アルトマンに告ぐ

プロンプトはGPTと対話して作成。ちょっとした洒落です。
なかなか難しいですね。



1枚目　サムネイル用
2枚目　プロンプト1
3枚目　プロンプト2
4枚目　プロンプト3
5枚目　加工例の元画像 https://www.chichi-pui.com/posts/be6afb9c-4376-4456-bd23-4a9387c7d3f3/
6枚目　バナー


■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
　2枚目の画像の変換プロンプト
■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■

添付画像を参照画像とし、以下の処理を行って、全く異なる印象となるイラストにしてください。
"""
Use the provided image as a reference.

Recreate it as a simple digital illustration.
Keep the image clean, flat, and slightly artificial.

The character should be drawn in a basic anime-like style,
with simple shapes and minimal detail.
Do not exaggerate facial features.

The face should be mostly symmetrical,
with medium-sized eyes.

Give the character a mild, simple expression,
such as a slight smile or a gentle, unclear emotion.
The expression should feel basic and somewhat imprecise.

Include a simple, loosely defined background.
The background should suggest a setting using basic shapes or colors,
but remain unclear and not detailed.
It should feel present, but not carefully designed.

Lighting should be soft and even.
Shading should be simple and flat.

Use a centered composition.

The result should look like an early text-to-image output:
straightforward, slightly generic, and imperfect rather than polished.
"""

■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
　3枚目の画像の変換プロンプト
■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■

添付画像を参照し、以下の指示に従い、同一のスタイル・印象ではないイラストを生成して。
"""
Use the provided image as a reference.

Recreate it as a simple digital illustration.
Keep the image flat, slightly artificial, and not fully refined.

The character should be drawn in a basic anime-like style,
with simple shapes and minimal detail.
Do not exaggerate facial features.

The face should be mostly symmetrical,
with medium-sized eyes.

Give the character a mild, simple expression,
such as a slight smile or a gentle, unclear emotion.
The expression should feel basic and somewhat imprecise.

Include a simple, loosely defined background.
The background should suggest a setting using basic shapes or colors,
but remain unclear and not carefully designed.

Allow slight roughness in the rendering:
linework may be a bit uneven,
shading may be inconsistent,
and some minor visual noise or diffusion artifacts may remain.
The image should feel slightly messy and unfinished,
as if generated by an early text-to-image model.

Lighting should be soft and even.
Shading should remain simple and flat.

Use a centered composition.

The result should look like an early text-to-image output:
straightforward, generic, and imperfect rather than clean or polished.

Do not correct or clean up small visual imperfections.
"""

■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
　4枚目の画像の変換プロンプト
■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■

{
"prompt": "Use the provided image as a reference.\n\nMake a simple, slightly rough digital illustration of the same scene.\nKeep it basic and a little uneven, as if it came from an older text-to-image system.\n\nCharacter style:\n- Simple anime-like character design with average proportions.\n- Medium-sized eyes (not oversized), simple nose and mouth.\n- Mild, friendly expression (a small smile or gentle, unclear emotion).\n- Avoid distinctive signature features (no extreme hairstyle, no heavy accessories).\n\nRendering:\n- Soft, even lighting.\n- Simple shading with occasional unevenness.\n- Allow minor imperfections (slight wobble in lines, soft edges, small mismatches).\n\nBackground:\n- Include a lightly suggested setting (basic shapes and colors).\n- Keep it simple and somewhat unclear, but not empty.\n\nOverall:\n- Straightforward and slightly generic.\n- Do not over-polish or clean up small imperfections.",
"negative_prompt": "photorealistic, ultra-detailed, cinematic lighting, dramatic shadows, sharp specular highlights, hyperreal textures, modern polished anime, heavy painterly brushwork, high contrast, complex intricate background, text, logo, watermark",
"style_hints": {
"detail_level": "low",
"finish": "slightly rough",
"lighting": "soft even",
"composition": "stable and centered-ish",
"background": "simple suggested setting (not empty)"
}
}