Stable Diffusion is a powerful AI image generation model that creates images from text prompts. Properly formatting prompts is key to generating high-quality images that match the desired output. This article provides prompt formatting guidelines and examples to help users create better Stable Diffusion images.
A well-structured prompt has three main components:
Subject and Style
The first part describes the subject or main focus of the image, as well as the desired style. For example:
A scenic oil painting of a girl in a flower field
Here, the subject is “a girl in a flower field” and the style is “scenic oil painting”.
The second part provides more details about the subject to guide the image generation, like:
A scenic oil painting of a girl in a flower field. The girl has long blonde hair and is wearing a white dress. She is running through the field on a sunny day with a big smile on her face. There are vibrant red, yellow, purple, and pink flowers all around her. The background shows towering snow-capped mountains and a bright blue sky with fluffy clouds.
The last section contains modifiers to refine the output, such as:
A scenic oil painting of a girl in a flower field. The girl has long blonde hair and is wearing a white dress. She is running through the field on a sunny day with a big smile on her face. There are vibrant red, yellow, purple, and pink flowers all around her. The background shows towering snow-capped mountains and a bright blue sky with fluffy clouds. Highly detailed, depth of field, illustration, trending on artstation
Terms like “highly detailed” and “depth of field” ask Stable Diffusion to pay special attention to certain qualities.
Here are some key formatting tips for writing better Stable Diffusion prompts:
The more details you provide, the better. Ambiguous prompts produce ambiguous results.
Use Full Sentences
Full sentences describe the image naturally compared to just listing words and phrases.
Specify Number of Subjects
If you want multiple or just one subject, make that clear, e.g. “a group of three dogs” or “one flower”.
Include colors, textures, sizes, etc. to define elements. For example, “a towering snow-capped mountain”.
Set the Tone
Inject some personality using descriptors like “whimsical”, “serene”, “magical”.
Direct the Focus
Guide attention to key parts of the image using emphasis words like “focused”, “prominent”, “main”.
Refine with Modifiers
Add modifiers like “intricately detailed”, “cinematic lighting” to refine the output.
Here are some prompt examples in different categories to spark ideas:
An epic wide landscape matte painting of a mystical golden temple surrounded by cherry blossom trees on the peak of a snowy himalayan mountain during sunset, intricate details, by Alphonse Mucha and Thomas Kinkade
A close-up portrait of a beautiful pensive woman with long wavy red hair wearing a flower crown in a fairy forest, intricate details, depth of field, by Waterhouse
A tabletop still life painting featuring a cut crystal vase with vibrant red roses and lilies, next to a plate with sliced fruit and drapped fabric, by Van Gogh, intricate realism, cinematic lighting
A abstract liquid art painting of an astronaut floating through a psychedelic galaxy nebula, intricate details, vibrant colors, by Mati Klarwein
Crafting the right prompts is key to generating stunning images with Stable Diffusion. Follow the tips outlined here on structure, formatting, and examples to create better prompts. Don’t be afraid to experiment with different styles, subjects, and modifiers to find what works best. Prompt engineering is an iterative process, but the AI capabilities of Stable Diffusion makes it fun and rewarding.