By the end of this guide you will have a repeatable way to write prompts for AI image generators: a short structure you can reuse, a method for changing one thing at a time, and a saved set of wording that works for your style. The job is done by a text-to-image tool, and the skill is in how you describe the picture.
If you want to write prompts for AI image generators and get closer to what you imagined on the first few tries, the key is specificity and iteration rather than long, clever phrases. This guide uses seven tools from our Image generation category to show which prompt-related features are available where.
Key takeaways
- A good prompt names the subject, the setting, the style, the lighting and the mood, in plain language, in that order.
- Change one element at a time between generations so you learn what each word does.
- Tools differ: some offer negative prompts and seeds, some rewrite your prompt for you, and some strip special characters, so check what your tool supports.
- Prompts are not a guarantee. Always check the image for flaws, and avoid asking for a real person's likeness or a living artist's style without permission.
What you need before you start
- An image generator account or a free web tool. Any tool in the comparison below will do.
- A goal for the image. Where it will be used and what it should show.
- A place to keep notes. A text file or spreadsheet where you paste prompts and results.
- Time. Around half an hour to learn the structure, then a few minutes per image.
- Difficulty. Easy. No technical background is needed.
- Care with sources. Avoid prompts that copy a living artist's name or a real person's likeness. Check the terms of your tool for what is allowed.
Which kind of AI tool fits this job
Prompting works best when the tool matches how you like to work.
Tools with deep prompt controls. Stable Diffusion AI lists negative prompt, CFG scale and seed controls, a Magic Prompt generator, prompt galleries and artist style references. It is free, with rate limits. Midjourney turns written prompts into images and short videos on the web or in Discord, and lets you browse a member gallery of other people's creations for ideas.
Tools that help with wording. AI Image Generator offers prompt-writing guides that explain how to combine a subject, setting, art style, lighting and atmosphere, plus mood and lighting settings and an optional reference image or sketch. Enterpix has a Prompt Enhancer that adjusts your wording and generation settings automatically, and saves the prompt and settings in each file's metadata.
Tools for text inside images. Ideogram focuses on designs that include legible typography, with image reference, prompt-based editing and character consistency.
Platforms where you can learn from others. Deep Dream Generator has a community gallery showing trending and latest creations with their prompts and settings, and visual prompts that turn an uploaded image into a prompt. Leonardo.Ai offers more than 80 models on paid plans, which lets you test the same prompt on different models. Its free daily token plan is limited to the Auto model.
Step 1: Describe the subject first
Start with what the picture is about, in plain words.
- Write the main subject as a short noun phrase: "a red bicycle", "an elderly fisherman", "a tiny cabin".
- Add one action or state: "leaning against a brick wall", "mending a net".
- Keep it to one main subject for now. Crowded scenes are harder to control.
- Avoid abstract words like "beautiful" or "amazing" that do not tell the model what to draw.
Check it worked by reading the subject line to someone and asking them to sketch it.
Step 2: Add setting, composition and light
Context turns a subject into a scene.
- Name the place and time: "on a cobbled street at dusk".
- Say how it is framed: close-up, wide shot, from above.
- Describe the light: soft morning light, harsh midday sun, neon at night.
- Add the mood in one or two words, such as calm or tense.
AI Image Generator's guide describes this as combining a subject, setting, art style, lighting and atmosphere.
Check it worked by confirming each sentence adds something visible rather than a vague feeling.
Step 3: Choose a style
Style tells the model how the image should look.
- Pick a medium: photograph, watercolour, pencil sketch, flat vector, 3D render.
- Optionally add a period or genre, such as mid-century poster or science-fiction concept art.
- Use a style reference if the tool supports one. Stable Diffusion AI lists artist style references, and Ideogram lists image reference.
- Be cautious with living artists' names. Naming a style by medium and period is safer and usually clearer.
Check it worked by seeing the same subject come back in a clearly different look when you change only the style line.
Step 4: Use the controls your tool offers
Settings can matter as much as words.
- Look for a negative prompt to exclude things you do not want. Stable Diffusion AI lists negative prompt controls, while Enterpix removes parentheses, braces and brackets from prompts, so weighting syntax does not work there.
- Look for a seed to repeat a result. Stable Diffusion AI lists seed and CFG scale controls.
- Try prompt helpers. Enterpix's Prompt Enhancer adjusts wording and settings, and Stable Diffusion AI offers a Magic Prompt generator.
- Read the tool's own guide before borrowing syntax from another tool, since special characters and parameters differ.
Check it worked by running the same prompt twice with and without a setting and noting the difference.
Step 5: Generate, compare and change one thing
Iteration is where prompts improve.
- Generate several images from the same prompt.
- Compare them with your goal and pick the closest.
- Change one element: the lighting, the framing or the style.
- Generate again and compare, and note which edit helped.
- Watch your allowance. Ideogram free accounts get 30 credits per week and Leonardo.Ai's free plan allows 150 fast tokens per day.
