Lesson 4 of 5 · 6 min

Creating Images With AI

Describe subject, style, lighting and composition to get the picture you imagine, improve it one change at a time, and learn what to avoid.

In this lesson

  • Write an image prompt with subject, style, lighting and composition
  • Improve an image by changing one thing at a time
  • Avoid real people, logos and copyrighted characters

How image generators work

An image generator turns a text description into a picture. You write what you want to see, and after a few seconds you get one or more images. Some chat assistants can create images directly in the chat. There are also dedicated tools such as Midjourney and Ideogram, and image features inside design apps like Canva.

The AI does not search the internet for an existing photo. It creates a new image based on patterns it learned. That means the result can be surprising, beautiful or slightly strange, for example a hand with six fingers or a sign full of spelling mistakes.

The four building blocks of an image prompt

A good image prompt describes more than the main object. Think about four things:

  • Subject: what is in the picture and what is happening? “An old fisherman repairing a net on a wooden boat.”
  • Style: what should it look like? A photo, a watercolour painting, a cartoon, a 3D render, a pencil sketch.
  • Lighting: soft morning light, bright midday sun, a warm sunset, the cool glow of a lamp at night.
  • Composition: close-up or wide view, from above or at eye level, subject in the centre or to one side, a simple or detailed background.

Here is the difference this makes when you put all four together.

Before: “A cat.”

After: “A fluffy orange cat sleeping on a windowsill, soft morning light coming through the window, watercolour illustration, close-up, green plants blurred in the background, calm and cosy mood.”

The first prompt gives you a random cat. The second gives you an image much closer to the one you imagined.

Iterate: change one thing at a time

Your first image is rarely perfect. Instead of writing a completely new prompt, change one part and compare the results. This way you learn which words have the biggest effect.

  • If the mood feels wrong, change the lighting.
  • If it looks too realistic or too childish, change the style.
  • If the subject is too small, ask for a close-up.
  • If something unwanted appears, say what you want instead, for example “an empty quiet street” rather than only “no people”.

Many tools also let you make variations of an image you like, or edit just one area of it. Save the prompts that work well so you can reuse them later. If you need words in the image, keep them very short and check the spelling, because AI often struggles with letters.

What to avoid

Image tools are fun, but some uses can cause real harm or legal trouble. As a simple rule, avoid:

  • Real people: do not create images of real people, famous or not, especially in situations that never happened. This can hurt reputations and breaks the rules of most tools.
  • Logos and brands: do not copy company logos or make images that look like official adverts or products.
  • Copyrighted characters: well-known cartoon, film and game characters belong to their owners. Create your own original characters instead.
  • Misleading images: when you share an AI image, be honest that it was made with AI, especially if it looks like a real photo.

Instead of asking for a famous superhero, describe your own: “a young inventor with a green scarf and round goggles, flying a home-made glider over rooftops.” It is more original, and it is truly yours.

Practice

Open any AI assistant you use and paste the prompt below, replacing the parts in [BRACKETS]. If your assistant cannot create images, it will still write the prompt for you to use in an image tool later.

Prompt
Help me write a detailed image prompt for [IMAGE_IDEA] in a [STYLE] style. Describe the subject, style, lighting and composition in one paragraph of under 60 words. Then give me two variations that change only the lighting or only the camera angle. Do not include real people, logos or famous characters. If you can create images, make the first version too.
[IMAGE_IDEA]
e.g. a small bakery on a quiet street, a fox reading a book
[STYLE]
e.g. watercolour, photo, flat cartoon, pencil sketch
See an example result

Main prompt:
A small family bakery on a quiet cobbled street, fresh bread and pastries in the window, a striped awning above the door, watercolour illustration with soft edges, gentle early-morning light and long shadows, wide view at eye level, a bicycle leaning against the wall, warm and peaceful mood.

Variation 1 (lighting):
Same scene, but in the blue light of evening, with the bakery window glowing warm yellow from inside.

Variation 2 (camera angle):
Same scene, seen from slightly above, as if from a window across the street, so the cobbles and awning fill more of the picture.

What to notice: Every part of the recipe is there, and each variation changes just one thing, so you can see what that change does. If the result feels crowded, remove details rather than adding more.

Key takeaways

  • Subject, style, lighting, composition: describe all four
  • Change one thing at a time and save prompts that work
  • Create original characters; avoid real people, logos and brands