I can't draw, so I started making factory 4-koma comics with AI

I have been using AI to create 4-koma comics about factories on this note for a little over three months.

Since note was holding a contest called "#StartedWithAI," I thought I would write about how I got started and what it's like to make them, in line with that theme.

I thought more people would read them if I turned them into comics

To begin with, the reason I started 4-koma comics on note is becauseI thought that if I could create something that was not just text, but also visually easy to understand and interesting, more people would read it.
I thought comics would be good, so I simply started making 4-koma comics on a whim.

Making 4-koma comics is fine, but I can't draw

Making 4-koma comics is fine, but I can't draw, and I'm not familiar with comic composition.

When I draw, I end up with dramatic, manga-style old men and women.I wanted art that was simpler and easier to read.

I tried drawing for the first time in a while, but my artistic ability hasn't changed.
I can't draw backgrounds either, andeven though I had the 4-koma scenario in my head, I couldn't give it shape.
So, I decided to ask AI to handle the drawing.

Creating 4-koma comics with GPT and Gemini

I use GPT and Gemini.
I use GPT to brainstorm prompts together, and Gemini to generate the images.

At first, I was also using GPT for image generation.
However, the Japanese text in the images was strange, and I didn't really like the art style, so now I divide the roles this way.

Specific prompts used

Using the 4-koma comic that recently got the most likes as an example, I will introduce some of the prompts I used to create it.

Specific prompt categories

Since it would be too long to post the entire content as is, I will only introduce the "4-koma flow" part specifically here.

【4コマ内容固定】

1コマ目
	•	出演人物はロックバンド風の作業員3人のみ
	•	工場のライン横で、ロックバンド風の作業員3人がちゃんと工場作業をしている
	•	3人とも手元は部品・治具・ライン作業に向いている
	•	楽器演奏ではなく、あくまで作業員として作業している
	•	構図はやや寄り
	•	ロックバンド風の作業員Aをセンターで大きめに描く
	•	BとCも近くにいて、3人組の一体感が見えるようにする
	•	ただの作業風景ではなく、少しMVのような空気感を出す
	•	セリフ：
ロックバンド風の作業員A
「僕に降り注ぐ」
「過ちの運命」
「君にも見せたい」

2コマ目
	•	出演人物は現場リーダーA、一般作業員B、ロックバンド風の作業員3人
	•	構図は、右側に現場リーダーA、左側に一般作業員B、奥に作業中のロックバンド風3人
	•	現場リーダーAが一般作業員Bに聞く
	•	ロックバンド風の作業員3人は引き続き工場作業をしている
	•	3人とも部品やラインに触れており、演奏はしていない
	•	セリフ：
現場リーダーA
「彼はなんて言ったんだ？」
一般作業員B
「この行程きついから」
「誰か代わってです…」

3コマ目
	•	出演人物は現場リーダーA、一般作業員B、ロックバンド風の作業員3人
	•	構図は、右側に現場リーダーA、左側に一般作業員B、その奥または反対側に無言で止まるロックバンド風3人
	•	真ん中に少し空間を作り、空気が凍った感じを出す
	•	現場リーダーAが感心した表情で、悪気なく言ってしまう
	•	一般作業員Bはすぐに止める
	•	ロックバンド風の作業員3人は無言で静止する
	•	効果線や動きは減らし、静止感を優先する
	•	セリフ：
現場リーダーA
「表現力すごいな…」
「さすがビジュアル系」
一般作業員B
「それ言っちゃダメ…」
ロックバンド風の作業員3人
「…………」

4コマ目
	•	出演人物はロックバンド風の作業員3人と現場リーダーAのみ
	•	一般作業員Bはこのコマに出さなくてよい
	•	構図は斜め後ろまたは斜め横から見た奥行きのある構図
	•	ロックバンド風の作業員3人が、工場のラインに乗って無言で去っていく
	•	普通に立っているだけではなく、明確に「遠ざかって退場している」構図にする
	•	ただ歩いて帰るのではなく、ライブ演出の退場みたいに、妙に美しくスタイリッシュに流れながら去る
	•	ただし最後まで工場作業員の見た目のまま
	•	ステージではなく、あくまで工場ライン上を退場している絵にする
	•	3人は後ろ向きまたは横向きで、奥へ流れていく感じを出す
	•	背景には集中線を入れて、このコマを強調する
	•	ただしライブ中ではなく、あくまで退場の勢いを強調するための集中線にする
	•	ロックバンド風の作業員の1人が静かに訂正する
	•	現場リーダーAは手前で申し訳なさそうに引き止める
	•	セリフ：
ロックバンド風の作業員
「ビジュアル系じゃなくて」
「ロックバンドだ」
現場リーダーA
「すまん…」
「ライブ演出みたいに」
「帰らないで…」

However, this didn't go perfectly on the first try either.
After generating it many times and correcting the parts I didn't like, the 4-koma flow section finally ended up like this.

In reality, there is more correcting than anything else

Even if you write it out in this much detail, you almost never get the 4-panel comic you imagined on the first try.
The positions might be reversed, the dialogue might be assigned to the wrong person, the expressions might be off, or the composition of the punchline might be hard to understand.

That's why, in reality, there is more editing than creating.
It often takes many iterations of fixing the parts that bother me and generating the image over and over before it finally takes shape.

How I come up with ideas

The process for coming up with ideas starts by throwing the content I thought of into GPT as a rough note.
From there, I decide on the composition, atmosphere, and flow of the dialogue, and once it's somewhat organized, I turn it into a prompt.
After that, I generate the image with Gemini and keep fixing the parts I'm not happy with.

The beginning of the 4-panel comic I mentioned earlier, 'What if a rock band worked in a factory,' looked something like this

At first, I throw it in roughly like this, and from there, I think about the situation and composition.
Once it takes some shape, since I've taught it a template for 4-panel comic prompts, I have it summarize it into a prompt for image generation.

After that, I have Gemini generate the image, and then I make it while fixing things that are wrong, deciding I actually want different dialogue, or noticing the positions or expressions are wrong.

Summary

For someone like me who can't draw, having AI do the drawing was a revolution.

With the development of AI, the frustration of having a little idea but not being able to give it form has decreased significantly.

If there's anyone out there like me, try using AI and enjoy it.

The 4-panel comic series I'm making ▼