Midjourney AI, best practices
I messed up sharing some of the Midjourney posts on Instagram, resulting in multiple copies on FB. I’m just gonna leave the duplicates.
Current best guess on how to use Midjourney:
1. Don’t expect perfect, or even reasonable, control. Allow for happy accidents a la Bob Ross.
2. The order you list things seems to be important, so list whatever is most important first.
3. Think in terms of three areas: content, style, and emotion/theme. Tell the engine what you want to see, how you want to see it, and how you want to feel while you’re seeing it.
Look through other people’s prompts for ideas, particularly on the style/how you want to see it aspect. For example, “3D,” “volumetric lighting,” and “cinematic lighting” will all prompt the AI to give your image depth, but will do so in different ways. You can use things like specific camera lenses, gaming engines (like Unreal), artist names (I LOVED playing with Hieronymus Bosch today!!!), years in history, contexts where the image would be seen (like “magazine ad” or “vintage Russian propoganda poster”), portrait/landscape, and “infinite detail.”
The emotion thing actually WORKS. It’s incredible, if not always perfect. I have recently used “black humor,” “transcendent love,” and “allegorical and virtuous” to good effect lately.
Can you use these for covers? Yes, but I’d wait until a few bugs got worked out on the legal side; the engine makes use of copyrighted images to create its images, and I’d hate to create a dozen covers and have to change them all at once because someone sued me to prove a point.