Quest 107 - Image Prompt Engineer

Quest 107: Image Prompt Engineer

easy 15-20 minutes

---

🎯 Learning Objectives

• ✅ How to structure prompts with subject, style, composition, lighting, and mood
• ✅ Why vague prompts produce vague images and specific prompts produce specific results
• ✅ How to improve prompts with before/after examples
• ✅ Model-specific tips for DALL-E, Midjourney, and Stable Diffusion

---

📖 Concept: Precision in Image Prompts

Image generation models like DALL-E, Midjourney, and Stable Diffusion are incredibly powerful — but they’re not mind readers. The quality of the output directly depends on the quality of your prompt. A vague prompt like “a cat” produces a generic, forgettable image. A structured prompt like “a fluffy orange tabby cat sitting on a sunlit windowsill, soft focus, warm afternoon light, cozy atmosphere” produces a specific, compelling image.

Prompt engineering for images is about giving the model enough structure to understand your vision. The best prompts follow a formula: Subject → Style → Composition → Lighting → Mood. This isn’t just theory — it’s a practical skill that separates effective AI artists from frustrated ones.

Think of it like martial arts kata: the structure isn’t a constraint, it’s a foundation that lets you express creativity effectively.

---

⚙️ How It Works

The Prompt Structure Formula

1. Subject — What is the main focus?
   "a warrior meditating under a cherry blossom tree"
         ↓
2. Style — What art style?
   "digital painting, Studio Ghibli inspired"
         ↓
3. Composition — How is it framed?
   "wide angle, low perspective, centered subject"
         ↓
4. Lighting — What's the light source?
   "golden hour, soft backlighting, warm tones"
         ↓
5. Mood — What feeling does it evoke?
   "peaceful, serene, contemplative"

Before vs After

Bad Prompt	Good Prompt
“a mountain”	“snow-capped mountain peak at sunrise, aerial photography, dramatic clouds, epic scale, warm golden light”
“a robot”	“friendly humanoid robot in a kitchen, soft pastel colors, Pixar style, warm lighting, cozy atmosphere”
“food”	“artisan sourdough bread on rustic wooden board, overhead shot, natural daylight, food photography, warm tones”

---

💡 Example: Building a Prompt Step by Step

Let’s build a prompt for a specific image:

Step 1: Start with the subject

a samurai standing in a field of bamboo

Step 2: Add style

a samurai standing in a field of bamboo, ink wash painting style, Japanese sumi-e

Step 3: Add composition

a samurai standing in a field of bamboo, ink wash painting style, Japanese sumi-e, vertical composition, bamboo framing the figure

Step 4: Add lighting and mood

a samurai standing in a field of bamboo, ink wash painting style, Japanese sumi-e, vertical composition, bamboo framing the figure, misty morning light, peaceful and contemplative

Each layer adds specificity. The model now has enough information to produce a focused, intentional image.

---

⚠️ Common Mistakes

Mistake 1: Being too vague

“Make me a cool image” → The model has no idea what “cool” means to you. Be specific about subject, style, and mood.

Mistake 2: Contradictory styles

“Photorealistic anime style watercolor” → Pick one primary style and stick with it. Contradictory instructions confuse the model.

Mistake 3: Ignoring model differences

“Same prompt for DALL-E, Midjourney, and SD” → Each model has different strengths. DALL-E excels at literal interpretations, Midjourney at artistic styles, Stable Diffusion at customization.

Mistake 4: Overloading with keywords

“cat, cute, fluffy, orange, tabby, sitting, window, sun, light, warm, cozy, home, indoor, paws, tail, whiskers” → A wall of keywords dilutes focus. Structure your prompt with clear hierarchy.

---

📝 Knowledge Check

📝 Knowledge Check

Q1:What is the recommended structure for an effective image prompt?

Adjective + noun + verbSubject → Style → Composition → Lighting → MoodA list of random keywordsJust describe what you want in natural language

Q2:Why is 'photorealistic anime style watercolor' a bad prompt?

It's too longIt uses contradictory styles that confuse the modelIt doesn't include enough keywordsAnime and watercolor are the same style

Q3:How do DALL-E, Midjourney, and Stable Diffusion differ in prompt handling?

They all interpret prompts identicallyEach model has different strengths — DALL-E for literal, Midjourney for artistic, SD for customizationOnly Midjourney supports image promptsStable Diffusion doesn't accept text prompts

---

🏋️ Quest: Image Prompt Engineer

Now it’s time to practice! Write a guide for effective image generation prompts.

1. Download the starter files:

   npx bluebeltdojo download quest-107-image-prompt
   cd quest-107-image-prompt

2. Open problem.js to read the design requirements

3. Create image-prompt-guide.md with all five required sections:

   • Prompt Structure (subject, style, composition, lighting, mood)
   • Style Keywords Library (organized by category)
   • Common Mistakes (vague prompts, conflicting styles)
   • Before/After Examples (bad prompt → improved prompt)
   • Model-Specific Tips (DALL-E, Midjourney, Stable Diffusion)

4. Run node test.js to validate your document structure

5. Verify all tests pass:

   node test.js

6. When all tests pass, submit your solution:

   npx bluebeltdojo submit

💡 Tip: The best way to learn prompt engineering is to experiment. Try the before/after examples with an actual image model and see the difference structure makes.

---

คำใบ้

• นี่เป็น design quest — งานจริงอยู่ใน image-prompt-guide.md
• โครงสร้าง prompt: Subject → Style → Composition → Lighting → Mood
• รวม before/after examples ที่แสดงให้เห็นว่า prompt ที่ดีกว่าสร้างภาพที่ดีกว่า
• ครอบคลุม model-specific tips สำหรับ DALL-E, Midjourney, Stable Diffusion
• ถ้าติดขัด ลองอ่าน “Common Mistakes” อีกครั้ง — อย่าดู solution โดยตรง