What do I use AI character reference art for? Mostly as reminders of what the character looks like when I write a scene with them. Once I have completed a character prompt that will generate the same recognizable character at least 9 times out of 10, I start putting them in different situations. Minnesota lake. City street. Forest. Or I want different emotions. What do they look like when happy, sad, angry, or curious? Some models won’t generate negative emotions at all, while others will go the whole horror route if asked to. I recently started using the JuggernautXL 5 model that generates darn near realistic pictures. Very nice for character references.
AI art can give you really good random images. Or it can give you the same character again and again in different situations if you devise the prompts right. I’m writer and a coder, so the prompts come fairly easy to me. I just describe the character I want to see and hit generate. Maybe do it a few or a few dozen times to see what the AI model tries to default to. And then I tweak the prompts. Sometimes a synonym will work better. Sometimes just changing the order of words. Earlier prompts tend to have more weight than later prompts. Most of my prompts are under 20 words. Though the latest Jasmine prompt is 48 words. I try to make my character prompts generate the same recognizable character at least 9 times out of 10, and I usually succeed at that.
When it comes to AI art, I have come up with a system that generally seems to work to get good results. I start with a scene, add a basic theme, and then the character prompts. Since I mostly do character references, that is usually just one character. Working from a previous example, I would start with a Minnesota lake. For a theme, let’s say it is nighttime, with a campfire. Then I add a prompt for a happy college woman with long blonde hair and big blue eyes. Maybe add a yellow sundress. Separate the prompts with commas, give it to the AI, and see what happens. Add the right focusing prompts and you might see Betty at night next to a campfire with a Minnesota lake and a bring moon rising in the background. Sounds like a nice place to be, huh?
One of the issues with AI art is they are kinda like idiot savants right now. Even the best can go down rabbit holes of crazy if you aren’t careful. And some do better at some things and worse at others. The key to getting good results from the various AI art generators is practice. Throw out some prompts and run a few generations to see what you get. If you don’t like it, modify your prompts and try again. There is a learning curve every time you try out a new model as well. Some are really good at showing emotions, some don’t emote at all. Find one with a style you like, and give it a go. You never know what you’ll get. For my uses, I create pictures of the characters in my stories so I can use them as references. It works really good for that.
The new AI art generators are amazing. They can take basic prompts and give you an image. Do I want a blonde-haired girl with blue eyes standing next to a Minnesota lake? I can ask for that. And it will generate something. I might like it. I might hate it. If you have a powerful enough computer you can run an AI model on your local computer. If you don’t, you can go many places on the web where you can try them out. Some are free. Some are paid. And some are a combo of both. I use the free side of dezgo dot com for character references and get good results that are awesome for putting up on the computer while writing a scene that features that character.
Forge of War on Amazon
Angel Flight on Amazon
Angel Strike on Amazon
Angel War on Amazon
Wolfenheim Rising on Amazon
Wolfenheim Emergent on Amazon