Stable Diffusion is a powerful AI system for generating images from text descriptions. Crafting effective prompts is key to guiding Stable Diffusion to produce high-quality outputs aligned with your creative goals. This article provides an overview of prompt design fundamentals and shares examples across common use cases.
What Makes a Good Stable Diffusion Prompt?
An effective Stable Diffusion prompt clearly communicates the desired image characteristics in a detailed yet concise manner. Here are some best practices:
- Be Specific: Include key details about the subject, style, composition, lighting, mood etc. Ambiguous prompts yield inconsistent results.
- Limit Scope: Focus on 1-2 central elements. Overly complex prompts often fail to capture all details.
- Use References: Specify relevant artists, art movements or sample images to define visual style.
- Try Stylistic Variations: Tweak prompts systematically to expand the creative possibilities.
Prompt Examples
Below are some examples of good Stable Diffusion prompts for common use cases:
Portrait Photography
A sharp focus, high resolution portrait photo of a 20 year old woman with long blonde hair, bright green eyes, flawless skin, wearing a white tank top, looking happy, soft natural lightingBreakdown:
- Defines gender, age, key facial features, expression, attire and lighting
- Focuses on core subject without unnecessary details
Product Concept Art
Futuristic watch concept design with sleek curved glass screen, rose gold metal body, digital interface, rendered in Unreal EngineBreakdown:
- Specifies product type, key physical traits, styling influences and target output medium
- Focuses on industrial design elements rather than environment/context
Architectural Rendering
Photorealistic exterior view of a modern 5 story brick apartment building, with large glass windows, green shrubbery along the entrance path, rendered in OctaneBreakdown:
- Outlines architectural style, size, materials, exterior elements and target rendering engine
- Prioritizes building characteristics over surroundings
Character Illustration
Trending on ArtStation. Portrait of a strong female warrior with long red hair, armor, sword, dramatic lighting, highly detailed digital painting by Artgerm and Ross TranBreakdown:
- Defines gender, hair color, attire, accessories, lighting and level of detail
- Specifies target artistic style by naming relevant artists
Prompt Structure
Effective Stable Diffusion prompts generally follow this basic structure:
Subject Details
- Gender, age, ethnicity for portraits
- Object type and key traits for still life
- Architectural features for buildings/interiors
- Character details like costume for illustrations
Composition and Framing
- Cropping, angle and perspective
- Background elements and surroundings
- Foreground focus details
Artistic Style and Medium
- Realism, painting, drawing, 3D rendering
- Relevant artists, movements or sample images
- Target output resolution and format
Emotion and Lighting
- Overall mood and expression
- Type of lighting (soft, dramatic etc)
- Time of day if applicable
Prompt Engineering Tips
Follow these prompt engineering best practices to create effective prompts:
Iterate on Test Inputs
- Start with simple prompts and build up complexity
- Adjust prompts based on output quality and relevance
Vary Key Parameters
- Swap out subject terms, artistic styles and composition
- Test prompt elements individually and in combination
Analyze Failure Modes
- Identify the mismatch between prompts and outputs
- Add or remove details to minimize anomalies
Curate and Share Winning Prompts
- Save prompts that produce desired outputs
- Share best practices with fellow creators
Conclusion
Crafting the right prompts is critical for unlocking Stable Diffusion’s creative potential. Start by focusing prompts on key details, iteratively test variations, analyze results and share effective prompts. With practice, you can learn to better communicate your visual goals to produce astonishing AI-generated images.