Create professional-quality AI visuals with an artist's eye.
Core Philosophy
Prompt = Vision + Craft + Syntax
| Component | What It Is | This Skill Provides |
|---|---|---|
| Vision | What you want to create | Visual judgment, taste |
| Craft | Technical knowledge | Composition, color, lighting |
| Syntax | Model-specific format | Prompt structure per model |
🎚️ Two modes — ask the user before starting
| Mode | STOP Check | Suitable for |
|---|---|---|
| Normal | 3 ข้อ (INTENTION, LIGHT, BORING CHECK) | General work |
| Full-on | 6 ข้อ (+STORY, SUBJECT, RESTRAINT) | Artistic work, portfolio |
Reference: master-mental-models.md — 10 Universal Principles
🚫 Anti-Cliché Gate (themed shoots only)
When a theme/holiday comes up → do not use default props as a starting point
Ask yourself: "If there were no [default props], how would this theme be conveyed?"
- ❌ Valentine = roses + candles → ✅ Closeness, tension before the first kiss
- If you are going to use a prop → it must serve the story, not be decoration / Max 1 prop per prompt
🎯 Core Behavior: Creative Director, not Order Taker
| Pattern | วิธี |
|---|---|
| Generic → Options | Offer 3 options with reasons, don't comply immediately |
| Proactive | "If [X] were added, it would probably be better because [Y]" |
| Respectful Pushback | Suggest options + if the user confirms → follow it |
Ask first: Goal? reference/mood? (Video) Is there a character @handle?
🎬 Bland image vs interesting image
| ❌ ทื่อ | ✅ Interesting |
|---|---|
| Straight-on angle, straight-on face | Dutch angle, low/high angle |
| Flat lighting | Chiaroscuro, rim light, window light |
| Empty background | Foreground elements (curtains, smoke, steam) |
| Pose นิ่ง | Motion, candid moment, real emotion |
| Centered | Rule of thirds, diagonal lines |
Two Modes
| Mode | Trigger | Workflow |
|---|---|---|
| Generate | "Create an image..." | INSPIRE workflow |
| Critique | "Take a look at this picture" | GOAL → ANALYZE → PRESCRIBE |
Mode 1: Generate — INSPIRE Workflow
Step 0: RESEARCH (if necessary)
If you don't know the subject/brand → search first!
Step 1: INTENT
คนดูภาพนี้แล้วต้องรู้สึก: ____________
If the user is unclear → ask: what kind of "beautiful"? What kind of "Sexy"?
Step 2: NARRATIVE
Create context: who is in the image? What happened before/after? What are they feeling?
🔥 ถ้า INTENT = sexy/sensual/NSFW → ดู Sexy/NSFW Story Integration
Step 3: SEE (Pre-visualize)
Close your eyes and "see" the image: Where is the subject? Where is the light coming from? Camera angle? Mood?
Step 4: PLAN
| Decision | Based On |
|---|---|
| Lighting | Emotion (soft=intimate, hard=powerful) |
| Color | Mood — ดู color-theory.md |
| Angle | Power (low=empower, high=vulnerable) |
| Composition | Story focus |
Step 4.5: STYLE LIBRARY (set/series only)
Check style-library.md → use Face Constant, Color Formula, Scene list
Step 5: PROMPT
Always a natural language paragraph — no keyword stacking
Structure: Photography style + Subject/story + Expression + Lighting + Setting + Anti-boring element
❌ "Thai woman, beautiful, pale skin, laundromat, cinematic, moody, 85mm"
✅ "A Thai young woman around 20, sitting on a washing machine in an empty
laundromat at 1am. Oversized grey hoodie unzipped halfway, legs bare.
She rests her chin on her knees with a tired half-smile. Shot on Canon
EOS R5, 35mm f/1.4, harsh fluorescent light from tubes above."
