Introduction

Writing effective prompts is the foundation of consistent AI character generation. Whether you're creating a mystical alchemist or a modern-day rogue, the way you structure your prompt directly impacts the quality and coherence of the results.

In this guide, I'll share the fundamental structure I use for all my character portraits, along with techniques that have helped me maintain visual consistency across dozens of images.

Basic Structure

A well-structured prompt typically follows this order:

  1. Quality Tags - Set the baseline quality
  2. Subject - Who or what is in the image
  3. Appearance - Physical characteristics
  4. Clothing/Equipment - What they're wearing or carrying
  5. Setting/Background - Where the scene takes place
  6. Style Modifiers - Artistic direction
  7. Technical Parameters - Lighting, camera angle, etc.
💡 Why Order Matters

AI models typically weight earlier tokens more heavily. Put your most important descriptors first!

Character Description

When describing a character, be specific but concise. Here's my template:

[age] [gender] [species/race], [hair color] [hair style], [eye color] eyes, [distinguishing features]

Example:

"young woman, human, long silver hair in a side braid, amber eyes, small scar on left cheek"

Key Elements to Include:

  • Hair: Color, length, style (ponytail, braided, loose, etc.)
  • Eyes: Color, shape, expression
  • Face: Shape, skin tone, distinguishing marks
  • Build: Height, body type (if relevant)

Style Modifiers

Style modifiers help control the artistic direction of your generation. Here are some of my favorites:

Art Style

  • anime style, manga style - Japanese animation aesthetics
  • digital painting - Smooth, painted look
  • concept art - Professional game/film style
  • illustration - Book illustration quality
  • semi-realistic - Balance between anime and realism

Mood & Atmosphere

  • soft lighting, dramatic lighting
  • warm colors, cool colors
  • dreamy atmosphere, dark atmosphere
  • ethereal, mysterious, cheerful

Quality Tags

Start your prompt with quality indicators to establish a baseline:

masterpiece, best quality, highres, ultra-detailed

These tags tell the model to prioritize quality and detail. You can adjust based on your needs:

  • masterpiece - Highest quality tier
  • best quality - Excellent quality
  • highres - High resolution details
  • ultra-detailed - Fine details
  • intricate details - Complex patterns and textures

Example Prompts

Example 1: Fantasy Character

masterpiece, best quality, 1girl, young woman, human, long wavy purple hair, green eyes, fair skin, wearing alchemist robe, blue and gold clothing, holding glowing potion bottle, medieval laboratory background, shelves with bottles, magical atmosphere, soft lighting, digital painting, detailed background

Example 2: Modern Character

masterpiece, best quality, 1girl, young woman, short black hair, brown eyes, wearing casual streetwear, hoodie, jeans, urban background, city street, sunset lighting, anime style, vibrant colors
✨ Pro Tip

Save your base character description as a text snippet. You can reuse it across multiple prompts by just changing the pose, outfit, or background!

Pro Tips

1. Use Parentheses for Emphasis

Most AI models support emphasis through parentheses:

  • (red hair) - Slight emphasis
  • ((red hair)) - Strong emphasis
  • [red hair] - De-emphasis

2. Be Specific with Colors

Instead of "blue dress," try "navy blue dress" or "sky blue dress." Specific color names yield more consistent results.

3. Use Reference Artists (When Appropriate)

Adding artist names can help guide the style:

in the style of [artist name]

4. Negative Prompts Are Your Friend

Use negative prompts to exclude unwanted elements:

low quality, blurry, bad anatomy, extra limbs, watermark, signature
⚠️ Important Note

Always respect artists' wishes regarding AI training. Some artists explicitly request their names not be used in prompts.