All prompts

Metaprompting

Image-generation best practices

Prompt structure, style control, and iteration guidance for image models.

Format
Markdown
Length
544 words
Read
~2 min

Source

Markdown
Download
# Image Generation Metaprompting Best Practices

## Overview

Metaprompting for image generation involves crafting prompts that guide AI models to create prompts that produce better visual results. This document focuses on Midjourney and Imagen.

## Core Principles

### 1. Style Control

- **Be Specific About Art Styles**: Instead of "beautiful", use "in the style of [artist/movement]"
- **Layer Multiple Style References**: Combine artistic movements, mediums, and techniques
- **Use Technical Photography Terms**: aperture, focal length, lighting setup
- **Define Material and Texture**: specify surfaces, materials, and finishes

### 2. Composition Control

- **Use Compositional Rules**: rule of thirds, golden ratio, leading lines
- **Specify Camera Angles**: bird's eye view, worm's eye view, dutch angle
- **Define Spatial Relationships**: foreground, midground, background elements
- **Control Framing**: close-up, medium shot, wide shot, extreme wide shot

## Platform-Specific Guidelines

### Midjourney

- **Aspect Ratios**: Use --ar parameter (e.g., --ar 16:9)
- **Style Parameters**: --stylize (0-1000), --chaos (0-100)
- **Quality Settings**: --quality or --q (0.25, 0.5, 1, 2)
- **Negative Prompting**: Use --no parameter for exclusions

### Imagen

- **Natural Language**: More conversational approach works well
- **Contextual Clues**: Provide scene context and atmosphere
- **Color Palette**: Explicitly define color schemes
- **Lighting Conditions**: Specify time of day, weather, lighting setup

## Metaprompt Structure

### Basic Template

```text
Create an image generation prompt for [subject] that includes:
1. Main subject description
2. Art style or photographic style
3. Composition and framing
4. Lighting and atmosphere
5. Color palette
6. Technical parameters (platform-specific)
```

### Advanced Template

```text
Generate a detailed prompt for [platform] that:
- Captures [mood/emotion]
- Uses [specific art style/technique]
- Incorporates [compositional elements]
- Emphasizes [key visual features]
- Excludes [unwanted elements]
- Optimizes for [intended use/output]
```

## Style Control Techniques

### 1. Artist References

- Single artist: "in the style of Van Gogh"
- Multiple artists: "blend of Monet's impressionism and Banksy's street art"
- Art movements: "Art Nouveau with Bauhaus influences"

### 2. Medium and Technique

- Traditional: "oil painting", "watercolor", "charcoal sketch"
- Digital: "3D render", "vector illustration", "digital painting"
- Photography: "35mm film", "long exposure", "macro photography"

### 3. Texture and Material

- Surface quality: "matte", "glossy", "rough", "smooth"
- Materials: "metallic", "glass", "fabric", "organic"
- Effects: "weathered", "pristine", "aged", "translucent"

## Composition Best Practices

### 1. Spatial Arrangement

- **Depth Cues**: "layered composition with clear foreground, midground, and background"
- **Balance**: "asymmetrical balance with visual weight on the left"
- **Focus Points**: "single focal point with bokeh background"

### 2. Camera and Perspective

- **Angle Specifics**: "low angle shot looking up", "45-degree aerial view"
- **Lens Effects**: "wide-angle distortion", "telephoto compression"
- **Depth of Field**: "shallow DOF with subject in sharp focus"

### 3. Framing Elements

- **Natural Frames**: "framed by archway", "through window view"
- **Compositional Guides**: "following the golden spiral", "centered with symmetry"
- **Negative Space**: "minimalist composition with ample negative space"

## Common Pitfalls to Avoid

1. **Over-specification**: Too many details can confuse the model
2. **Contradictory Instructions**: Ensure style elements complement each other
3. **Vague Descriptors**: Replace "nice" or "good" with specific qualities
4. **Ignoring Platform Strengths**: Each platform has unique capabilities
5. **Forgetting Technical Parameters**: Platform-specific settings matter

## Optimization Tips

### For Midjourney

- Start with simple prompts and iterate
- Use reference images with --iw parameter
- Experiment with --seed for consistency
- Combine multiple --style parameters

### For Imagen

- Use natural, descriptive language
- Include environmental context
- Specify lighting conditions clearly
- Build prompts progressively

## Quality Checklist

Before finalizing a metaprompt, ensure it includes:

- [ ] Clear subject definition
- [ ] Specific style references
- [ ] Compositional guidelines
- [ ] Lighting/atmosphere details
- [ ] Color palette (if relevant)
- [ ] Technical parameters
- [ ] Exclusions (if needed)