Type “a nice picture of a dog” into an AI image tool and you will get a dog. It will probably be fine, and also very similar to what everyone else gets. The gap between a generic result and the image you pictured in your head is almost always the prompt: how clearly you described the subject, the setting, the style and the mood.
This guide is for complete beginners. You will learn how image prompts work, a simple structure you can reuse, how to describe style, light, composition and colour, how to fix common problems such as odd hands and unwanted text, how to improve an image step by step, and how to use AI images responsibly. No design background is needed, and the advice applies across most image tools.
Key takeaways
- A strong image prompt describes the subject, the setting, the style, the lighting, the composition and the mood in plain, specific language.
- Start simple, look at what you get, then change one thing at a time so you learn what each change does.
- Tell the tool what you want, and where supported, what to leave out, but do not rely on negative wording alone.
- Expect flaws such as distorted details or garbled text, and plan to regenerate, edit or fix them in other software.
- Check the tool’s terms on commercial use, avoid imitating living artists or real people, and be honest about AI-made images.
How image tools read a prompt
An AI image tool turns your words into a picture by drawing on patterns it learned from a very large collection of images and descriptions. It does not understand your sentence the way a person does. It matches your words to visual patterns and builds an image that fits them.
Why vague prompts give generic results
If you write “a woman in a city”, the tool fills in every unstated detail with the most typical choice it knows: a typical face, a typical street, typical light. Every detail you leave out is a detail you hand to chance. The more you specify, the more control you keep.
Why more is not always better
Specific does not mean long. A prompt stuffed with thirty descriptive words can confuse the tool, and some details will be ignored or will clash. Aim for a clear picture with a handful of well-chosen details.
Every result is a draw
Run the same prompt twice and you will usually get two different images. That is normal. Treat each result as one draw from a range of possibilities, and generate several before judging the prompt. The same idea of treating output as variable appears in our guide to writing better prompts, and it applies to pictures as well as text.
Tools differ
Different tools read prompts differently. Some prefer natural sentences, others respond to short phrases, and some offer extra controls such as aspect ratio or a way to upload a reference image. Read the tool’s help pages and run a few test prompts to learn its habits.
A simple prompt structure
You do not need a secret formula, but a consistent order helps you remember what to include. Here is a structure that works for most beginners.
| Part | What to say | Example |
|---|---|---|
| Subject | Who or what is in the picture | An elderly baker holding a tray of bread |
| Action or pose | What the subject is doing | Smiling and looking towards the window |
| Setting | Where it takes place | A small village bakery early in the morning |
| Style or medium | How the image should look | Warm documentary photograph |
| Lighting | The kind of light | Soft light from a side window |
| Composition | Framing and viewpoint | Medium shot, slightly from the front |
| Mood and colour | The feeling and palette | Cosy, warm tones |
Putting it together
Join the parts into a natural description: “An elderly baker holding a tray of bread, smiling and looking towards the window, in a small village bakery early in the morning. Warm documentary photograph, soft light from a side window, medium shot from slightly in front, cosy warm tones.” This reads like a brief to a photographer, and that is the right mindset.
Order and emphasis
Many tools give more weight to what comes first. Put the most important element, usually the subject, at the start, and keep the main idea in one clear sentence before adding finishing details.
Use plain words
Choose ordinary, concrete words. “Dusty”, “crowded”, “glowing” and “weathered” paint pictures. Abstract words like “beautiful” or “amazing” are too loose to guide the result.
Describing the subject
The subject is the heart of the image, and the place where beginners most often stay too vague.
Be specific about the details that matter
Instead of “a car”, say “a small, old, blue hatchback with a dented door”. Instead of “a person”, describe age range, clothing, expression and posture if they matter to your purpose. Only describe what you care about. Leave the rest open and you may be pleasantly surprised.
Describe the action
Pictures with clear action feel alive. “A cyclist climbing a steep hill, leaning forward” gives the tool a pose to work with. “A cyclist” alone may produce a stiff, static figure.
Limit the number of subjects
Prompts with many characters or objects are harder for tools to handle, and details tend to blur or merge. Start with one main subject and add secondary elements gradually. Tools are also often unreliable with exact numbers, so check counts in the result.
Be careful with real people
Do not try to produce realistic images of real, identifiable people without their permission, and avoid anything that could mislead others about what a real person did or said. We return to this in the section on responsible use.
