From Idea to Image

About this series

From Idea to Image is a weekly self-tutorial in making images with AI. The text of the series is itself written with the help of AI, and it reaches you edited and published by Owrang.

Each part teaches one specific skill and ends with a simple exercise you can do the same day. The three images at the foot of this page were made with exactly the steps you are about to read.

To make an image, you do not need to be a designer. The starting point is being able to see what is in your head, and to describe it.

Anyone who has ever wanted a poster, pictured a book cover, imagined an image for a poem or a scene for a film knows this gap: the picture is clear in the mind, but the page is still blank.

For a long time the only way across was to know how to draw and design yourself, or to hand the work to someone who could. Today there is another way. You can describe the picture in ordinary words, in Persian or English, and very soon see a first version of it.

At this stage the most important skill is not drawing. What matters is knowing exactly what you want.

This series practices that skill step by step. One step each week, and at the end of each part an image you made yourself.

Before the software, five questions

Choose the picture in your head and answer these five questions in plain sentences.

What is the subject?
Instead of "a street," write "an empty side street in Tehran in the 1970s."
Where are we, and when?
Night or morning? Winter or summer? Indoors or outside?
What should the viewer feel?
Calm, threat, nostalgia, joy? For now choose one feeling, not four at once.
What must be in the picture, and what must not?
For example: a single figure under a lamp. No cars. No visible violence.
Where will the image be used?
A wide poster? A square Instagram post? A book cover? The shape of the frame affects the composition.

Now write the five answers together as one short paragraph.

That paragraph is the first stage of making your image, before you have even turned on the computer.

If you like, draw a very simple sketch on paper as well. A few lines and a few boxes are enough. It does not need to be beautiful. Just mark roughly where things go. We will come back to that sketch in later parts.

Why ChatGPT first?

For the first exercise we use ChatGPT. Not because it is necessarily the best image tool, but because it understands ordinary language well, including Persian, and after the image is made you can talk to it about the result right there.

At this stage our goal is not to learn a piece of software. The goal is to learn the language of images through conversation and trial.

Other tools, including Midjourney, will be introduced in later parts of this series.

Step by step

Step 1
Go to ChatGPT and sign in to your account.

Step 2
Put the paragraph you wrote in the message box, and before it write:

"Make an image of this scene."

Our example is this:

"Make an image of this scene. An empty side street in Tehran in the 1970s, late at night. A single figure stands under a street lamp. The mood is threatening but nothing violent is shown. Minimal, quiet, realistic. Wide format."

The resulting image is version one of our exercise, and you can see it at the bottom of this page.

Step 3
Once the image is made, do not rush to make a new one. First look at the one you have.

Step 4
Compare it with your five answers:

Is the subject right?
Are the place and time what you asked for?
Does the image carry the feeling you wanted?
Are the things that should and should not be there respected?
Is the shape of the frame right?

Something will probably not match what was in your head. That is normal. This exercise is about exactly that distance between what you imagined and what was made.

Step 5
Now change only one thing.

For example, write:

"Remove the figure, keep everything else exactly the same."

The result is the second image.

Step 6
Look at the two images side by side. Check whether only what you asked for has changed, or whether other things moved as well.

Start practicing an important habit right now: at each stage, ask for only one specific change.

This lets you understand what effect each instruction has on the image, so you are not forced to start over every time.

Step 7
Now try one more change, this time about the light or the shape of the image.

For example:

"Make the lamp the only light source, keep everything else exactly the same."

This is version three.

Step 8
Keep the versions that matter to you. Later, comparing these stages will help you see which instruction gave the result you wanted and which one pulled the image away from what you had in mind.

This week's exercise

Choose one picture from your head.

Answer the five questions in five lines.
Make the first image.
Change only one thing with one sentence.
Keep the first and second versions side by side.

For now the goal is not to make a masterpiece. The goal is to see how a thought becomes an image with a few sentences, and how that image can be brought, step by step, closer to what you have in mind.

Next week we turn to the most important part of this work:

How do we explain what we see in our mind to a machine?
The prompt, without formulas.

The three images of this exercise

Version one: the side street with a figure under the lamp
Version two: the same scene, without the figure
Version three: the same scene, with the lamp as the only light source