SYSTEM NOTICE

Auto translation by AI. Be sure, accuracy, nuances and authorial intent may not be fully reflected.
見出し画像

[Definitive Guide] End the AI Image Generation Gacha! YAML-Style Image Generation Prompts!

Nyanto@AIお役立ち/都市伝説


~Create professional-quality images even as a beginner! Templates included~

🎯 This article is recommended for people like this!

  • You've tried AI image generation, but it's not going well...

  • The character is good, but the background is different every time...

  • It's painful to think of prompts from scratch every time...

  • I want to mass-produce with the same character, but I don't know how!

Presented to you, the ultimate syntax to save you from prompt confusion
—that is—
💥 YAML-style prompts!

Hello, I'm Nyanto, a writer who brings you surprising discoveries!
Today, I'd like to introduce a method that makes it much easier to control the randomness of generated images, which almost everyone faces in AI image generation!

What is a YAML prompt?

YAML is a data description format that is "easy for humans to read," used in programming configuration files and the like.
Applying this writing style to AI image generation instructions is what we call a "YAML-style prompt."

Conventional prompt:

a cute cat, red background, pop art style

YAML-style prompt:

character:
  species: cat
  color: black and white
background:
  color: red
style:
  - pop art
prompt: >
  A cute black and white cat on a red background in pop art style.

The point is that you can organize image elements by category!
This clears your mind, making modifications and reuse very easy.

The 3 Major Benefits of YAML Style

1. No more confusion by turning prompts into a "blueprint"!

You can write down "I want an image like this!" from your head by separating it into elements such as:
・Character features
・Background types
・Overall atmosphere
…and so on, which also organizes the AI's thinking.

2. Once you make it, you can mass-produce it!

If you save it as a template, you can perform

partial changes in an instant

such as:・Changing only the character's color・Changing only the background・Changing only the pose

3. Swap only the background → The atmosphere changes completely!

Since you can change the background from "summer sea" to "winter night" or "neon city" while keeping the character the same,
it's easy to change the world setting while keeping the same character!

Basic YAML Prompt Template [Free Distribution]

*The following were generated using image FX, which allows for free image generation.

🐱 For Characters


character:
  species: squirrel
  color: light brown and white
  eye_detail: large, sparkling eyes
  expression: surprised and happy
  pose: holding a sunflower seed with both hands
  mouth: open, joyful
style:
  - bold pop art
  - Japanese shonen manga style
  - simple composition suitable for stickers
prompt: >
  A cartoon squirrel character with large sparkling eyes, 
a joyful open-mouthed expression, and holding a sunflower seed with both hands. 
Illustrated in a bold pop-art and Japanese shonen manga style. No background. 
Designed to be used as a YouTube icon or sticker.



日本語版
キャラクター:
種類: リス
色: ライトブラウンと白
目の特徴: 大きくてキラキラした瞳
表情: 驚いていて嬉しそう
ポーズ: 両手でひまわりの種を持っている
口元: 開いていて楽しげ
スタイル:太めの線のポップアート風
日本の少年マンガ風 
ステッカー向きのシンプル構図
プロンプト: >
大きなキラキラした瞳をしたリスのキャラクター。
口を開けて楽しそうな表情で、両手でひまわりの種をもったポーズ。
ポップアートと日本の少年漫画風のイラスト。
背景なし。YouTubeアイコンやステッカーに適したデザイン。

🏞 For Backgrounds


scene:
  place: bamboo forest
  time: twilight
  season: summer
mood: mystical
style:
  - Japanese watercolor
prompt: >
  A mystical bamboo forest at twilight in Japanese watercolor style.
日本語版
シーン:
場所: 竹林
時間帯: 夕暮れ
季節: 夏
雰囲気: 神秘的
スタイル:日本の水彩画風
プロンプト: >
夏の夕暮れ時、竹林に広がる幻想的な風景。日本の伝統的な水彩画スタイルで描かれた静寂で神秘的な情景。

How is using YAML convenient?

🎯 Example: "I want to change the background of the squirrel character from earlier"

With YAML, you just need to add the "background" section! 👇

character:
  species: squirrel
  color: light brown and white
  eye_detail: large, sparkling eyes
  expression: surprised and happy
  pose: holding a sunflower seed with both hands
  mouth: open, joyful
style:
  - bold pop art
  - Japanese shonen manga style
  - simple composition suitable for stickers
prompt: >
  A cartoon squirrel character with large sparkling eyes, 
a joyful open-mouthed expression, and holding a sunflower seed with both hands. 
Illustrated in a bold pop-art and Japanese shonen manga style. No background. 
Designed to be used as a YouTube icon or sticker.

