How To Write Effective Prompts for Stable Diffusion (2024)

Images add power to brand and marketing campaigns, since consumers are more likely to remember content with a visual component. Traditional image creation methods—including photo shoots and graphic design—are great ways to create memorable, unique assets. Unfortunately, these approaches can also be time-consuming and expensive.

Start-ups and small companies might not have the resources to produce enough original imagery to support social media, email marketing, web page development, and advertising on an ongoing basis. Learn how using Stable Diffusion, an artificial intelligence (AI) tool for image creation, can help you or your marketing team produce powerful assets for a fraction of the cost of traditional methods.

What is Stable Diffusion?

Stable Diffusion is a generative AI tool focused on creating digital art. More specifically, it’s a text-to-image tool that anyone can use to generate visual assets by inputting a simple prompt. The platform operates using a deep learning diffusion model, meaning its creators trained it on a large set of text and image pairs. Unlike other text-to-image models, like Midjourney and Dall-E, Stable Diffusion is an open-source tool, meaning anyone can use, edit, or enhance the system.

Stable Diffusion offers three different license types; non-commercial and community licenses are free, while pricing is custom for enterprise licenses.

Elements of Stable Diffusion prompts

The best Stable Diffusion prompts include guidance on an image’s content and stylistic elements, such as its color, composition, and contrast. All these things affect the viewer’s understanding of an image.

Addressing multiple aesthetic components in your prompt also increases the chances that the output will meet your expectations. There are some visual elements, also known as keyword categories in Stable Diffusion, to consider including in your prompt.

  • Subject: What is the focus of your image? Include a line of instruction describing the central character or item you’d like to see.
  • Setting: What should the background look like? Specify the scenery, real or imagined, of your image.
  • Lighting and tone: Specify the atmosphere and lighting; for example, “a bright and sunny, clear summer afternoon” or “a warm and cozy interior bathed in candlelight.”
  • Color palette: Specify which colors to use. For business content creation, you may choose to input the hex codes for your specific brand colors.
  • Artistic style: Refer to a specific artistic style—for example, hyperrealism, oil painting, or ’90s cartoon—to inform what overall texture Stable Diffusion should apply to your images.
  • Reference: Referencing a specific cultural movement or artist’s name can help refine the style. For example, adding the instruction “in the style of Claude Monet” would tell Stable Diffusion to create an impressionist oil painting using green, blue, pink, and purple tones.

How to write a Stable Diffusion prompt

  1. Set the scene
  2. Add more keywords
  3. Use a negative prompt
  4. Check for common mistakes

Prompt engineering with Stable Diffusion AI is often an iterative process. Creators sometimes generate multiple images and then adjust the prompt. Here’s how to use both iterative and negative prompting techniques to create an output that matches the picture in your head:

1. Set the scene

Start out with a basic prompt describing your desired image. Address multiple keyword categories by describing the subject, setting, and color palette. Generate several images with your initial prompt and review the results.

2. Add more keywords

If the initial outputs don’t align with your expectations, use iterative prompting to build on your initial instructions. Provide new keywords and additional details to modify the image. For example, if the image is too abstract, try inserting “hyper realism” in the prompt.

3. Use a negative prompt

Negative prompts are instructions that tell the model what not to include in the generated image. A negative prompt can be a powerful tool for refining imagery by eliminating unwanted elements.

For example, say you input the following prompt: “A young woman with brown hair and blue eyes wearing a denim jacket stands in front of a vintage movie theater in a small town.” If Stable Diffusion produces an image with garbled text on the marquee, you could resubmit your instructions, this time with the negative prompt “no text in the image,” to yield a more polished final product.

4. Check for common mistakes

Look out for common mistakes in your generated images such as unorganized keywords and conflicting instructions. Stable Diffusion weights keywords based on their placement within the prompt. If elements of your prompt are missing or lacking prominence, try revising the prompt by placing the most important keywords near the front. You’ll find that adjusting the word order of a prompt can dramatically change the output.

