Stable Diffusion is a powerful AI image generation tool that creates images based on text prompts. To get the most out of Stable Diffusion, it’s important to understand the special prompt symbols and formatting options available. This guide will walk through the key prompt symbols to use for consistent and high-quality AI-generated images.
Brackets for Weight Adjustment
One important technique is using brackets [[]]
to adjust the weight given to certain words or concepts. Putting a word or phrase inside double brackets reduces its weight, telling Stable Diffusion to de-emphasize it. For example:
[[Woman]] with red hair and green eyes
This prompt will likely generate images focused on the red hair and green eyes while placing less emphasis on the woman herself. The more brackets used, the less weight given.
Colons for Step-Wise Changes
Another useful prompt symbol is the colon :
to make step-wise changes during image generation. The basic format is:
[Concept A: Concept B: Step]
For example:
[Daytime scene: Nighttime scene: 10]
This tells Stable Diffusion to generate 10 steps using the daytime version before switching to the nighttime version. The step value can be any number to control when the switch happens.
Pipes for Alternation
The pipe symbol |
inserts alternation into the prompt, switching between concepts/styles with each step. For example:
[Realistic style | Cartoon style | Photorealistic style] of a lion
This prompt will cause the style to change between realistic, cartoon, and photorealistic with each image generation step.
Advanced Prompt Formatting
In addition to special symbols, Stable Diffusion supports Markdown formatting in prompts for more control. Here are some advanced formatting options:
LaTeX Math Expressions
Wrap any math expressions in double dollar signs $$
so they are rendered properly, like this:
$$x^2+y^2=z^2$$ sphere
Code Blocks
Use Markdown code blocks with language identifiers to insert code into prompts:
import torch
x = torch.rand(3, 4)
Markdown Formatting
Use bold, italics, lists, headings, etc. to structure prompt concepts:
Subject: Woman
Description: Smiling while on a beach
- Wearing a blue shirt
- Long blonde hair
Image Insertion
Insert images with Markdown syntax and URL:
Example Image
Key Resources
Here are some useful Stable Diffusion prompt resources:
Let me know if you need any clarification or have additional questions!