↓

character:
  species: squirrel
  color: light brown and white
  eye_detail: large, sparkling eyes
  expression: surprised and happy
  pose: holding a sunflower seed with both hands
  mouth: open, joyful
background:
  scene: forest
  time: morning
  lighting: soft sunlight streaming through the trees
  mood: peaceful and fresh
style:
  - bold pop art
  - Japanese shonen manga style
  - semi-realistic natural background
prompt: >
  A cartoon squirrel character with large sparkling eyes, 
a joyful open-mouthed expression, holding a sunflower seed with both hands, 
standing in a peaceful morning forest with soft sunlight streaming through the trees. 
Illustrated in a bold pop-art and Japanese shonen manga style, with a natural, semi-realistic background.


日本語版
キャラクター:
種類: リス
色: ライトブラウンと白
目の特徴: 大きくてキラキラした瞳
表情: 驚いていて嬉しそう
ポーズ: 両手でひまわりの種を持っている
口元: 開いていて楽しげ
背景:
シーン: 森の中
時間帯: 朝
光の演出: 木々の間からやさしく朝日が差し込む
雰囲気: 穏やかで爽やか
スタイル:太めの線のポップアート風
日本の少年マンガ風
背景は自然寄りのセミリアルタッチ
プロンプト: >
大きなキラキラした瞳をしたリスのキャラクター。口を開けて楽しそうな表情で、両手でひまわりの種を持っている。
やさしい朝日が木々の間から差し込む穏やかな森の中に立っている姿。ポップアート風と日本の少年マンガ風のテイストで、
自然なセミリアルな背景付きのイラスト。


This allows you to generate images with completely different atmospheres!

Common mistakes and how to fix them

Common Mistake Solution Too much information packed in Limit elements to 3-4. Clarify the theme! Too abstract expressions Make it concrete, e.g., instead of "cute," use "smiling with red cheeks"

How to convert images to YAML

You've seen an AI image and thought, "I want to recreate this!" right?
Actually, it is possible to reverse-engineer a YAML prompt from an image!

Analysis with ChatGPT

Just upload the image to ChatGPT and ask it this:

この画像をYAMLプロンプト形式に分解して。

Example of the result you get:

character:
  gender: female
  hairstyle: long straight hair
  hair_color: golden blonde
  eye_color: purple
  expression: calm and slightly serious
  clothing: black long-sleeve shirt
  accessories:
    - round black-rimmed glasses
  pose: touching her glasses while looking forward
background:
  color: soft beige (solid color)
  props:
    - wooden table
style:
  - anime-style
  - clean and simple
  - slice-of-life atmosphere
prompt: >
  A calm and slightly serious anime-style female character 
with long golden blonde hair and purple eyes.
She is wearing a black long-sleeve shirt and round black-rimmed glasses, 
gently touching her glasses while looking forward.
The background is a soft beige solid color with a wooden table visible in front. 
The illustration has a clean, slice-of-life anime aesthetic.

日本語版
キャラクター:
性別: 女性
髪型: ロングストレート
髪色: ゴールドブロンド
目の色: 紫
表情: 落ち着いていて、少し真剣
服装: 黒の長袖シャツ
小物: 黒縁の丸メガネ
ポーズ: メガネに手を添えてこちらを見つめている
背景:
色: 淡いベージュ（単色背景）
小道具: 木目のテーブル
スタイル:アニメ調
クリーンでシンプル
日常系
プロンプト: >
落ち着いた表情のアニメ風の女子キャラクター。ゴールドブロンドのロングヘアで、紫色の瞳。
黒いシャツを着て、黒縁の丸メガネをかけ、メガネに手を添えて静かにこちらを見ている。
背景はシンプルな淡いベージュで、木目のテーブルが写っている。
クリーンで整った日常系アニメスタイルの構図。

It describes the details very concretely! It's quite difficult to put it into words like this yourself, isn't it? While there seem to be various ways like using image analysis tools, I think ChatGPT is sufficient.

✅ Summary: YAML prompts are an "extension device for creativity"

The YAML style is not just a way of writing.
It is a magical prompt for creating "words that get through" to AI.

  • Highly reproducible

  • Easy to modify

  • Can accurately instruct what you want to create

This is a highly recommended method for anyone who wants to get serious about working with ChatGPT and AI image generators!

いいなと思ったら応援しよう！