Prompt Engineering Concepts
The practice of crafting effective text prompts to guide AI image generation. Involves understanding how models interpret text and using techniques to achieve desired results.
Prompt engineering is the skill of communicating your vision to AI image generators effectively.
Core Formula
A well-structured prompt typically includes:
[Subject] + [Style] + [Medium] + [Lighting] + [Composition] + [Quality]Key Techniques
Be Specific
- Weak: “a dog”
- Strong: “a golden retriever puppy with fluffy fur”
Order Matters
Models often weight earlier terms more heavily. Put important elements first.
Use Quality Boosters
Common enhancers:
- “highly detailed”
- “8k resolution”
- “masterpiece”
- “professional”
- “award-winning”
Style References
- Artist names (use ethically)
- Art movements: “impressionist”, “art deco”
- Media: “oil painting”, “digital art”
Negative Prompts
Specify what to avoid:
- “blurry, low quality”
- “bad anatomy, extra limbs”
- “watermark, signature”
Advanced Techniques
Token Weighting
Emphasize or de-emphasize terms:
(important word:1.3)(less important:0.7)
BREAK Keyword
Separate concepts into chunks (SD):
portrait of a woman BREAK
forest background with sunlightPrompt Templates
Build reusable structures for consistent styles.
Platform Differences
| Platform | Prompt Style |
|---|---|
| Midjourney | Comma-separated, keywords |
| DALL-E | Natural language, conversational |
| Stable Diffusion | Comma-separated, weighted |