AI can already create images. But who turns them into a children’s book?

The first AI-generated image is rarely the disappointing one. At that point, everything still feels promising: the main character looks charming, the forest feels atmospheric, the colours are appealing, and the story world suddenly seems visible.

The questions usually start to appear later, when those images have to become actual book pages. The main character’s face changes by the fourth image. One scene feels like a soft bedtime story, the next one looks like cinematic fantasy, and the third feels as if it came from a different book altogether. During layout, it may also become clear that the focus of the image is not where the page rhythm needs it to be, the character’s gaze leads the reader in the wrong direction, or the most important gesture clashes with the text area.

That is why the most useful question is not simply whether an image looks good. A better question is whether it belongs in a book. Does it keep the same character? Does it belong to the same world as the other pages? Can it work with the text? Does it support the reading experience, or does it try too hard to stand alone as a visual spectacle?

AI is not the problem. An unguided image sequence is.

It is worth being precise here, because this topic is easy to oversimplify. The point is not that AI is “bad”, or that it cannot produce interesting, atmospheric or even genuinely strong visual foundations. These tools are far more capable today than they were only a few years ago. They can help explore style, test character directions, compare moods and, to some extent, support character continuity.

The difference is not a dramatic battle between machine and artist. The real question is whether there is a professional decision-making process behind the images that can eventually turn them into one coherent children’s book.

AI can be useful for ideation. It can help test visual directions quickly. It can also help an author show what kind of visual world feels close to the story. But once images have to become a book, it is no longer enough to have many pictures. Decisions have to be made: what should stay, what needs to be redrawn, what should be simplified, what needs to be made consistent, and how the whole visual world can function as a publication.

Unguided AI use often runs into trouble because every image tries to win on its own. In a children’s book, however, not every page should behave like a poster. Some scenes need to stay quiet. Sometimes the character’s face matters most. Sometimes the space leads the eye. Sometimes the image needs to hold the story gently rather than carry it away.

A striking image does not automatically work as a page

An AI-generated image on a screen often feels larger, richer and more finished than it will feel as part of a book page. On a monitor, it can carry tiny lights, a detailed background, dense forest textures, glowing objects and atmospheric effects. On a children’s book page, those same details can start competing with one another.

Imagine a scene in which the main character enters a mysterious garden for the first time. AI creates a beautiful image: flowers, vines, lanterns, butterflies, misty light and a tiny house in the background. At first glance, it feels magical. But the main character stands near the edge of the image, the strongest light falls on a decorative gate rather than on the character, and a bright red flower in the background attracts more attention than the actual moment of the story.

The image is beautiful, but it leads the reader’s attention to the wrong place.

A children’s book illustration does not work because it contains many impressive details. It works because it knows what deserves attention. A child is not analysing composition; they are following the scene. If the image does not help them do that, even the most beautiful background can become a distraction.

That is the difference a single successful generated image cannot guarantee. An image may be appealing, but that does not mean the emphasis is in the right place, the visual focus is clear enough, it works beside the text, or it fits the rhythm of the whole book.

A character is not a costume. It is behaviour.

One of the most common challenges in AI-assisted children’s book illustration is character consistency. It rarely starts with a huge mistake. More often, it begins with small shifts. In the first image, the little fox has a narrower face. In the second, the face is rounder. In the third, the ears sit differently. By the fourth, the character looks slightly older. By the fifth, even the movement no longer feels like theirs.

At that point, the viewer may only feel that something is off.

Character consistency is not the same as giving the character a similar hair colour, fur colour or outfit on every page. That is only the surface. A real character has proportions, movements, gestures, posture, facial habits and an internal pace. They shrink differently when afraid. They look differently when curious. They express joy in a way that belongs to them and not to another character.

A little bear, for example, is not the same little bear simply because it is brown and round. It remains the same because it holds itself in the same way, carries the same shy warmth in its gestures, looks up at larger characters in a recognisable manner, and stays identifiable even when it is not smiling.

That is why it is not enough to ask for “the same style” again and again. A book needs character logic. The proportions that hold the figure together have to be defined. The recurring visual traits must be clear. The range of facial expressions and movement has to remain believable. The character must appear in different scenes without losing themselves.

Children notice the shift before they can name it

An adult might say the character is not stable. A child will not put it that way. They may simply stop at the picture, turn back a few pages, and ask: “Is this the same bear?”

That question says more than any technical analysis.

The images in a children’s book are points of orientation. A child uses recurring faces, colours, shapes and places to understand that they are still inside the same world. If those points shift too much from page to page, the child may not consciously analyse the problem, but the sense of safety in the story becomes weaker.

