A Guide to Writing Effective AI Art Prompts

|

AI-generated images have completely changed how we create content by being cost-efficient, customizable, high quality, and bring new creative ideas. With just a single sentence prompt, you can create an endless cascade of unique and captivating visuals ready to use to promote your products or services. Yet, beneath the surface of this magical technological leap lies a sharp distinction – a stark contrast between those who casually dip their toes into the realm of AI-generated art and the accomplished practitioners who have masterfully honed their craft into what is now being referred to as AI prompt engineering or simply prompting.

This emerging discipline involves not only typing a few words into a text field but also understanding and wielding the language of AI to craft mesmerizing art that continues to transcend our expectations of the possible. Let us take you on a journey that will navigate you from the realm of a curious hobbyist to that of a proficient AI prompt engineer skilled at generating images with these powerful new tools.

How do AI image generators think?

First, it’s essential to grasp how these image generators operate. While each model interprets the input differently, they are all the same in this regard: unlike humans, they lack an inherent understanding of what words really mean. AI image makers use iterative improvements based on vast data to “learn” and make images, but this comes from mathematical calculations, rather than human-like thinking.

With this in mind, a precise text prompt is crucial when working with AI image generators because, even without understanding language, they rely on clear guidance to produce accurate and desired visual outputs. Some tips for creating an effective input include:

  1. Be clear and specific. Have a clear goal in mind when you articulate what you want the AI to create. Use precise language to describe the elements you want using terms that describe colors, styles, subjects, and compositions. The more specific you are, the better the AI will understand your vision.
  2. Use visual references. Have an idea of art styles or even names of specific artists you want the AI to emulate. This can be a powerful detail to include because not only does it provide something specific for the AI to comprehend your intentions, but it is also concise.
  3. Experiment with constraints. You won’t likely get the result you want on the first click. Like working with a real artist, it’s part of the process. You can steer the AI in the right direction by providing certain boundaries and details so that future results will stop a specific habit you are noticing it repeating.
  4. Understand the AI’s capabilities. Familiarize yourself with the strengths and limitations of the AI model you’re using. AI models excel at different styles or specific subjects, so you may have to tailor your prompts a certain way, or you may even decide to seek out a new one entirely! Some powerful models even utilize a unique language that enables prompts to do very special and specific things.
  5. See what others are doing. Explore prompts that others have been using that generate the results that you’re looking for. There are many online communities where users share AI-generated artwork and the prompts they used to create them. Some even provide the seed number to an AI model so you can continue where their AI left off.

Prompting with style. Examples in the wild.

Word choice is key to the descriptors you’ll want to include in your prompts. Here are a few samples.

Describe specific subjects

The foundation of every prompt is the subject, the central theme, object, or concept. Provide clear and specific details about the subject’s attributes, such as its physical appearance, emotional context, or any stylistic elements you envision. Including some adjectives that highlight things such as mood, texture, and color. Adding spatial relationships can also greatly enhance the AI’s understanding of your intent.

  • “A muddy red Lamborghini”
  • “A beautiful Greek goddess wearing a white toga”
  • “An old wizard wearing elaborate robes studying a book near a candle”
  • “A mysterious stranger wearing a black coat holding a torch in the middle of a foggy cemetery at night”

Word choice is key to the descriptors you’ll want to include in your prompts. Here are a few samples.

Include color and mood descriptors.

A more refined prompt incorporates color and mood descriptors to help focus the AI on generating images. The transformative power of such descriptors can greatly change the image’s overall treatment and tone to fit better what you had in mind.

Use specific moods and color descriptors to fine-tune the image, such as:

  • Vivid, vibrant, or colorful
  • Pastel
  • Earth tones
  • Monochromatic or black and white
  • Cool or warm
  • Muted

Prompt used: “An old, wooden shipwreck on a beach in __________.”

“pastel colors”
“warm tones”
“very vibrant colors”
“only black and white”

Add details about the composition

Compositional descriptors will help position subjects and frame your AI artwork the way you intend. Like a director on a film set, you want to describe the scene to the cameraman on how you want things displayed.

Use specific references to angles and shot types to help guide the framing of your AI-generated images, such as:

  • Portrait
  • Off-center
  • Overhead
  • Close-up
  • Wide-angle

Prompt used: “A(n) __________ of a bowl of colorful, assorted fruits on a table.”

“off-center shot”
“wide-angle shot”
“close-up”
“overhead view”

Make references to specific styles or artists/photographers.

Arguably, one of the most powerful descriptors you can include in your text prompt is to refer to specific artistic styles for the image. This descriptor can be broad, indicating a specific genre of art or—more effectively—mentioning a specific artist or group of artists that match what you imagine your final image would look like.

Specific artist names are exceptionally effective due to their being very clear and specific and usually short in nature. There is very little room for the AI to be wrong in this regard as it searches its database for relevant data it is trained on. It is, however, better used when referring to popular artists or art style terminology, as the AI may not have been trained on more obscure terms or names.

Examples of effective references are:

  • Cartoony
  • Realistic
  • Illustration
  • Psychedelic
  • Renaissance painting
  • Popular historical artists with a distinct style, such as Vincent van Gogh or Salvador Dali
  • Recent popular artists with a niche style

Prompt used: “A __________ portrait of an astronaut wearing a full helmet.”

“very psychedlic”
“photorealistic”
“sketchy, illustrated”
“classical oil painting”

Get Digital Marketing Results

Here at Digital Agility Media, we’re committed to producing remarkable creations using the latest technologies and trends. We continuously provide excellence for our clients through this combination of cutting-edge technology and years-tested artistry. Contact us to see how we can help you use our mix of tools and experience to propel your next marketing or web project.

Oliver Ooytas Avatar