The big criticism of AI art is that by plugging in an artist name, it can quickly copy their style. However, AI art has tons of benefits — as a writer, for example, I enjoy being able to generate pictures of characters and scenes that I can use to supplement my descriptive writing; this is especially useful for me, because I have aphantasia and can’t conjure up mental images.
Truth is, as Stable Diffusion continues to be refined, it is actually becoming more of a burden to use artist names, since they cause it to throw in random artifacts like signatures that are extremely difficult to get rid of. Plus, the artist community is pretty pissed about it.
So, for the sake of ethical AI use, I quit using artist names.
To cover our bases, my standard negative prompt is:
signature, text, fonts, words, watermark, badly drawn, bad composition, boring, collage, surreal, nonsensical, incoherent, ugly, disfigured, deformed, beginner, amateur, asymmetrical, low contrast,
I customize this as necessary for the specific images that I create.
I also include standard at the end of the positive prompt: professional quality, well drawn, attractive
So, let’s see what we can do with the prompt, “Cherry blossom”. Same seed and other settings.
As you can see, specifying medium alone makes a huge difference in the type of images that are generated — no artist names necessary.
As I’ve said before, AI art appeals to me because I think so strongly in words. I enjoy the process of crafting out prompts and put a lot of thought into writing them, and I hope that one day AI art is viewed as a legitimate form of creative expression — alongside traditional and other digital mediums.
Here is a giant list of different mediums that can be used in Stable Diffusion. There are a lot more possibilities that weren’t included, so experiment!
By Autumn Rain 