Stable Diffusion is a powerful AI image generation model that creates images based on text prompts. Writing effective prompts is key to getting good results from the model. Here are some prompt engineering tips and examples to help you create better Stable Diffusion images.
Basic Prompt Structure
A basic Stable Diffusion prompt has a simple structure:
[Subject], [Style], [Details]
For example:
A still life painting of a bowl of fruit, highly detailed oil painting by Rembrandt
This prompt structure clearly defines the subject (still life bowl of fruit), style (Rembrandt oil painting), and details (highly detailed).
Key things to include:
- Subject – What is the main focus of the image? Be specific.
- Style – What artistic style should it follow? Reference a well-known artist for consistency.
- Details – Extra descriptive details about the image.
Using Weights and Brackets
You can control the emphasis of prompt terms using parentheses and brackets:
(term)
: Slightly more emphasis((term))
: Even more emphasis[[term]]
: Slightly less emphasis
For example:
A large golden retriever dog playing in a field full of flowers ((wearing a blue bandana around its neck))
The triple parentheses around “wearing a blue bandana” tells Stable Diffusion to pay extra attention to that detail.
Prompt Ordering
List prompt terms in order of importance:
A majestic wizard with a long white beard and purple robes, highly detailed digital art
Here “wizard” is most important, followed by physical details, then style details.
Using Negative Prompts
Add -[term]
to exclude unwanted elements, like:
An astronaut floating in space -[poorly drawn]
This reduces chances of a crudely drawn image.
Image Size and Quality
Include resolution to get larger images:
An ornate hourglass on a stone pedestal, digital art, 4k
You can also request higher quality:
A regal lion illustration, trending on ArtStation, extremely detailed
Guiding Overall Style
Give guidance on the overall mood/style:
A cute puppy and kitten cuddling, heartwarming
Terms like “heartwarming”, “serene”, “gloomy” help set the tone.
Advanced Prompting Techniques
Once you have a handle on basic prompting, try these more advanced techniques:
Stylistic References
Name a specific artist for stylistic consistency:
A still life oil painting by Vincent Van Gogh of a vase of sunflowers
Compare it to:
A still life oil painting of a vase with sunflowers
Naming Van Gogh makes a big difference!
Descriptive Scene Setting
For landscape images, describe the scene in detail:
A misty autumn forest scene with golden morning light shining through the orange and red leaves of the tall trees, digital painting trending on ArtStation
The more descriptive the better for complex scenes.
Directing Viewpoint
Guide the viewpoint for consistent compositions:
A small bird sitting on a tree branch viewed from below, digital art
Terms like “from below”, “from above”, “profile view”, etc. help set viewpoint.
Guiding Lighting
Describe lighting for more realistic images:
A still life painting by Caravaggio of fruit in a bowl with strong chiaroscuro lighting
Words like “chiaroscuro”, “dramatic lighting”, etc. guide lighting style.
Comparative Language
Use comparisons for subjective qualities:
An astronaut floating gracefully in space, more beautiful than any Hubble space image
Comparisons like “more beautiful than” help communicate intent beyond literal meaning.
Combining Techniques
You can combine multiple techniques in one prompt for powerful results:
A majestic snow leopard walking through a misty Himalayan forest at dawn, viewed from the side, highly detailed National Geographic photography by Steve Winter
This prompt uses weights, ordering, scene details, viewpoint, stylistic reference, and lighting guidance all in one!
Useful Stable Diffusion Websites
Here are some useful sites for learning more about Stable Diffusion prompts:
Stable Diffusion Prompt Design
Stability AI’s official prompt design guide for Stable Diffusion with tips.
AITuts Stable Diffusion Prompts
Prompt engineering guide for beginners from AITuts.
sd-prompts.org
Community site with a large database of prompts and images.
With practice, you’ll get better and better at writing prompts that produce exactly what you imagined. Mastering prompt engineering is crucial to get the most out of Stable Diffusion.