A blog related to original AI artworks. I make AI artworks using coined words and emoji as prompts.
If you enter a word with “🖼” as a prompt in Stable Diffusion 1.5, it will usually draw a picture. However, Stable Diffusion 2.0/2.1 does not draw the picture as you think. Adding “painting” often draws a picture, but it tends to use bold colors. Perhaps, if the spelling includes paint, it means wet with a bright color. So one way to fix it is to specify “painting” as a positive prompt and at the same time I tried specifying it as a negative prompt as well. Then, they often draw a picture in which bright colors are suppressed.
So we specify “painting” for the negative prompt, and “painting” for the positive prompt. I tried drawing a picture with various other words.
For some reason, trees are often drawn when one or two numbers are combined.
Interesting things other than trees are drawn.
Combining one letter of the alphabet can draw various pictures, but it seems that it is not as interesting as numbers.
I also tried combining words with repetitions like “lili”. Both in China and in the West, girls and pandas are named like this. In other words, we want them to draw cute things. I will write more about the results in another article, but here are some examples. The following images are some of the results obtained with repeated words.
Words such as “juju” and “susu” draw cute things such as girls as expected, but they also draw things like scary faces.