Conflicting instructions can lead to low-quality images. For example, if your prompt includes both “standing in a crowd” and “dancing,” the AI might try to depict both of these actions, leading to an image with bad anatomy, like extra limbs or malformed hands. Read over your instructions and make sure they don’t include contradictions or redundancies.

Stable Diffusion prompt examples for business owners

You can use Stable Diffusion and other generative AI models to help you create digital assets for your branding and marketing efforts, including for social media posts, web imagery, and ad campaigns. These tools have the potential to save you time and enhance your creative processes. Consider these examples to learn how generative AI could fit into your asset creation process:

Social media content

Streamline your asset creation for social media with AI image generation. Here are four examples of prompts:

  • A smiling young man with black hair and brown eyes sits on a red and white picnic blanket with grapes and a cheese plate. Bright summer day, warm glowing light. Sharp focus.High-res image.
  • A crayon drawing of a smiling stick figure in the style of children’s artwork. Poorly drawn face, poorly drawn hands.
  • A futuristic city with flying cars flying between towering skyscrapers and neon signs in vibrant colors. Dark blue night sky with a full moon. Neon lights reflected on the pavement. Dark moody lighting.
  • Scattered assorted doughnuts fill the screen. Colorful frosting and sprinkles. Overhead photograph. Blue photo background.

Ad campaigns

Create ad imagery from scratch or generate background art and add text using a third-party design tool. Refer to your style guide and provide the model with your brand colors to create cohesive imagery. Consider creating many images and using your favorites for A/B testing. Here are four example prompts:

  • An ombre gradient of abstract shapes using the colors 3be8b0, 1aafd0, 6a67ce. Digital painting.
  • Concept art for a sci-fi-inspired fashion line. Strong effect.
  • A neoclassical building facade. White marble. Volumetric lighting. Black and white image.
  • A background of abstract shapes. Clean lines. Use the colors b3dcff, 5ecc62, 00ad45. Vector image.

Let Shopify Magic write product descriptions for you

Ever wished a product description would just write itself? With Shopify Magic—Shopify’s artificial intelligence tools designed for commerce—it will. Create product descriptions in seconds and get your products in front of shoppers faster than ever.

Learn about Shopify Magic

Web imagery

Enhance your website with custom imagery using generative AI. AI imagery can take the place of stock photos—unlike generic stock imagery, you can customize AI photos to include brand-specific details. Here are four examples:

  • A vintage photograph of a field. Soft light. Film grain. Sepia tones. No humans.
  • An animated cartoon giraffe looking over a fence. Long neck visible over the fence. Bright primary colors in the style of a ’90s cartoon. Use the Unreal Engine implementation.
  • A close-up image of a woman wearing bright pink and white metallic ombre eyeshadow. Soft smile. Studio lighting. High resolution. White background.
  • Realistic photograph of a birthday cake with colorful candles on a cake stand. Pink and white frosting and sprinkles. Studio lighting. Photo backdrop.

Prompts for Stable Diffusion FAQ

What are a few good Stable Diffusion prompts?

The best stable diffusion prompts are packed with intricate details. Effective prompts include a thorough description of the subject as well as instructions to guide the image’s tone, lighting, mood, and color story.

How long are Stable Diffusion prompts?

Stable Diffusion uses a process called tokenization to parse text prompts into cues. The maximum prompt length is 75 tokens—this translates to approximately 350 characters. Longer, highly detailed prompts tend to generate better results.

How can Stable Diffusion prompts be used in business?

Brands may be able to save time and money with generative AI technology. You can use Stable Diffusion to create graphic assets for marketing efforts, such as ad campaigns or social media posts. Anyone can input a simple text prompt and produce digital art—this method is faster and less expensive than designing images from scratch or conducting a photo shoot.

Source

Leave a Reply

Your email address will not be published. Required fields are marked *