Stable Diffusion is a revolutionary AI system for generating images from text descriptions. To create high-quality images, users must provide detailed prompts – text descriptions of the desired image content, style, composition etc. Prompt engineering is key to mastering Stable Diffusion.
Viewing past prompts is useful for improving one’s prompt writing skills and reproducing old images. However, Stable Diffusion lacks built-in prompt history. This article provides methods to access prompt history in Stable Diffusion and examples of effective prompts.
Why Prompt History is Useful
Here are some reasons why accessing past prompts is valuable:
- Reusing and improving prompts – Viewing past prompts allows prompt details to be copied instead of rewritten from scratch. Old prompts can also be refined over time.
- Learning prompt writing skills – Analyzing previous prompts, especially those that produced good images, helps develop prompt engineering abilities.
- Regenerating lost images – If an image file is deleted, the prompt can be retrieved from history and reused to regenerate it.
- Tracking prompt iterations – Viewing a history of past variations shows the evolution of a prompt.
Methods to View Prompt History
Since Stable Diffusion lacks native prompt history, here are methods to retrieve past prompts:
1. Prompt History Browser Extension
- A community-built browser extension adds a Prompt History tab showing past prompts.
- Install it from the Extensions tab by entering the extension URL.
- After installing, restart Stable Diffusion to access the new Prompt History tab.
2. Image Metadata
- Stable Diffusion stores prompt details and settings in image metadata.
- Go to the PNG Info tab and upload a generated image.
- The prompt and parameters used to create that image will be displayed.
3. Telegram Chatbot
- The Stable Diffusion Telegram chatbot has a
/historycommand showing recent prompt activity.
- Add a search term after
/historyto filter results e.g.
- Click returned links to open images and prompts in the web interface.
Crafting Effective Prompts
Here are some examples of effective prompts and tips for writing better prompts:
Specifying a Clear Subject
- Clearly state the primary subject e.g. “a portrait of a girl”, “a still life painting of fruit”.
Describing Visual Details
- Use adjectives and visual descriptors for colors, textures, shapes etc.
- For example, “a bright red bird with smooth feathers” or “a swirling abstract acrylic painting”.
- Specify the art style e.g. minimalist, deco, cubist.
- State the medium used e.g. colored pencil, vector graphics.
- Describe the composition e.g. symmetrical, rule of thirds.
Using Brackets to Weight Keywords
- Put keywords in brackets to make them more influential, like (75%) or (highly detailed).
Providing Quality Reference Images
- Upload sample images to guide the output visuals.
Employing Negative Prompts
- Add negative prompts prefaced by “-” to exclude unwanted elements e.g. “-text -logo”.
Here are some full prompt examples covering various styles and approaches:
Realistic Portrait Focusing on Lighting
A realistic close up portrait of a beautiful young woman with smooth glowing skin, looking into the distance wistfully. Dramatic lighting illuminates half her face. High key lighting, depth of field background blur.
Whimsical Landscape Painting With Animation
A vibrant, whimsical landscape digital painting of a meadow with dancing foxes and rabbits. Watercolor and oil paint texture. Smooth cartoon shading. Cinematic color grading. Animated with swaying trees and flowing grass. 4k high resolution. Trending on ArtStation.
Minimalist Geometric Logo Design
A professional minimalist logo for a technology company. Geometric shapes integrated together in a seamless design. Modern and sophisticated style. Vector image.
Useful Resources on Prompt Engineering
Here are some websites with more AI prompt examples and guides:
Accessing prompt history in Stable Diffusion is vital for improving prompt writing skills and reproducing old images. Install the Prompt History extension, check image metadata, or use Telegram bot commands. Analyze past prompts, especially effective ones, to incrementally enhance new prompts over time. Use the examples and resources provided to continue honing abilities.