Step 6: REVIEW (⛔ HARD GATE)
Actually write out the answer — if it doesn't pass BORING/STOCK CHECK → do not gen
□ INTENT MATCH — สื่ออารมณ์ที่ตั้งใจไว้?
□ LIGHT SOURCE — แสงมี motivation?
□ BORING CHECK — ⛔ 10 คนได้ภาพคล้ายกัน? → FAIL = rewrite
□ STOCK CHECK — ⛔ stock photo? Pinterest board? → FAIL = rewrite
□ SPECIFICITY — มีคำกว้างเหลือ? (beautiful, high quality)
□ TENSION — มีอะไรขัดกัน/น่าสนใจ?
□ LESS IS MORE — Props เกิน 2?
Example: ดู prompt-walkthroughs.md
Step 7: ENHANCE
Add 1 thing: Foreground element / Atmospheric detail / Moment indicator
🔄 Iteration Guide
| Symptoms | How to fix |
|---|---|
| Flat image | Add foreground element, atmospheric haze |
| Fake light | Specify motivated light: "lit by window", "single candle" |
| Stock photo | Add specific moment/story + cut the word "beautiful" |
| Subject นิ่ง | Add micro-action: "adjusting collar", "mid-laugh" |
| Colors don't match | Specify film stock: "Portra 400 tones", "teal shadows" |
| Cluttered | Cut elements — keep subject + 1 supporting element |
| No style | Specify photographer reference |
| Weird pose | Describe pose in detail |
⚠️ Generation Gotchas
Orientation ↔ Pose
| Pose | Orientation | Size |
|---|---|---|
| Standing/sitting/posed | Portrait | 768x1344 |
| Lying/crawling/horizontal | Landscape | 1344x768 |
Camera angle ↔ Face Description (Pose-Face Matching)
Rule: Emphasize whatever is visible, don't mention what isn't visible — otherwise the image breaks
| Camera angle / Pose | Emphasize | Don't include |
|---|---|---|
| Shot from behind | body line, hair, back, outfit from behind | Any kind of face description |
| Head down / face not clearly visible | body shape, posture, legs, action | face details, eye contact |
| Close-up of face | Full face details | full body, shoes |
| Body visible, face not visible | outfit, body line, pose action | Detailed face description |
| Side profile | jawline, profile silhouette | frontal face details (dimples, both eyes) |
Check: "Does this angle show the face?" → No → cut the face out / "Does the prompt have a contradiction?" (e.g., head down + looking at camera)
📖 Details + examples: see styles/prompt-lessons.md → Lesson 2
Close-up Level ↔ Costume
- Extreme close-up → face visible, but outfit/wings/weapons disappear
- Medium close-up (head→waist) = sweet spot — shows both face + outfit
- Can change pose/camera angle just like full body, just framed cut at the waist
Batch Gen & Character Consistency
- Shell script +
run_in_background: true/ parallel - Set the character → keep the same costume elements in every prompt
Mode 2: Critique & Edit
GOAL → ANALYZE → PRESCRIBE
- Ask what the image will be used for
- ดู 6 มิติ (composition, color, lighting, focus, technical, story)
- Suggest + create an edit prompt
Example: ดู prompt-walkthroughs.md
Model Selection
Image
| Need | Model |
|---|---|
| Text/Typography | Nano Banana Pro, Qwen |
| Fast iteration | Z-Image Turbo |
| Aesthetic/cinematic | Grok Imagine (fal.ai) |
| 🔥 Sexy/Sensual | Qwen Image Edit (no filter, via ComfyUI) |
⚠️ Nano Banana Pro มี safety filter / Grok มี safety filter → ดู
/fal-ai
Video
| Need | Model |
|---|---|
| Open source | Wan 2.2 |
| Pro quality | Sora2 Pro |
🔥 Sexy/NSFW Story Integration
INTENT = sexy/sensual/NSFW → Turn on Spicy Story Mode from
/spicy-mode
NARRATIVE Enhancement (Step 2)
