【画像生成AI】FLUX.1で複数人物を描き分けるプロンプトテクニック

FLUX.1を使ってみての第一印象は「プロンプトめっちゃ効くわ!」でした。

それまでの画像生成AIの「あたなの言うことあまり聞きません」な感じに日々イライラさせられていた身としては、これは画期的。

それもそのはずで、FLUX.1は120億(12B)のパラメータを持っています。これは非常に大規模なモデルであり、高度な画像生成能力の源となっています。

ちなみに、Stable Diffusion XL (SDXL)は約33億(3.3B)のパラメータを持っています。Midjourneyは非公開ですが、その性能から推測すると、数十億のパラメータを持つ大規模なモデルであると考えられます。

という訳で、そんなFLUX.1なら複数人物の描き分けも簡単に実現できてしまいます。今までひたすら生成ボタンをクリックして、ガチャを回してきた苦労がウソのようです。

以下がそのプロンプトの例です。モデルは「flux1-dev-fp8」を使っています。

プロンプト
Photograph of three men engaged in a lively discussion about smartphone design. Scene set in a modern, minimalist conference room with a large window showing a city skyline. In the center, a sleek white table with a prototype smartphone and design sketches spread out.

Left: A distinguished older man with glasses and a black turtleneck sweater, gesturing enthusiastically.

Middle: A man with a kind face and graying beard, wearing a colorful, casual shirt, leaning forward with interest.

Right: A younger man in his 30s with a trendy haircut, wearing a tech company t-shirt and blazer, holding a tablet.

All three are focused on the smartphone prototype, with animated expressions suggesting an intense but friendly debate. Subtle Apple-inspired decor in the background.

日本語訳:

スマートフォンデザインについて活発に議論する3人の男性の写真。舞台は、大きな窓から都市のスカイラインが見える現代的でミニマルな会議室。中央には洗練された白いテーブルがあり、その上にスマートフォンのプロトタイプとデザインスケッチが広げられている。

左:眼鏡をかけ、黒のタートルネックセーターを着た、風格のある年配の男性が熱心にジェスチャーをしている。

中央:優しい表情で白髪交じりのひげを生やし、カラフルでカジュアルなシャツを着た男性が、興味深そうに前かがみになっている。

右:30代の若い男性で、流行の髪型をし、テクノロジー企業のTシャツとブレザーを着て、タブレットを持っている。

3人全員がスマートフォンのプロトタイプに注目しており、表情は熱心だが友好的な議論を示唆している。背景にはAppleにインスパイアされた控えめな装飾がある。

スティーブ・ジョブズ風、スティーブ・ウォズニアック風、テックブロガー風、3人の人物を登場させてみます。

単にそれぞれの人物の配置と見た目の記述を追加しただけで、特別なテクニックは必要ありません。

出力結果

ちゃんと、プロンプト通りに風貌やファッションが描き分けられています。

では、次に5人まで増やしてみましょう。

プロンプト
Photograph of five people engaged in a lively discussion about smartphone design. Scene set in a modern, minimalist conference room with a large window showing a city skyline. In the center, a sleek white oval table with a prototype smartphone and design sketches spread out.

Left side:
1. A distinguished older man with glasses and a black turtleneck sweater, gesturing enthusiastically.
2. A woman in her 40s with short, stylish hair, wearing a smart business suit, pointing at a design sketch.

Center:
3. A man with a kind face and graying beard, wearing a colorful, casual shirt, leaning forward with interest.

Right side:
4. A younger man in his 30s with a trendy haircut, wearing a tech company t-shirt and blazer, holding a tablet .
5. A woman in her late 20s with long hair tied back, wearing a colorful dress and a cardigan, sketching on a digital tablet.

All five are focused on the smartphone prototype, with animated expressions suggesting an intense but friendly debate. Their body language indicates active participation and engagement. Subtle Apple-inspired decor in the background, including a minimalist poster of a classic Apple product.
出力結果

5人、いけましたね。では、9人でやってみます。

プロンプト
A hyper-realistic image in the style of a high-end movie still, depicting a diverse group of 9 people in an intense iPhone design meeting. Scene set in a sleek, modern conference room with floor-to-ceiling windows showcasing a dramatic city skyline at dusk.