This is especially important in stories with a subtle emotional tone. In a bedtime story, it matters whether one page uses soft, homely light while the next suddenly turns into a dramatic, high-contrast cinematic scene. Both images may be beautiful, but they speak different emotional languages.

In a children’s book, it is not only the character that has to return. The inner rules of the world have to return as well. If the house is small and cosy on one page, but later feels large and castle-like, the child is no longer returning to the same place. If the main character is fragile and hesitant in one scene, but suddenly stands with a confident, almost advertising-like smile in the same emotional situation, the image says something different from the story.

A book has rhythm, not just a list of images

A children’s book does not work by placing twenty impressive images one after another. The images have relationships with each other. One introduces. One pauses. One releases tension. One increases it. One may hold nothing more than a small emotional movement.

That is why page spreads matter. It matters how two facing pages speak to each other. It matters where the visual weight sits, where the page needs breathing room, where the text belongs, and where an image should become quieter for a moment.

With AI, every image can easily speak at the same volume. Every scene becomes striking, every background rich, every light dramatic, every character expressive. Each image may be attractive on its own, but as a book, the result can become tiring.

Not every page in a children’s book should be a major scene. If every page is a climax, there is no real rhythm. The child does not get a moment to rest, the story cannot breathe, and the important moments lose their weight.

Professional illustration work is therefore not only image-making. It is also selection, restraint and emphasis. Sometimes the right decision is to use fewer details. Sometimes it is to keep the background from becoming more interesting than the character. And sometimes a scene should not receive a big visual explosion, because the story is not at that point yet.

A book page is not just an image. It is image and text working together.

In a children’s book, the most important question is not whether an image can technically be placed on a page. During layout, many things can be arranged: where the text will go, how much space the illustration will take, how the image fits the format, and how it appears within the page spread.

The more difficult question is whether the content of the image can truly work with the page.

An AI-generated illustration may be beautiful, but if the main focal point falls in the wrong place, if the character’s gaze leads the reader out of the page, if the most important gesture sits too close to the text area, or if the background carries more visual weight than the actual scene, then the image no longer exists on its own. It begins to affect how the page is read.

This is where a merely striking image may not be enough. The text needs breathing room, the image needs focus, the spread needs rhythm, and the character needs a position that does more than illustrate the scene: it has to guide it. Without that coordination, a page may be technically arranged and still feel uncertain.

In a quiet evening scene, for example, it matters whether the child’s face carries the attention or whether an overly strong background light takes over. On a dialogue page, it matters whether the characters’ gaze supports the reading direction or scatters attention. In a full-page illustration, it matters whether the key moment of the story is held in the centre of the image or lost among decorative details.

This is not a printing trick. It is a book design decision. The content of the image, the position of the text, the rhythm of the spread and the weight of the scene together determine whether the page works.

That is why fast AI images often raise questions before the printing stage, already at the point of page planning. The image may work on its own, but once it has to work with text, the neighbouring page and the key moment of the scene, new decisions appear.

Where does AI belong in a children’s book, then?

The best place for AI is not to take over the direction of the book. It is far more useful as a tool for early orientation.

It can help when the mood is still being explored. It can help compare several visual directions quickly. It can also help when an author finds it difficult to describe the world they imagine, but can react more easily when they see a few visual attempts.

In a professional workflow, AI is not the final destination. It can be raw material. Something like a fast visual notebook: it helps us talk about possibilities, but it does not decide for us. In the final book, human professional judgement is still needed to decide what works, what is merely attractive, what can be kept, what has to be redrawn, and where the entire visual system needs to be reconsidered.

Someone who uses AI well is not using it to avoid design. They are using it to see certain early decisions faster, and then build the book more deliberately.

When is it worth pausing and looking at the material professionally?

The best moment is not when all the images are finished and the layout suddenly reveals that something does not come together. At that point, every correction may be slower, more complicated and more expensive.

It is worth pausing earlier if the character sometimes feels like someone else. If the images are beautiful separately, but feel unrelated together. If it is not clear how the image will communicate with the text. If the cover and the inner illustrations speak in different visual languages. If the story is quiet, but the images are too loud. If the main character sometimes disappears into their own background.

These are not small aesthetic inconveniences. These are the points where visual uncertainty can later become extra time, extra revisions and extra cost.

If you already have AI-generated visual material, it does not automatically have to be thrown away. It may contain useful directions, promising moods, character ideas worth keeping, or details that can serve as a starting point. But it is worth looking at the material professionally to understand what kind of illustration or book design work would be needed to turn it into a real book.

Written by: Ujréti János – Galantusz Grafika, 2026