From pencil sketch to refined AI-generated hummingbird artwork: a black, green and gold visual illustration of iterative image prompting

How to Create a DALL·E Prompt That Works (2026 Guide)

Gary Whittaker

Jack Righteous · AI Visuals for Creators · Updated October 2026

A good image prompt is not the longest description you can write. It is a clear creative direction that tells the image tool what matters, what the image is for, and what you want to change next.

Important DALL·E update for 2026: This article keeps the familiar DALL·E name so earlier readers can find it, but OpenAI has retired the official DALL·E GPT from ChatGPT. For new images and edits in ChatGPT, use ChatGPT Images. The prompt-writing lessons below are useful across several image tools, although their controls and results differ. See OpenAI's current Images instructions.

Maybe you are making a cover for your first song, an image for your Shopify store, or artwork to help tell a story. You already have something in mind, but the result keeps coming back wrong. I have been there. The temptation is to add more adjectives and hope the next attempt magically understands you.

There is a better way. Treat the prompt like a brief you would give a creative partner: communicate the job, show what matters, review the result, and give useful feedback. You remain the one making the creative decisions.

1. Start with what the image is supposed to do

Before writing the prompt, answer: Where will this image be used? A square album cover, a blog thumbnail, a vertical social post and a horizontal website banner need different compositions. If you wait until the image is finished to think about the format, important details may be cropped or text may not fit.

  • Subject: What is the main thing people should notice first?
  • Purpose: Is this a song cover, tutorial image, campaign concept or product-related visual?
  • Audience: Who needs to understand or connect with it?
  • Format: Square, vertical or wide? Where must there be space for titles?
  • Feeling: Warm, reflective, energetic, peaceful, mysterious or something else?

For a useful real-world example, I will build a visual around an emerald-and-gold hummingbird for an imaginary music release. The purpose is to make the image memorable even when it is shown as a tiny cover on a phone.

2. Use this six-part prompt formula

You do not need to memorize technical language. Describe the following six elements in plain English. Include enough detail to direct the image, but do not add details that do not serve the result.

  1. Subject and action: An emerald hummingbird hovering in the air.
  2. Setting: Tropical foliage around a softly glowing gold circle.
  3. Composition: The bird is the focal point, with enough clear space around it.
  4. Visual treatment: Detailed editorial album art with rich texture rather than a cartoon.
  5. Lighting and palette: Deep green, black, warm gold and controlled highlights.
  6. Usage and exclusions: Square music-cover layout; no title text or unrelated objects.
Copy-and-edit prompt framework

Create a [format and purpose] showing [main subject and action] in [setting]. Make [focal point] the first thing viewers notice. Use [composition or camera angle] with [style or visual medium], [lighting], and [color palette]. The feeling should be [mood]. Keep [important detail] visible and leave space for [text or layout requirement]. Avoid [specific unwanted elements].

The formula is a starting point, not a rule that every prompt must contain six long sentences. If your idea is simple, the prompt can be simple. What counts is whether you communicated your creative direction.

3. Watch how a prompt gets better

First attempt: too open-ended

Prompt 1

A beautiful hummingbird in a jungle.

You will probably get a hummingbird and a jungle, but almost everything else is left to chance. Maybe it is a photograph. Maybe the bird is tiny, the background is crowded, or the colors clash with the music you are releasing.

Second attempt: give it a job

Prompt 2

Create square cover artwork for an uplifting reggae-inspired instrumental. An emerald hummingbird hovers in front of a large warm-gold circular glow, surrounded by deep green leaves. Place the bird prominently in the center-right area. Use rich black shadows, fine golden accents, and a textured, cinematic editorial-art treatment. The atmosphere should feel hopeful and rhythmic. Keep the image readable at thumbnail size, with no words or logos.

Now the image has a purpose, a central subject, a mood, a composition and a visual language. This gives you something you can review against your original intention.

Third attempt: correct one visible problem

Follow-up edit

Keep the hummingbird, gold circle, color palette and square composition. The leaves are competing with the subject. Reduce the background detail and increase the contrast around the bird's wings. Keep everything else as close as possible to the current image.

Notice what changed: I am not asking for an entirely new image. I am telling the tool what worked, what did not, and what to preserve. That is how the image starts to become your creative direction rather than a random collection of outputs.

4. Four practical prompts you can adapt today

For a music release

Square cover

Design square cover art for an independent soul-reggae single about resilience. Show a solitary lighthouse casting a warm beam across dark ocean waves just before sunrise. Keep the lighthouse as the focal point. Use cinematic contrast, deep navy, muted gold and subtle film-grain texture. No people, text or logos. The visual must remain clear as a small music-platform thumbnail.

