🤰
ImageFXで女子大学生のリアルな日常他撮り写真を生成する
課題
ImageFXの人物画像は「ハイクオリティ」になりすぎるため、その要素を薄めたリアル路線の画像を生成したい。
ImageFX
Imagen 3というモデルで画像生成できるGoogleのサービス。無料で使用できます。
どうしてもAPIで叩きたい場合は、有料ですがこちらから申請可能です。
生成した画像
プロンプト
ImageFXにダイレクトに突っ込むときのプロンプトです。次の章で汎用的な書きかたにします。
A realistic photo of a Japanese woman in her 20s, smiling shyly and making a peace sign with her hand at a theme park.
She has natural makeup, long black hair, and is wearing casual, trendy clothing like a pastel-colored sweater and jeans.
The background shows a vibrant theme park with colorful attractions, crowds, and a sunny sky.
The image has a shallow depth of field, mimicking a smartphone normal camera application, with the subject in sharp focus and the background slightly blurred. The lighting is natural and bright, with soft shadows.
The overall vibe is candid and amateurish, as if taken by a friend with a phone, with no professional editing or overly polished look.
Her expression is natural and not overly posed, conveying a slightly embarrassed or shy feeling, with a genuine and unforced smile that feels authentic and relatable.
The composition and quality resemble a TikTok or Instagram-level post, with a casual, everyday aesthetic that feels like it was shared on social media.
The colors are slightly muted, as if taken with a normal camera rather than a high-end one, giving it a more realistic and less saturated tone.
The angle is from the eye level of a male photographer, creating a slightly downward perspective that feels natural and immersive.
ほかのLLM用の汎用プロンプト
Variables:
{subject}: a Japanese woman in her 20s
{mode}: smiling shyly and making a peace sign with her hand
{location}: a vibrant theme park
{subject_description}: She has natural makeup, long black hair, and is wearing casual, trendy clothing like a pastel-colored sweater and jeans.
{expression_description}: Her expression is natural and not overly posed, conveying a slightly embarrassed or shy feeling, with a genuine and unforced smile that feels authentic and relatable.
{background}: a lively theme park with colorful attractions, crowds, and a sunny sky
----
Prompt:
A realistic photo of {subject}, {mode} at {location}. {subject_description}. {expression_description}.
The background shows {background}.
The image has a shallow depth of field, mimicking a smartphone camera, with the subject in sharp focus and the background slightly blurred. The lighting is natural and bright, with soft shadows. The overall vibe is candid and amateurish, as if taken by a friend with a phone, with no professional editing or overly polished look. The angle is from the eye level of a male photographer, creating a slightly downward perspective that feels natural and immersive.
プロンプトの概要
- 20代女性
- 笑っていて、ピースサイン
- テーマパーク(カラフルなアトラクション、人混み、晴れ)
- 流行の服
- 被写界深度を浅く、ボケを弱く
- 友だちの他撮り
- 男性の高さから撮影
参考(元ネタ)
とめさんのこのツイートが流れてきて、「ImageFXの胡散臭い画像の正体は被写界深度なのか!」と勉強させていただき、上記のプロンプトをつくってみました。
Discussion