Style and medium
Style decides whether the same subject looks like a photograph, a children’s book illustration or a technical diagram. Naming it clearly is one of the biggest improvements you can make.
| Style type | Words to try | Typical use |
|---|---|---|
| Photographic | Photograph, documentary, studio portrait, candid | Realistic scenes, product-style images |
| Illustration | Flat illustration, line drawing, children’s book style | Articles, presentations, explainers |
| Painting | Watercolour, oil painting, gouache, pastel | Artistic or decorative pieces |
| Graphic design | Minimal vector style, bold shapes, limited palette | Icons, posters, backgrounds |
| 3D | Clay render, isometric 3D, soft plastic look | Playful product or concept visuals |
| Sketch | Pencil sketch, ink drawing, storyboard frame | Concepts and early ideas |
Describe qualities instead of copying artists
It can be tempting to name a living artist’s style. That raises ethical and sometimes legal questions, and it makes your work look derivative. A better approach is to describe the qualities you like: “muted palette, thick brush strokes, simple shapes, textured paper”. You get a style of your own and avoid the problems of imitation. This is general information, not legal advice.
Keep the style consistent across a set
If you need several images that belong together, such as a series for a blog, write down a short style description and reuse it word for word in every prompt. Keep the same palette, level of detail and framing. This is a simple way to get a coherent set without any design software.
Light, colour and mood
Lighting and colour carry most of the emotion in an image. They are also easy to control with a few words.
Lighting words
- Soft, diffused light: gentle shadows, flattering and calm
- Hard, direct light: strong shadows, dramatic and high contrast
- Golden hour: warm, low sun shortly after sunrise or before sunset
- Overcast daylight: even, quiet and slightly muted
- Backlit or rim light: a glowing edge around the subject
- Studio lighting: clean and controlled, common for products and portraits
- Night with artificial light: pools of light against darkness
Colour
Name a palette instead of leaving colour to chance: “warm earth tones”, “cool blues and greys”, “bright primary colours”, “black and white with one red accent”. For brand or presentation work, describe your actual colours in words, and check the result because tools do not always match exact shades.
Mood
Words such as calm, tense, playful, nostalgic, clinical or mysterious steer the whole feel. Pair a mood with a lighting choice so they agree: a calm mood and harsh light will pull in opposite directions.
Composition and framing
Composition is how the subject sits in the frame. Borrowing a few terms from photography gives you good control.
| Term | Meaning | Effect |
|---|---|---|
| Close-up | Subject fills the frame | Intimate, detailed |
| Medium shot | Subject from about the waist up | Balanced, conversational |
| Wide shot | Subject small within a larger scene | Sets the scene |
| Overhead view | Looking straight down | Flat lays, tables, layouts |
| Low angle | Camera below the subject, looking up | Powerful or imposing |
| Eye level | Camera at the subject’s height | Natural and neutral |
| Shallow depth of field | Subject sharp, background blurred | Focus on the subject |
Leave room for text
If the image will carry a headline, ask for empty space: “plenty of empty space on the left for text”. It is much easier to add words later in a design tool than to expect the image generator to spell correctly.
Aspect ratio
Decide the shape you need before you start: wide for headers and slides, square for many social posts, tall for stories or posters. Where the tool lets you choose, set it first, since cropping afterwards may cut off important parts. For planning visuals alongside posts, see our guide to planning social media content with AI.
Refining step by step
Few people get the perfect image on the first try. A calm, systematic approach beats random tweaking.
- Write a simple first prompt. Subject, setting and style in one or two sentences.
- Generate several versions. Look at the range, not just the first one.
- Decide what is wrong. Name the problem: the colours are dull, the framing is too wide, the style is wrong.
- Change one thing. Adjust only the part that addresses the problem so you can see its effect.
- Repeat. Keep the changes that help and undo the ones that do not.
- Save the winner. Store the prompt next to the image so you can reuse it.
Keep a prompt notebook
Save prompts that gave good results, along with a note about what worked. Over time you will build your own library of phrases for light, style and framing. The principle is simple: write good prompts down.
Use reference images carefully
Some tools let you upload an example image to guide the style or composition. Only use images you have the right to use, and avoid uploading private photos of other people. Check the tool’s data terms first. Our guide to protecting company data when using AI tools applies to image uploads too.
Fixing common problems
Image tools are impressive but imperfect. Knowing the usual failure points saves frustration.
| Problem | Why it happens | What to try |
|---|---|---|
| Distorted hands or limbs | Complex shapes are hard to produce consistently | Regenerate, crop the area out, or pose hands out of sight |
| Garbled or misspelled text | Tools draw letters as shapes, not as language | Leave text out and add it later in a design tool |
| Wrong number of objects | Counting is unreliable | Regenerate or edit the image afterwards |
| Strange details in the background | Less attention is paid to areas you did not describe | Describe the background, or simplify it |
| Generic or stereotyped people | Typical patterns dominate when you give little detail | Be specific about the people and review the result critically |
| Style drifts between images | Each result is a new draw | Reuse the same style wording and settings |
| Image looks too perfect or plastic | Smooth, idealised defaults | Ask for natural texture, imperfections and everyday light |
| Ignoring part of the prompt | Too many instructions, or conflicting ones | Shorten the prompt and remove contradictions |
Use exclusions with caution
Some tools offer a way to list things to avoid. It can help, but describing what you do want is usually more reliable than listing what you do not. Writing “no people” in some tools can even make people appear, because the tool latches onto the word. Test how your tool handles it.