Check it worked by being able to say which word or setting caused each change.
Step 6: Fix common flaws
Know the patterns that tend to go wrong.
- For garbled lettering, shorten the text and try a tool built for typography, such as Ideogram, then proofread.
- For odd hands or faces, simplify the pose or crop closer, or use an editing feature to repaint the area. Stable Diffusion AI lists inpainting to repaint part of a picture.
- For a cluttered scene, remove secondary subjects from the prompt.
- For the wrong mood, adjust the lighting and colour words rather than adding more adjectives.
Check it worked by having an image where the problem areas are fixed or cropped out.
Step 7: Save a prompt library
A saved structure saves time later.
- Copy successful prompts into a note with the tool, the settings and a thumbnail.
- Turn them into templates with blanks for subject and setting.
- Use stored metadata where it exists. Enterpix saves the prompt, negative prompt and model configuration in each file's metadata, which helps you recreate a result.
- Browse community prompts for ideas, such as Deep Dream Generator's gallery with prompts and settings, and adapt them rather than copying whole prompts.
Check it worked by recreating a past image from your notes alone.
Tools for this job compared
| Tool | Good fit for | Pricing | Stated limits |
|---|---|---|---|
| Midjourney | Prompt-driven artwork on web, Discord, iOS and Android | Paid, from $10/mo (checked 24 Oct 2026) | No free plan; unused Fast GPU time does not roll over; Stealth mode only on Pro and Mega |
| Ideogram | Prompts that include readable text, with editing | Freemium, paid from $15/mo (checked 24 Oct 2026) | Free accounts get 30 credits per week; images public by default unless Plus or Pro; credits do not roll over |
| Stable Diffusion AI | Learning prompts with negative prompt, seed and CFG controls | Free | No-account UI fixed at Stable Diffusion 1.5 and 512 px; shared GPU; rate limited |
| AI Image Generator | Beginners using prompt guides and a reference sketch | Free | Reference images limited to 1 MB; resolution, usage limits and licensing not described |
| Enterpix | Automatic prompt enhancement and saved metadata | Free | In beta; brackets and parentheses removed so weighting syntax does not work |
| Deep Dream Generator | Learning from community prompts and settings | Freemium, paid from $9/mo (checked 24 Oct 2026) | Every generation uses energy; free users get 3 Pro-model images per day; commercial rights need a paid plan |
| Leonardo.Ai | Testing one prompt across many models | Freemium, paid from $12/mo (checked 24 Oct 2026) | Free plan allows 150 fast tokens per day and only the Auto model; Free plan creations are public |
Prices and plan limits change, so confirm the current terms on each vendor's site before you subscribe.
Common mistakes to avoid
- Packing in dozens of adjectives. Long lists of style words dilute each other and the model may ignore half of them. Fix it by writing a clear subject, setting, style and light, and adding words only when a result needs them.
- Changing everything at once. If you rewrite the whole prompt each time you never learn what helped. Fix it by changing one element between generations and noting the result.
- Copying syntax from another tool. Weighting symbols and parameters differ, and Enterpix strips brackets entirely. Fix it by reading your tool's own prompt guide first.
- Asking for a real person or living artist. This raises consent and rights problems and may be refused. Fix it by describing the look in general terms, such as medium, period and mood.
- Not checking the output. Lettering, hands and small objects can be wrong even when the overall picture looks right. Fix it by zooming in on every image before you use it.
How we put this together
This is an evidence-based guide built from each tool's listing in our directory, which is compiled from the vendor's own site and pricing page. It is not a review based on using the tools. We don't rank tools, and nobody paid to be included. Some outbound links may be affiliate links; see our disclosure.
Frequently asked questions
Is there a free way to practise writing image prompts?
Yes. Stable Diffusion AI is free with rate limits, Ideogram free accounts get 30 credits a week, and Leonardo.Ai's free plan gives 150 fast tokens per day. The AI Image Generator site also publishes prompt-writing guides, though it does not describe its usage limits.
How long should an image prompt be?
Long enough to name the subject, setting, style and light, and no longer. A few clear sentences usually beat a long list of keywords. Add detail only to fix something specific that the first results got wrong.
Do negative prompts work in every tool?
No. Stable Diffusion AI lists negative prompt controls, while other tools have different options or none. Check your tool's own guide, and use positive wording ("empty street") when a negative option is not available.
Can I reuse a prompt across tools?
Plain-language prompts transfer well, but settings and special syntax do not. Enterpix, for example, removes parentheses and brackets from prompts. Keep a plain version of each prompt and adapt the extras for each tool.
Can I use someone else's prompt?
Community galleries, such as Deep Dream Generator's, show prompts and settings so you can learn from them. Use them as a starting point and adapt them. Check the tool's terms, and do not use them to imitate a living artist or a real person's likeness.
Compare more options in our Image generation category. Once your prompts work, see how to generate images from text with AI.
