By name: ai-portrait-prompt description: >- Build structured AI portrait prompts using a 9-module photography direction framework. Transforms photographic intent (lens language, lighting design, depth of field, color grading) into precise, production-ready prompts for any image generation model (Flux, SDXL, Midjourney, etc.). Use when the user asks to:; "인물 사진 프롬프트 만들어줘", "portrait prompt", "AI 인물 생성"; "ai-portrait-prompt", "촬영 방안", "인물 프롬프트"; "프로필 사진 만들어줘", "매거진 인물", "시네마틱 인물"; "portrait shot plan", "character photo prompt"; "인물 이미지 생성", "AI 초상화", "프롬프트 설계"; Any request to generate a portrait/character image with specific photographic intent; Landscape or product photography prompts (handle directly or use muapi-cinema for video); Logo or...
AI Portrait Prompt Builder
Build structured AI portrait prompts using a 9-module photography direction framework. Transforms photographic intent (lens language, lighting design, depth of field, color grading) into precise, production-ready prompts for any image generation model (Flux, SDXL, Midjourney, etc.).
Triggers
Use when the user asks to:
- "인물 사진 프롬프트 만들어줘", "portrait prompt", "AI 인물 생성"
- "ai-portrait-prompt", "촬영 방안", "인물 프롬프트"
- "프로필 사진 만들어줘", "매거진 인물", "시네마틱 인물"
- "portrait shot plan", "character photo prompt"
- "인물 이미지 생성", "AI 초상화", "프롬프트 설계"
- Any request to generate a portrait/character image with specific photographic intent
Do NOT use for:
- Landscape or product photography prompts (handle directly or use muapi-cinema for video)
- Logo or graphic design (use canvas-design)
- Video generation prompts (use pika-text-to-video or muapi-cinema)
- General image generation without portrait/character focus (use muapi-image-studio)
Core Principle
효과적인 인물 프롬프트의 핵심은 스타일 단어의 수가 아니라, 촬영 판단이 명확한지 여부에 있다.
A portrait prompt is a simplified shooting plan. It must communicate:
- What this photo is FOR (purpose)
- WHO the subject is and what FEELING they convey
- Camera distance and perspective (lens language)
- Where light comes from and its quality
- How much background information to retain
- Color direction
- Where realism details should concentrate
Workflow
Step 1: Determine Purpose (Module 1)
Ask or infer the image's intended use. This anchors ALL subsequent decisions.
| Purpose | Key Requirements |
|---|---|
| Social media profile | Face clear, gaze stable, clean background |
| Professional headshot | Trust, naturalness, professionalism, approachability |
| Magazine editorial | Composition, negative space, atmosphere, overall mood |
| Commercial advertising | Clean lighting, stable skin, strong visual control |
| Lifestyle portrait | Natural, relaxed, feels shot in a real scene |
| B&W documentary | Emotion, texture, subject state, presence |
| Cinematic night | Color, shadows, atmosphere, mood |
Step 2: Define Subject & Mood (Module 2)
Never write just "a woman." Answer two questions:
- Who is this person? (identity/role)
- What feeling do they give? (mood/atmosphere)
Then specify: expression, skin texture preference, and any identity-preserving details.
Step 3: Choose Lens Language (Module 3)
| Focal Length | Feel |
|---|---|
| 24mm | Environmental, strong perspective distortion |
| 35mm | Street/documentary distance, story feel |
| 50mm | Natural, balanced, everyday portraits |
| 85mm | Classic portrait, stable face proportions, clean BG |
| 105mm | Beauty/commercial close-up, stronger compression |
| 135mm | Restrained, detached, pronounced BG compression |
Write as: 85mm portrait lens look, natural facial proportions, clean background separation
Step 4: Set Composition & Angle (Module 4)
Composition options: face close-up, bust-up, half-body, three-quarter body, full-body, cover composition
Angle options: eye level, slightly high angle, slightly low angle, frontal gaze, 45-degree side, candid offset
Step 5: Set Depth of Field (Module 5)
| Aperture | Effect |
|---|---|
| f/1.4 | Dreamy, emotional, strong bokeh (risks looking fake) |
| f/1.8 | Strong atmosphere, good for night/emotional portraits |
| f/2 | Subject stands out, works for lifestyle and style |
| f/2.8 | Stable portrait, face sharp, background moderately soft |
| f/4 | More environmental info retained (clothing, props, scene) |
| Deep DOF | Documentary, environmental, street scenes |
Step 6: Design Lighting (Module 6)
| Lighting | Best For |
|---|---|
| Soft window light | Natural, everyday, comfortable skin |
| Large diffused window | Softbox feel, headshots |
| Overcast soft light | Urban, low contrast, stable |
| Golden hour backlight | Warm, soft, outdoor mood |
| Clamshell lighting | Beauty, commercial, clear catchlights |
| Hard direct light | Fashion, dramatic, intense |
| Neon + tungsten mixed | Night, cinematic, urban |
| Available light | Documentary, presence, minimal control |
Always specify: direction, quality, and effect on face.