Central focus:
- A large, curved glass table with a cutting-edge prototype iPhone in the center, surrounded by holographic design projections and scattered paper sketches.
- Warm, cinematic lighting with a mix of cool blue from outside and warm indoor lights, creating a dynamic atmosphere.

Participants (ensure diverse representation in age, gender, and ethnicity):
1. A distinguished older man with silver hair and glasses, wearing a black turtleneck, standing and gesturing passionately.
2. A friendly-looking man with a graying beard in a colorful, casual shirt, leaning forward, engaged in conversation.
3. A trendy young man with a modern haircut, wearing a tech company t-shirt and blazer, holding a tablet.
4. A confident woman in her 40s with short hair, in a sharp business suit, pointing at a hologram.
5. A creative-looking woman in her late 20s, with vibrant-colored hair, casual-chic attire, sketching on a digital pad.
6. An older man with a distinguished look, wearing a classic suit, observing intently.
7. A young man of South Asian descent, wearing smart casual attire, explaining something with enthusiasm.
8. A middle-aged woman with an artistic flair, colorful glasses, gesticulating while speaking.
9. A young woman of East Asian descent, in business casual attire, taking notes on a futuristic-looking device.

Composition:
- Participants arranged around the table in various poses - some seated, some standing, all actively engaged.
- Camera angle slightly elevated, capturing the entire scene with cinematic depth of field.
- Background details include subtle Apple-inspired decor, state-of-the-art technology, and a wall-mounted screen displaying complex design schematics.
出力結果

まだいけるかも。

FLUX.1くんの限界を見てみたい。18人に挑戦です。

プロンプト
a hyper-realistic image in the style of an epic movie still, depicting a large-scale iPhone design meeting with 18 diverse participants. Set in a vast, ultra-modern conference space with a panoramic view of a futuristic city skyline at sunset.

Central focus:
- A massive, U-shaped smart glass table dominating the room, displaying interactive 3D models of iPhone prototypes and holographic design projections.
- Dramatic lighting mixing the warm glow of sunset with cool, high-tech blue accents from various screens and projections.

Participants (ensure diverse representation in age, gender, ethnicity, and roles):

Core Team (prominently positioned):
1. Distinguished older man, silver hair, black turtleneck, standing, gesturing at a hologram
2. Friendly man with graying beard, colorful shirt, engaged in animated discussion
3. Trendy young man, modern haircut, tech t-shirt and blazer, capturing the scene with a futuristic camera
4. Confident woman, 40s, sharp suit, short hair, pointing at a projection
5. Creative woman, late 20s, vibrant hair, casual-chic, sketching on a floating transparent tablet

Extended Team:
6. Distinguished older man, classic suit, observing intently
7. Young South Asian man, smart casual, excitedly explaining a concept
8. Middle-aged woman, artistic flair, colorful glasses, gesturing while speaking
9. Young East Asian woman, business casual, organizing holographic task lists
10. African man, 30s, stylish tech-wear, manipulating a complex algorithm visualization
11. Older woman, elegant attire, listening attentively
12. Young Middle Eastern woman, hijab, modern outfit, examining a deconstructed device
13. Latino man, 40s, business casual, presenting market trend holograms
14. Woman with visible disability, smart casual, demonstrating an adaptive interface
15. Androgynous person, avant-garde outfit, pointing to a timeline projection
16. Teenage prodigy, casual but sharp, excitedly sharing ideas via holographic sketch
17. Older Asian man, traditional-modern fusion attire, analyzing regional data
18. Woman, 50s, eco-chic style, showcasing green tech materials

Composition:
- Multiple focal points creating a dynamic, layered scene
- Participants arranged in small groups around the U-shaped table and standing areas
- Varied poses: some seated, some standing, others moving, all actively engaged
- Elevated camera angle capturing the entire space with cinematic depth of field

Environment details:
- Cutting-edge technology seamlessly integrated into the room's design
- Wall-sized screens displaying complex schematics and global market data
- Floating holographic displays scattered throughout the space
- Sustainable design elements: living plant walls, eco-friendly materials

Atmosphere:
- A palpable sense of innovation, collaboration, and high-stakes decision-making
- Mix of intense concentration and excited discussion, capturing a moment of collective creative breakthrough
- Visual cues suggesting a global, forward-thinking approach to design
出力結果

おお、ちゃんと18人いる!さすがにここまでくると男女の描き分けはだんだん微妙になってきていますが・・・。

コメント

コメントする

目次