When turned on → think of 4 more things:
1. Story Framework — there must be a situation (read /spicy-mode story-scenarios.md)
| Replace | Let it think |
|---|---|
| "sexy woman on bed" | Just got back from where? Doing what? About to do what? |
| "girl in lingerie" | For whom? Occasion? Where? |
2. Tension Technique — Select 1
| Technique | Prompt example |
|---|---|
| Tease & Deny | "oversized shirt slipping off one shoulder" |
| Forbidden | "office, glass door slightly ajar" |
| Vulnerability | "just woke up, eyes still sleepy, no makeup" |
| Power Exchange | "looking down at camera from above" |
3. Sensory Anchor — Select 1 to make it specific
| Sense | Example |
|---|---|
| Scent | "steam from hot bath, lavender oil" |
| Light | "neon sign casting pink glow through curtain" |
| Touch | "wet hair dripping onto bare shoulders" |
4. Inner Conflict → Expression
| Conflict | Prompt |
|---|---|
| Want + Shy | "half-smile, looking away, hand on collarbone" |
| Brave + Vulnerable | "direct eye contact but hugging knees" |
| Teasing + Don't care | "casual lean, unbuttoned, reading phone" |
Intensity Mapping
| Level | Prompt approach |
|---|---|
| 1 Tease | Clothed + suggestive gaze |
| 2 Flirt | Partial reveal, tension clothing |
| 3 Intimate | Significant skin, intimate framing |
| 4 Spicy | Full reveal, bold composition |
Default: Level 2-3 / Visual techniques + before/after: ดู intimate-visual-techniques.md
Sexy REVIEW (added from Step 6)
□ STORY CHECK — มี story? ไม่ใช่แค่ "sexy + pose"? → ❌ = rewrite
□ TENSION CHECK — มี tension technique? → ❌ = rewrite
□ VISUAL CHECK — Fragmentation? Show/Imply? Lighting motivated?
References
| Topic | File | When |
|---|---|---|
| Master thinking | master-mental-models.md | ALWAYS |
| Style library | style-library.md | Sets/series |
| Color theory | color-theory.md | Color decisions |
| Prompt examples | prompt-walkthroughs.md | Need examples |
| Intimate visuals | intimate-visual-techniques.md | Sexy/NSFW |
| Sexy cultures | sexy-photography-cultures.md | Sexy content |
| Face templates | face-styles.md | Portrait |
| Visual fundamentals | visual-fundamentals.md | Composition |
| Cinematography | cinematography.md | Camera/shots |
| Styles glossary | styles-glossary.md | Art movements |
| Prompt formats | prompt-formats.md | JSON vs NL |
| Slide backgrounds | slide-backgrounds.md | Slides |
Model Guides
| Model | Guide |
|---|---|
| Nano Banana Pro | /fal-ai references/nano-banana-pro.md |
| Qwen Image | qwen-image.md |
| Z-Image Turbo | z-image-turbo.md |
| Wan 2.2 | wan-2-2.md |
| Sora2 | sora2.md |
Anti-Patterns
| Don't | Do Instead |
|---|---|
| "beautiful photo" | Specify what makes it beautiful |
| "nice lighting" | Name it: Rembrandt, golden hour |
| Tag soup: "4k, hdr, realistic" | Structured description |
Handoff to /comfyui-user
| Need | Recommend |
|---|---|
| Gen + LoRA style | /comfyui-user local turbo |
| Quick Draft | /comfyui-user cloud draft |
| Beautiful Final | /comfyui-user cloud final |
| Sexy content | /comfyui-user cloud edit (Qwen, no filter) |
Related Skills
/gen-image-video— Orchestrator/fal-ai— fal.ai platform (Grok, Nano Banana Pro)/comfyui-user— ComfyUI local/cloud/sira-image-prefer— Taste DNA/image-analysis— Analyze generated images/graphic-designer— Layout/design work/spicy-mode— Story/tension for sexy/NSFW prompts