Edit afterwards
Treat the generated image as raw material. Cropping, adjusting colour, removing small flaws and adding text in an editor will often turn a good result into a finished one. Many tools also have built-in editing features for changing a selected area.
Prompts for everyday uses
These are adaptable examples, written to show structure. Change the details to suit your own subject.
A blog header image
- Prompt: Flat illustration of a person at a desk sorting paper cards into neat piles, simple shapes, limited palette of navy, orange and off-white, plenty of empty space on the right, wide format.
A product-style photo concept
- Prompt: Studio photograph of a ceramic mug on a pale wooden table, soft diffused light from the left, shallow depth of field, neutral background, clean and minimal.
A presentation visual
- Prompt: Minimal vector style illustration of a bridge connecting two cliffs at sunrise, calm mood, cool blues with a warm orange sky, wide shot, space at the top for a title.
A social media post background
- Prompt: Abstract soft gradient background in warm peach and pale blue, subtle paper texture, square format, no objects, plenty of calm empty space.
Using images responsibly
Generating images is easy, which makes it easy to cause problems without meaning to. A few habits keep you on the right side.
Check the licence and terms
Different tools have different rules on commercial use, ownership and attribution, and the rules may differ between free and paid plans. Read them before you use an image in client work, products or advertising. Laws on AI-generated images also vary by country and are still developing. This is general information, not legal advice.
Avoid misleading images
Do not present an AI-generated picture as a real photograph of an event, a place or a person’s actions. Realistic fake images can deceive and cause harm. Where an image could be mistaken for a real record, label it clearly as AI-generated.
Respect people and creators
Do not generate images of real people without consent, and avoid copying a living artist’s distinctive style. Avoid prompts that produce demeaning content or reinforce stereotypes, and review results with a critical eye.
Be open about AI use
If your audience would reasonably want to know that an image was AI-generated, tell them. For organisations, it is worth deciding the rule in advance. Our guide to writing an AI use policy for your team can help you include images in it.
Protect private information
Do not upload photos of customers, colleagues or children to an image tool without checking the terms and getting permission where needed.
Prompt checklist and template
Before you generate
- Do I know what the image is for and where it will appear?
- Have I chosen a shape that fits that place?
- Have I described one clear main subject?
- Have I named the style, light and mood in plain words?
- Have I left space for any text I plan to add?
After you generate
- Did I look at several versions, not just the first?
- Have I checked hands, faces, text and background details?
- Does anything look misleading, stereotyped or inappropriate?
- Have I confirmed I can use the image the way I intend?
- Have I saved the prompt and settings?
Reusable template
Copy and fill in: “[STYLE OR MEDIUM] of [SUBJECT] [ACTION], in [SETTING]. [LIGHTING]. [COMPOSITION AND VIEWPOINT]. [MOOD] mood, [COLOUR PALETTE]. [SHAPE AND SPACE FOR TEXT].” Then change one element at a time until you are happy.
Frequently asked questions
How long should an image prompt be?
Usually one to three sentences. Include the subject, setting, style, light and framing, and stop when each additional word is not clearly improving the result. Very long prompts often cause conflicts and ignored details.
Why do my images look so similar to everyone else’s?
Generic prompts pull results towards the most typical patterns. Add specific details about setting, light, colour and composition, and describe qualities you want in your own words to move away from the default look.
Can I get the same image again?
Not always. Many tools offer a way to reuse a setting that makes results more repeatable, but results can still differ. If you find an image you like, save it and the prompt straight away.
Why can AI not spell the words in my image?
Many tools generate letters as visual shapes without understanding spelling, so words often come out wrong. The practical fix is to generate the image without text and add the words later in a design or presentation tool.
Is it okay to use AI images commercially?
It depends on the tool’s terms, your plan and the laws where you work. Read the licence for the specific tool and plan, avoid copying protected work or real people’s likenesses, and take proper advice for important commercial uses.
How do I make images that match my brand?
Write down a short description of your palette, style, level of detail and framing, and reuse it in every prompt. Check each result against your brand guidelines, and expect to adjust colours and details in an editor.
Conclusion
Good image prompts are less about tricks than about clear description. Say who or what is in the picture, where it is, what style it should have, how it is lit, how it is framed and what mood it should carry. Start simple, change one thing at a time, expect flaws and fix them in an editor, and keep a record of the prompts that work.
Above all, use the tools thoughtfully: check the licence, avoid misleading or disrespectful images and be open about AI use when it matters. For help with the wider skill of clear instructions, read how to write better prompts, and if you are choosing between image tools, how to choose an AI tool offers a method for comparing them on your own tasks.