Step 7: Choose Color/Style Anchor (Module 7)
Pick ONE per prompt:
| Anchor | Direction |
|---|---|
| Kodak Portra 400 | Soft skin tones, lifestyle, natural film feel |
| Kodak Gold 200 | Warm, sunny, light retro |
| Fujifilm Classic Chrome | Low saturation, urban, restrained, editorial |
| Fujifilm Eterna | Soft, narrative, cinematic |
| CineStill 800T | Night, neon, tungsten, subtle halation |
| Ilford HP5 | B&W documentary, grain, subject state |
| Neutral editorial tone | Clean, neutral, professional |
| Clean commercial tone | Advertising, beauty, clean skin tones |
Step 8: Specify Realism Details (Module 8)
Target specific areas — never just write "realistic":
- Skin:
natural skin texture, subtle pores - Eyes:
realistic catchlights, focused gaze - Hair:
natural flyaway hairs - Fabric:
real fabric texture - Background:
optical bokeh, not artificial blur - Highlights:
natural highlight roll-off - Face:
slight facial asymmetry, realistic facial detail - Retouching:
not over-retouched, no plastic skin
Step 9: Add Negative Constraints (Module 9)
Standard negative set (always include):
no AI fake face, no plastic skin, no over-retouched skin, no westernized facial features,
no influencer look, no excessive bokeh, no unnatural catchlights, no distorted hands
Context-specific additions:
- Night scenes:
restrained neon glow, no excessive halation - Commercial:
natural expression, not stiff, not overly posed
Assembly Formula
A [purpose] portrait of [subject], with [mood], framed as [composition] from [camera angle],
using a [lens look] and [aperture feel], lit by [lighting setup], in [environment],
wearing [wardrobe], with [tone/style anchor], emphasizing [realism details],
avoiding [negative constraints].
Order: Purpose → Subject → How to shoot → Lighting & Color → Realism & Constraints
Style Presets (Quick Select)
| # | Style | Recipe |
|---|---|---|
| 1 | B&W Documentary | 35/50mm, available light, Ilford HP5, natural texture, grain |
| 2 | Film Lifestyle | 50mm, f/2, window light, Portra 400, half-body, relaxed |
| 3 | Low-Sat Urban | 85mm, f/2, overcast, Classic Chrome, bust-up, quiet editorial |
| 4 | Professional Headshot | 85mm, f/2.8, soft window light, neutral editorial, half-body |
| 5 | Commercial Beauty | 105mm, f/2.8, clamshell, face close-up, clean commercial |
| 6 | Cinematic Night | 85mm, f/1.8, neon+tungsten, CineStill 800T, subtle halation |
| 7 | Japanese Soft | 50mm, f/2.8, soft daylight, warm tone, clean BG, gentle |
| 8 | High-Contrast Fashion | 35/85mm, hard light, strong shadow, fashion editorial |
| 9 | Founder Portrait | 85mm, f/2.8, window light, minimal workspace, professional |
| 10 | Outdoor Golden Hour | 50/85mm, golden hour backlight, warm, soft flare |
Output Modes
Mode A: Full Prompt (Default)
Generate a complete English prompt following the assembly formula. Include all 9 modules.
Mode B: Fill-in-the-Blank (Interactive)
Ask the user to choose from each module sequentially, then assemble the final prompt.
Mode C: Style Preset + Customization
User picks a preset number (1-10), then specifies customizations (subject, environment, wardrobe).
Mode D: Korean Brief → English Prompt
User describes intent in Korean ("카페에서 자연스러운 분위기의 반신 인물 사진"), skill translates to structured English prompt.
References
Read references/portrait-modules.md for:
- Full module option lists with Korean/English mappings
- Extended style presets with example prompts
- Common mistakes and fixes
- Model-specific tips (Flux vs SDXL vs Midjourney)
Quality Checklist
Before outputting any prompt, verify:
- Purpose is explicit (not just "portrait")
- Subject has identity AND mood (not just "a woman")
- Exactly ONE lens focal length specified
- Aperture feel matches the intended DOF
- Lighting has direction + quality (not just "good lighting")
- Only ONE color/style anchor (no mixing film stocks)
- At least 2 realism detail targets specified
- Negative constraints included
- Total prompt length: 50-150 words (sweet spot for most models)
Structured Presets (구조화 — 모듈 조합 고정)
9-모듈을 매번 고르지 말고 프리셋으로 lens/조리개/조명/색감(필름스톡)/구도/사실감/네거티브를
잠근다. 프롬프트-온리(생성 API 없음) — 출력은 Flux/SDXL/MJ/Imagen에 붙여넣는 프롬프트.
컴파일은 공유 엔진([[gpt-image-2]] scripts/spec_compiler.py), 스키마는
gpt-image-2/references/SPEC_SCHEMA.md.
python3 scripts/portrait_prompt.py --list
python3 scripts/portrait_prompt.py --preset founder-headshot --subject "a young Asian woman"
python3 scripts/portrait_prompt.py --preset bw-documentary,cinematic-night --subject "a young man"
프리셋: assets/portrait-presets.json. 규칙: color_grade는 1개만(필름스톡 혼용 금지),
realism 2-4개, constraints에 universal 네거티브 포함. 새 룩은 키 추가만.