[Hand-drawn/Generative AI] The time I tried having AI draw my own artwork
I saw people on my X timeline having AI draw their own artwork,
and it looked interesting, so I tried it out myself.
The thumbnail image for this note is an AI-generated image from that project.
[Introduction] How to read this article
This article is intended as an observation log.
I have compiled the prompts and processes used when feeding my hand-drawn art into the AI,
keeping them as raw as possible.
Since it is more of a record than an explanatory article,
it is structured to be skimmed or read selectively.
I think it will be the kind of thing where you can just freely look at
the images and comments that catch your eye.
1st image: Ideal side-tail little sister
I tried having it generated.
For some reason, she's clutching a sharp pencil,
which gives off a sense of childhood madness, haha.
2nd image: Mascot keychain
this is one that confused the gallery, sweat.
3rd image: Tardigrade person
4th image: Pokemon's Jigglypuff
"Please generate an image of this character flying in the sky.
It is a round, overall pink character,
with fluffy bangs."
5th image: Overseas animation-style doodle
For some reason, the generated image looks like a bird, haha.
↓Prompt
"Please generate an image of this character walking through an American town
in the style of an overseas cartoon."
6th image: Three-eyed monster
It has more hair volume and a more solid center of gravity than the original,
and its chubby, squat look is super cute.
↓Prompt
"Please draw a picture of this monster."
7th image: Yukkuri Reimu
↓Prompt
"Please generate a Yukkuri Reimu with a sharp look based on this illustration"
8th image: A woman mimicking a koi fish
I was surprised that the text in this generated image didn't break at all.
Even so, the coloring is kind of... lewd, isn't it? 🧐
↓Prompt
"Please generate a picture of a woman mimicking a koi fish receiving food"
9th image: Beast fairy
there are parts that weren't reflected in some places.
Still, I think it's mostly reproduced.
↓Prompt
"Please generate a color picture of this beast fairy. The horns are milky white, the hair and fur are bright purple, and there are yellow highlights on the bangs and the tips of the ears. The fishhook on the tail is metallic gold. The chest fur is white, the hoodie is black, and the eyes are 'violet'."
[Postscript] I mistook 'violet' for scarlet, and here is the corrected image↓
10th image: Manga title page
I didn't specify the colors, but
I was surprised that the character's color scheme was just as I imagined.
↓Prompt
"Please generate a color picture of this title page"
11th image: Country junior high girl with a center-parted short haircut
It only fails when I type in a prompt full of my own quirks... 😣
↓Prompt
"Please generate a full-body picture of this character. The hairstyle is black hair with a center-parted short cut, wearing a white scarf and a black sailor uniform. Please generate a picture of her carrying a black backpack and a school bag, walking to school on a country road."
12th image: One-shot gag manga
The generated image has turned into something amazing, lol
↓Prompt
"Please turn this image into a color manga"
13th image: Group of NanJ users
It seems like the 'Nan-J' character has been trending in short videos lately,
so I used the model rendering images of the Nan-J character
and its color variations that I previously made in Blender as a reference to recreate them.
↓Original image
↓Process
14th image: Fairy
↓Prompt
'Please generate an image of this fairy gracefully flying in a blue sky with fluffy clouds floating by'
15th image: Junior high school boy
but that was a mistake.
The original image is a junior high school boy.
↓Prompt
'Please generate a full-body image of this high school boy'
16th image: Teary-eyed Hanakappa
so I included similar instructions in the prompt.
↓Prompt
'Please generate an image of Hanakappa with teary eyes like this'
Bonus (Keychains)
