How to Create Custom Illustrated Storybooks for Kids Using Midjourney

For decades, creating a professional children's book required thousands of dollars to hire an illustrator or years of practice to learn to draw. Because of this, millions of amazing stories—stories about your students, your children, or your specific life lessons—were never told.

That era is over. With the release of Midjourney v6 and its revolutionary Character Reference (--cref) feature, you can now generate a character that looks 100% consistent across 20 different pages. No more "morphing faces." No more random outfit changes.

Whether you want to create a personalized gift for a student, a teaching aid for your classroom, or a passive income stream on Amazon KDP, this guide will walk you through the entire workflow, step-by-step.

Phase 1: The Story Script (ChatGPT)

A common mistake is asking ChatGPT to "Write a children's book." It will give you a wall of text that is impossible to illustrate. To create a real book, you need to think in Spreads (two open pages of a book).

We need ChatGPT to act as an Art Director, splitting the text into chunks that match specific images.

Teacher Tip: If you are creating this for a class, ask ChatGPT to include "Sight Words" appropriate for your grade level (e.g., "Include Dolch First Grade sight words").

Phase 2: The Master Character

This is the most important step. You cannot generate the book page-by-page yet. You must first generate The Master Reference. This is the single image that Midjourney will use to "clone" your character onto every other page.

1. Generating the Reference

You want a neutral pose on a simple background. Complex backgrounds in your reference image can confuse the AI later.

2. Getting the URL

Once you have the perfect character:

1. Upscale the image (Click U1, U2, etc.).
2. Right-click the upscaled image.
3. Select "Copy Link" (or "Copy Image Address").
4. Save this URL in a notepad. You will use it 20 times today.

Phase 3: Consistent Generation (--cref)

Now we use the --cref (Character Reference) parameter. This tells Midjourney: "Ignore your random training data and use this specific face".

The Formula

/imagine prompt: [Scene Description] --cref [URL] --cw 100 --ar 3:2

Understanding Character Weight (--cw)

This is where most beginners fail. You must adjust the "Weight" depending on the scene.

Phase 4: Locking the Art Style (--sref)

Consistency isn't just about the character; it is about the vibe. You don't want Page 1 to look like a photograph and Page 2 to look like a watercolor painting.

Midjourney v6 introduced Style References (--sref). These are codes that lock in a specific art style. Add these codes to the end of your prompt (e.g., --sref 1803230093):

• Whimsical Storybook (--sref 1803230093): Soft, painterly, classic fairytale look. Good for fantasy.
• Bold & Vector (--sref 4158632630): Bright, thick lines, flat colors. Good for toddlers/Pre-K.
• Pastel Dream (--sref 2049067160): Calm, soothing, watercolor vibes. Good for bedtime stories.
• Retro Comic (--sref 4158632630): Punchy, vintage comic book style. Good for action/adventure.

Phase 5: Advanced Consistency (Backgrounds)

The hardest part of AI art is keeping the bedroom looking the same from different angles. Here are two pro techniques to solve this:

Technique 1: The "Seed" Method

When you generate the first image of the bedroom, find the "Seed Number" (react with an Envelope ✉️ emoji in Discord to get the bot to DM it to you). Add --seed [Number] to future prompts to keep the random noise consistent.

Technique 2: Pan and Zoom

Instead of generating a new image, click the "Pan Right" ➡️ arrow on your original bedroom image. Change the prompt to include your character. This physically extends the image, guaranteeing the bed and window stay in the exact same place.

Phase 6: Assembly & Publishing

Midjourney creates the art, but it cannot handle text well (it often spells things like "Hppay Brithday"). You need an external tool.

Using Canva for Layout

1. Search Canva for "Storybook" (8x8 inch is a standard size).
2. Upload your upscale Midjourney images.
3. Drag them onto the canvas. Important: Leave space for text! You might need to use Midjourney's "Zoom Out 2x" feature to create empty space around your subject.
4. Add your text using readable fonts like Fredoka One, Chewy, or Comic Neue. Avoid cursive as kids struggle to read it.

Monetization: Amazon KDP

If you want to sell your book, Amazon KDP (Kindle Direct Publishing) allows you to upload your PDF and sell paperback copies with zero upfront cost. Amazon prints the book when someone buys it and pays you a royalty.

Frequently Asked Questions

Who owns the copyright?

As of 2025, in the US, images generated by AI cannot be copyrighted. This means you cannot stop someone from copying your images. However, the story text and the layout/compilation of the book ARE copyrightable by you.

My character has 6 fingers. How do I fix it?

Use the "Vary Region" tool. Click "Vary Region" on the image, use the rectangle tool to select the bad hand, and type "hand" in the prompt box. Midjourney will re-roll just that specific square until it looks right.

Can I do this with free AI?

Tools like Bing Image Creator (DALL-E 3) are free but lack the specific --cref consistency tools. For a professional book where the character looks the same on page 20 as page 1, Midjourney is currently the only viable option.