17 Writing prompts
cmdrgrunt edited this page 2024-03-31 00:17:28 -04:00
    1. The Basics

Keep your prompts short don't write full sentences. Separate different aspects of your image by commas:

girl on a swing, green grass, pink trousers

Due to the way Stable Diffusion was trained, there are a few words that often improve the quality of images, e.g. highly detailed or trending on artstation:

image

You can use the names of art styles, artists, or computer programs to influence the look of the rendered images. A wide choice of modifiers can be found in the Image Modifiers section:

image

You can also put the modifiers directly into your prompt to get many different styles:

image

Emphasizing parts of the prompt

You can provide the importance ("weight") as numbers: girl on a (swing)1.2 (green grass)1.0 (pink trousers)0.8 Enclose the words with parentheses (round brackets) and put the weight after them. For e.g. in the above example, swing has a weight of 1.2, green grass has a weight of 1.0 and pink trousers has a weight of 0.8.

Testing variations of prompts

Use curly brackets in prompts to try different words, e.g. the prompt man riding a {horse,motorcycle} creates two rendering jobs: One for man riding a horse and one for man riding a motorcycle:

image

You can include entire portions of a prompt that include commas inside curly brackets by using the \ escape character before each comma you don't want to be treated as a variation separator. For example:

{a tall\, brown hair\, green eyed\, man, a short\, blond hair\, blue eyed\, woman} shopping at a store will create the following two tasks:

  • a tall\, brown hair\, green eyed\, man shopping at a store
  • a short\, blond hair\, blue eyed\, woman shopping at a store

Note that the \ escape characters while visible in the prompt, will be ignored when generating your image.

Prompt Matrix - Separate your prompt with the | character to explore variations quickly. E.g. girl holding a rose | illustration | cinematic lighting creates four task combinations automatically:

  • girl holding a rose
  • girl holding a rose, illustration
  • girl holding a rose, cinematic lighting
  • girl holding a rose, illustration, cinematic lighting

Negative prompts

The negative prompt can be used to describe what you don't want to see in the image. The AI sometimes ignores parts of the negative prompt, especially if the prompt and negative prompt contain very similar terms. In other cases, the negative prompt has big impact on the composition of the image. For example, if you put "shoes" into the negative prompt, you might get a barefoot person - or you get an image where the feet are out of frame.

A useful negative prompt is:

Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, mutated hands and fingers, out of frame

Further reading

Prompt Syntax (weights, emphasis etc).

The PromptBook from OpenArt, a platform and community dedicated to AI-native content, and is written by members from their community.

Stable Diffusion Modifier Studies This is a list of hundreds of different style modifiers.

lexica. Art This site is a search engine for prompt and image ideas. Warning: Some of the images on this website can be NSFW (adult-rated), please use it at your own discretion.

Prompt Examples and Experiments