numbers increase its importance, negative numbers decrease importance.
Using "::" can have really big effects, so be careful.
2. HD (high definition), black and white line art with soft shading, sharp lines. (line detail was a typo but it seems to have been ignored, so no harm).
These are the styles I want the image done in. I wanted line art with soft shading to give it the feel of a hand drawn image, but with sharp lines so it looks more finished. I could have specified almost any way of illustration or photography. HD wil give it better clarity, as high definition often does in other ways.
I have seen prompts, and made them, where this is a very long section depending on what you want. You can specify specific artist styles as well. If you go back through my images, you'll see a few I labeled as being inspired by Joe Jusko (who did the John Carter of Mars book illustrations). If I add a specific artist the AI will mimic that style.
3. --q 2 (quality 2) --ar 3:5 (aspect ratio) --v 5.1 --no weapons
These are midjourney specific qualifying terms, I'm sure other AI have their own.
--q (quality) is what quality you want. Without adding this, the quality is baseline 1, but you can increase that.
--ar (aspect) is the aspect ratio - so width x height. This can make a big difference. I specified earlier "full body image" so I need a "tall" image to accomodate. so I used an ar of 3:5. 3 units wide, 5 units tall. If you want a scenery piece, then wide may be better than tall. But if you want a tall image and give it a wider aspect, it will cut off the top of the image.
--no - this specifies something you don't want. I put "--no weapons" because the word "ranger" automatically inspires swords and bows. AI weapons mostly suck right now.
--v 5.1 this is the version of midjourney I wanted. 5.1 is the newest and is now standard, but it wasn't initially. And I can regress to versions 1-5.0 easily enough.
The version, as well as other functions, can be accessed using "/settings"
