Have you ever typed an idea into an AI image tool and got something that looked nothing like what you imagined? You're not alone. The good news: writing a great prompt isn't a secret skill. It's a simple habit. Once you know what the AI needs to hear, your results get better almost straight away.
This guide walks you through a simple formula, shows you real before-and-after examples, and covers the mistakes most beginners make.
Why short prompts give random results
AI image tools fill in every detail you don't give them. If you type "a woman in a city", the tool has to guess her age, clothes, pose, the time of day, the weather, the camera angle, the colours and the style. Each guess is a coin flip, and that's why you get something different from what you pictured.
A good prompt doesn't need to be long. It just needs to answer the questions the AI would otherwise guess.
The 6-part prompt formula
Think of every prompt as six building blocks. You don't need all six every time, but the more you include, the more control you have.
1. Subject: who or what is in the picture?
Be specific. Not "a dog", but "an old golden retriever with a grey muzzle". Add age, expression, clothing or texture where it matters.
2. Action and pose: what are they doing?
"Sitting on the floor with knees pulled up", "laughing while looking away from the camera", "walking through puddles". Action makes images feel alive.
3. Setting: where is it happening?
"A narrow alley with peeling blue paint", "an empty night train", "a minimalist white studio". The background sets the story.
4. Lighting: how is it lit?
Lighting changes everything. Try:
- Golden hour: warm, soft and flattering
- Hard light: strong shadows and drama
- Soft window light: calm and natural
- Neon light: colourful, moody and urban
- Chiaroscuro: deep darkness with one bright area, like an old painting
5. Camera and composition: how is it framed?
Even if no real camera is involved, photography words guide the AI: "close-up portrait", "wide shot", "shot from a low angle", "shallow depth of field" (blurry background), "85mm lens" (classic portrait look).
6. Style and mood: what should it feel like?
"Editorial fashion photography", "cinematic film still", "soft watercolour illustration", "moody and quiet", "bright and playful".
Before and after: see the difference
Before:
a girl on a train
After:
A young woman with curly dark hair sitting alone on an empty night train, wearing an oversized graphic T-shirt and denim shorts, misty light coming through the windows, neon city lights outside, cinematic film still, moody and calm
The first prompt leaves almost everything to chance. The second tells the AI the subject, setting, lighting, style and mood, so the result matches the picture in your head.
Another example:
Before:
product photo of perfume
After:
A glass perfume bottle with a gold cap standing on wet black stone, small water droplets on the glass, soft rim light from behind, dark background, luxury product photography, close-up, sharp focus
Tool-specific tips
Different tools understand prompts in slightly different ways. Features change often, so always check your tool's own help pages for the latest options.
Midjourney responds well to descriptive phrases separated by commas. It also supports extra settings added at the end of the prompt, for example:
--ar 2:3 sets the shape (aspect ratio) of the image; 2:3 is tall, 16:9 is wide
--no text asks it to avoid something (here, text in the image)
--seed 1234 helps you recreate a similar result later
Gemini, ChatGPT and other chat-based tools understand normal sentences very well. You can write as if you're describing the picture to a friend, and say the shape in words: "make it a tall, phone-wallpaper shape".
For any tool: if something keeps going wrong, say what you don't want. Many tools have a "negative prompt" box for this, for example "blurry, extra fingers, watermark".
5 common beginner mistakes
- Too vague. "Beautiful landscape" means nothing specific. Which place, season, time of day?
- Too many ideas at once. A dragon, a city, a beach and a sunset in one image often turns messy. Pick one main idea.
- Forgetting the lighting. Lighting is the fastest way to make an image look professional.
- Giving up after one try. Professionals generate several versions and change one thing at a time.
- Copying other artists' names into prompts. Describe the look you want instead ("bold colours, thick brush strokes, dreamy mood"). It's more respectful to artists, and you'll learn more about style.
How to improve a prompt step by step
When a result is close but not quite right, don't start over. Change one thing at a time:
- Keep the prompt that's closest.
- Change only the lighting, or only the setting, or only the pose.
- Compare the results side by side.
- Keep what worked and repeat.
This is how you learn which words really change the image.
Learn faster by remixing real prompts
One of the quickest ways to improve is to study prompts that already work. Every image on Imagetheus comes with the full prompt used to create it. Copy one, change the subject or the setting, and see what happens. Each prompt page also has a remix guide to show you what to try.
Frequently asked questions
How long should an AI image prompt be?
Usually one to three sentences is enough. Length matters less than covering the key details: subject, setting, lighting and style.
Do I need to use special words?
No. Plain, clear descriptions work well. Photography words like "close-up" or "soft light" simply give you more control.
Why do I get different images from the same prompt?
Most tools add randomness on purpose so you get variety. Some, like Midjourney, let you use a "seed" number to get more similar results.
Can I use the same prompt in different AI tools?
Yes, but expect different results. Each tool has its own style, so small adjustments may be needed.
Start practising today
Pick one image idea and write it using the six building blocks: subject, action, setting, lighting, camera and style. Then generate it, change one thing, and try again. After a few rounds, you'll be surprised how much control you have.
Looking for inspiration? Browse the Imagetheus gallery and remix a prompt that catches your eye.