[Stable Diffusion] Using the popular image generation AI, we asked Van Gogh to draw a hamster

Hello.
This is Kawa from the Systems Solutions Department, and I've been having back pain lately.

Happy New Year. It's 2023

You might have a feeling about this too, but looking back, 2022
felt like a year in which a lot of new technologies emerged.
AI technology in particular has evolved incredibly without us even realizing it.
It's a little frightening in the sense that it might make workers obsolete.

So, I decided to learn about AI technology so that I could be on the side of using it rather than being used by it

The thumbnail image for this Beyond blog list was
actually generated by AI.
Space hamster, space ham.

Using Stable Diffusion , an image generation AI that was released around summer and became quite a hot topic , I was able to create a space ham surprisingly easily. The power of science is amazing!

If a master painter drew a hamster

A while ago, the "What if Van Gogh painted a biker gang?" series was really popular, but
this time I'd like to have him paint a hamster.

"A painting of a hamster in Van Gogh style" -

Van Gogh.
It has a "The Starry Night" vibe, doesn't it?

"An impressionistic painting of a hamster in Monet style

. It has that nice Impressionist feel.
By the way, the top image of this article is also by Monet, but that one is a version without the word "impressionism."

A picture of a hamster in the style of

Katsushika Hokusai.
It has a very Japanese feel to it.

"A painting of a hamster in the style of Piccasso's Guernica."

Finally, Picasso.
I think I captured the Guernica feel.

So, how do you do it?

If you do some research, you'll find various methods, but this time I'm doing it in an Ubuntu environment that fully utilizes the CPU rather than the GPU. Since the generation uses a model based on training data, I used Hugging Face 's Stable-Diffusion-v-1-4 model. (It has a generous license that allows commercial use as long as the content is not malicious or harmful to others.)

Prompt Tips

To output the image you want, you need to devise a prompt sentence

For example, the hamster below (??),

The image output using "space hamster" often doesn't turn out as intended because the expression is too abstract, causing the model to automatically combine elements and generate its own interpretation, resulting in something that doesn't even resemble a hamster. Therefore, if you want to represent a hamster in space, using something like "a hamster in space" might be a better approach. (It's easier to understand if you imagine the old game "Where is Who and What is Happening?")

Other expressions you can use to make the image more concrete include "in Van Gogh style" or "in the style of Van Gogh,
" "a painting of,"
"a photograph of,"
"abstract." and

Example:
"a painting of Tokyo Sky Tree in Renoir style"

"a painting of Tokyo Sky Tree in Van Gogh style"

Give it a try!

complete