For a blog article

Editorial feature image

Create a clean, horizontal editorial image about independent creators planning their work with AI. Show a simple notebook with thumbnail sketches, headphones and a tablet displaying abstract color studies on a light desk. Natural window lighting, realistic materials, gentle shadows and a calm modern atmosphere. No readable text or logos. Leave uncluttered space near the left edge for a headline if needed.

For a social campaign

Vertical social post

Create a vertical visual for a creator's new-project announcement. Show bold sculptural paper shapes rising upward against a warm ivory background, with a restrained black-and-gold palette. Use a simple central composition, strong contrast and plenty of empty space in the upper third for later typography. No generated words, people or branding.

For a product idea, not a customer-facing product claim

Visual concept exploration

Develop an editorial concept image for premium music-inspired stationery: a blank black notebook, matching pen and abstract gold sound-wave motifs arranged on a bright neutral tabletop. Soft natural lighting, tasteful styling and clear composition. Treat this as creative inspiration only, not a depiction of a manufactured product. No labels, writing or logos.

If you sell physical products, use accurate photographs or verified mockups for the actual listing. Do not present fictional AI-generated features, packaging or finishes as something your customer will receive.

5. How to make changes without starting over

Current ChatGPT Images lets you ask for edits after generating an image. You can also open an image and select an area to change. Targeted selections are helpful, but they are not always perfectly precise: edits can sometimes affect nearby details. Check the whole image again after making a change. OpenAI explains the current editing controls here.

Try directing changes this way:

  • Wrong subject size? “Make the main object 20% larger, preserving its position and the overall square composition.”
  • Background too busy? “Simplify the scenery. Keep the main subject sharp and reduce competing detail.”
  • Wrong mood? “Keep the composition, but replace the cold blue lighting with warmer late-afternoon light.”
  • Need space for a title? “Preserve the subject and move it slightly lower so the upper third remains clear.”
  • Something must stay? State exactly what to preserve: palette, subject, camera angle, texture or proportions.

One useful rule: change the biggest problem first. When you ask for seven major corrections in one sentence, it becomes harder to tell which change helped and which introduced another problem.

6. What to do when your prompt still fails

It keeps looking generic. Add details that are meaningful to the story, not just adjectives such as “amazing,” “epic,” or “award-winning.” Describe the physical setting, the main visual metaphor, or the relationship between objects.

The layout looks wrong. Ask for a specific focal point, viewpoint and arrangement. Say where to leave open space. Choose the image format with the tool's aspect-ratio control when available.

One detail is incorrect. Try a localized edit and explicitly preserve everything else. Inspect the result for unintended changes.

The style changes across a series. Reuse the same palette, lighting rules, composition principles and approved reference material where the tool supports it. For the broader workflow, read How to Keep AI Images Consistent With Your Brand.

Generated text is imperfect. Some image tools can produce readable words, but always verify spelling, positioning and legibility. For important cover titles, prices, disclaimers or calls to action, I recommend checking and finishing typography in a design editor.

7. Review the result before you publish

My image-prompt QA checklist

  • Does the image fulfill the original job rather than simply look attractive?
  • Is the subject clear at mobile thumbnail size?
  • Are hands, objects, proportions, reflections and backgrounds believable where realism matters?
  • Are colors, layout and atmosphere consistent with the creator's visual identity?
  • Is any writing correct, large enough and high-contrast?
  • Does the artwork avoid misleading likenesses, fabricated endorsements and inaccurate product details?
  • Have I checked relevant tool terms, third-party rights and platform rules before using it commercially?
  • Did I save the useful prompt, approved output, edit decisions, and exported formats?

A tool's permission to use its output is not the same as a guarantee of exclusive copyright protection. Human creative contribution, third-party content, trademarks and likeness rights may raise separate questions depending on the jurisdiction. For general U.S. guidance, see the U.S. Copyright Office's AI information.

Keep learning: turn one good image into a useful system

One great prompt is a starting point. Your next challenge is making the image fit a larger body of work without losing the personality that makes it yours.

Here is what I want you to remember: the prompt is not supposed to replace your imagination. It is how you explain your imagination to the tool. Give your idea a purpose, make one good decision at a time, and keep improving the result until it represents something you actually want to put your name on.

This is an educational guide. Product availability and interfaces may change; follow current platform instructions. Rights information is general and not individualized legal advice.

Back to blog

Leave a comment

Please note, comments need to be approved before they are published.

articleall levelsHow to Use Jack Righteous
On this page

    Keep Jack Righteous in your Google results

    Make Jack Righteous a preferred source.

    Google can highlight preferred publications more prominently for you in Top Stories, AI Mode and AI Overviews when those features are available.

    The Righteous Beat

    Get the week’s most useful creator guidance, platform changes and free resources.

    Join the free newsletter →