General prompts for Stable Diffusion

Stable Diffusion is an open-source deep learning model for text-to-image generation. It can create high-quality and realistic images from text descriptions. Stable Diffusion is based on latent diffusion models and has been trained on a large dataset of image-text pairs from the internet.

Some key capabilities of Stable Diffusion:

  • Generate photorealistic images from text prompts
  • Control image styles, attributes and composition
  • Inpainting and outpainting of images
  • Image translations guided by text
  • Animation and video generation

Prompt Format

When providing a prompt to Stable Diffusion, it is important to be as descriptive as possible. A good prompt clearly specifies the subject matter, attributes, style, composition and other details.

Here is the basic structure of a Stable Diffusion prompt:

  • Subject matter and attributes
    • Specify the main subject and key attributes
  • Style
    • Artistic style (e.g. oil painting), genre (e.g. sci-fi), time period
  • Composition
    • Framing, angle of view, number of subjects
  • Fine details
    • Facial expression, clothing, background elements
  • Quality and rendering
    • Output size, level of realism/abstraction

Example Prompts

Realistic portrait

A high resolution close up portrait photograph of a beautiful young woman with long wavy blonde hair, bright blue eyes, dark eye makeup, and red lipstick, warm lighting, 35mm depth of field

Landscape concept art

An epic wide landscape matte painting of a lush green valley with a crystal lake, surrounded by tall snowy mountains and waterfalls, fantasy style, highly detailed, by Greg Rutkowski and Thomas Kinkade

Futuristic cityscape

A futuristic neon cityscape digital painting, with towering skyscrapers, flying vehicles, and trains, in night time, cinematic lighting, volumetric fog, in the style of Blade Runner movie, 8k resolution

Prompt Engineering Techniques

There are several techniques that can help optimize prompts to generate better images with Stable Diffusion:

Specify an Artist Style

Referencing a specific artist helps guide the model to match their style:

A portrait by [Picasso|Van Gogh|Monet] of a [man wearing a suit|woman in a dress] in [blue and yellow|bright vibrant colors], [oil painting|watercolor painting]

Use Comparisons

Compare attributes to provide more context:

A Siamese cat with fur as soft as silk and eyes as blue as the ocean

Include Emotional Cues

Evoke an emotional response to render more expressive images:

A sad clown with a single tear running down his cheek

Direct the Focus

Call attention to important elements:

A close up of a puppy with sharp focus on its cute face

Limit Unwanted Elements

Add negative prompts to exclude unnecessary elements:

A burger advertisement, focus on the burger, extremely detailed, cinematic lighting, volumetric rays, award winning photograph, best quality, (negative prompt) no people, no utensils, no drinks, no background

Advanced Prompting Techniques

Here are some more advanced techniques for optimizing Stable Diffusion prompts:

Split Prompts

Split long prompts into multiple segments to emphasize different elements:

(A) A close up portrait of a beautiful woman, soft lighting, 35mm depth of field. (B) She has bright green eyes and dark red hair. (C) She is wearing an elegant black evening gown with intricate gold accents. 

Weighted Prompts

Increase weights for important keywords:

(1.5) An oil painting of a cute puppy

Nested Prompts

Nest prompts to isolate components:

A macro photograph of a ladybug on a (flower with large pink petals)

Story Prompts

Build a narrative to render logical, sequential images:

(1) An astronaut stepping onto an alien planet for the first time, (2) she takes samples of the purple trees and orange soil, (3) looking up at three moons in the green sky

Prompt Websites

There are many websites that offer Stable Diffusion prompts across different categories:

These can be great sources of inspiration when designing your own prompts. You can browse popular prompts for ideas or search for specific styles and subjects.

Conclusion

Crafting effective prompts is key to generating stunning images with Stable Diffusion. Follow prompt formatting guidelines, leverage prompt engineering techniques, and browse prompt websites when ideating prompts. Experiment with different prompts and continuously refine them based on results. Mastering prompt design unlocks the full creative potential of AI image generation with Stable Diffusion.

Useful Websites