Chapter 3: Creating Effective Image Prompts
The Art of Visual Communication with AI
Prompt engineering for images is both an art and a science. Master these techniques to consistently generate the images you envision.
Anatomy of an Effective Image Prompt
The Formula
[Subject] + [Action/Context] + [Style] + [Lighting] + [Composition] + [Quality modifiers]
Component 1: Subject
Be Specific
| Vague ❌ | Specific ✅ |
|---|---|
| A dog | A golden retriever puppy |
| A woman | A 30-year-old woman with curly red hair |
| A building | A Victorian Gothic mansion |
| A car | A 1967 cherry red Mustang convertible |
Describe What You Want
❌ "A cat"
✅ "A fluffy orange tabby cat with green eyes,
sitting on a windowsill"
Component 2: Style
Art Styles
- Photorealistic
- Oil painting
- Watercolor
- Digital art
- Anime/manga
- Comic book
- Pencil sketch
- Low poly
- Pixel art
Artist References
Many platforms understand artist names:
- "in the style of Studio Ghibli"
- "reminiscent of Monet's impressionism"
- "Greg Rutkowski fantasy style"
- "Alphonse Mucha art nouveau"
Era/Movement
- Art Deco
- Renaissance
- Baroque
- Minimalist
- Cyberpunk
- Steampunk
- Vaporwave
Component 3: Composition
Framing
- Close-up / portrait
- Medium shot
- Wide shot / establishing shot
- Bird's eye view
- Worm's eye view
- Dutch angle
Perspective
- First person POV
- Over the shoulder
- Rule of thirds
- Centered composition
- Symmetrical
- Asymmetrical balance
Examples
"Portrait shot, shallow depth of field,
subject centered, negative space on left"
Component 4: Lighting
Types of Lighting
| Term | Effect |
|---|---|
| Golden hour | Warm, soft, romantic |
| Blue hour | Cool, moody, twilight |
| Dramatic lighting | High contrast, shadows |
| Soft lighting | Even, flattering |
| Rim lighting | Glowing edges |
| Volumetric | God rays, atmosphere |
| Neon | Cyberpunk, urban |
| Candlelight | Warm, intimate |
Examples
"Dramatic side lighting with deep shadows"
"Soft diffused lighting, overcast day"
"Volumetric sunbeams through forest canopy"
Component 5: Quality Modifiers
Resolution & Detail
- High resolution
- 8K, 4K
- Highly detailed
- Sharp focus
- Intricate details
- Hyperdetailed
Rendering Quality
- Photorealistic
- Ray tracing
- Octane render
- Unreal Engine
- Cinema 4D
- Professional photography
Trending Terms
These often boost quality:
- "Award winning"
- "Trending on ArtStation"
- "Featured on Behance"
- "National Geographic photo"
Negative Prompts
What They Do
Tell the AI what to AVOID in the image.
Common Negative Prompts
ugly, deformed, blurry, bad anatomy,
bad hands, extra fingers, disfigured,
poorly drawn face, mutation,
low quality, worst quality, watermark,
text, signature, cropped
Platform Support
- Stable Diffusion: Full support
- Midjourney: Use
--no [item] - DALL-E 3: Not supported
Practical Examples
Example 1: Product Photography
A sleek modern smartwatch floating on
a gradient background, studio lighting,
product photography, minimalist,
professional, sharp focus, 8K resolution
Example 2: Fantasy Art
An ancient elven library carved into a
giant tree, magical floating books,
warm candlelight mixed with blue magical
glow, intricate wood carvings, fantasy
concept art, highly detailed,
by Greg Rutkowski
Example 3: Portrait
Professional headshot of a confident
businesswoman in her 40s, natural smile,
wearing navy blazer, soft studio lighting,
neutral grey background, Canon 85mm f/1.4,
shallow depth of field
Example 4: Architectural Visualization
Modern minimalist house with floor-to-ceiling
windows, infinity pool, sunset golden hour,
architectural photography, wide angle lens,
real estate photography, warm tones,
photorealistic
Platform-Specific Tips
DALL-E 3
- Be natural and descriptive
- Works well with conversational prompts
- Excels at text in images
- Refine through ChatGPT dialogue
Midjourney
- Use parameters (
--ar,--v,--style) - Shorter prompts often work better
- Aesthetic keywords are powerful
- Use
/describeto learn from images
Stable Diffusion
- Use negative prompts liberally
- Experiment with CFG scale
- Different models have different strengths
- Try multiple samplers
Advanced Techniques
Prompt Weighting
Some platforms support emphasis:
(important concept:1.5), regular concept,
(de-emphasized:0.8)
Seed Consistency
Use the same seed to:
- Create variations of a concept
- Maintain character consistency
- Reproduce results
Upscaling Workflow
- Generate at lower resolution
- Select best candidates
- Upscale for final quality
Common Mistakes to Avoid
1. Being Too Vague
❌ "A pretty picture of nature"
✅ "Misty mountain lake at sunrise,
pine forest reflection,
Banff National Park style,
landscape photography"
2. Contradictory Instructions
❌ "A dark bright sunny night scene"
✅ "A moonlit night scene with
soft ambient glow"
3. Overcomplicating
❌ [500 word prompt with every detail]
✅ Focus on key elements that matter most
4. Ignoring the Platform
- Don't use Midjourney syntax in DALL-E
- Adjust approach for each tool
Prompt Templates
Template: Product Shot
[Product description], [surface/setting],
[lighting type] lighting, product photography,
commercial, professional, [quality modifiers]
Template: Character Art
[Character description], [pose/action],
[setting/background], [art style],
[lighting], highly detailed,
[artist reference if desired]
Template: Landscape
[Location/scene type], [time of day],
[weather/atmosphere], [style],
[composition], [quality modifiers]
Key Takeaways
✅ Structure prompts: Subject → Style → Lighting → Composition → Quality
✅ Be specific about what you want
✅ Use appropriate style and artist references
✅ Leverage negative prompts to avoid common issues
✅ Adapt your approach to each platform
✅ Iterate and refine based on results
Module Summary
You've completed Module 4. You now understand:
- How diffusion models work: Denoising guided by text
- Major platforms: DALL-E, Midjourney, Stable Diffusion
- Effective prompting: Structure, components, and techniques
Next Steps
Test your knowledge with the Module 4 Quiz, then continue to Module 5: Audio and Music Generation.