How to Create a Children's Book with AI Images in One Weekend

How to Create a Children's Book with AI Images in One Weekend

My niece's fifth birthday was three weeks away when I had the idea: make her a personalized children's book featuring a character that looked like her, going on an adventure based on her specific interests.

Custom illustrated children's books from professional illustrators cost $800–$2,000 and take months. I had three weeks and approximately zero illustration skills.

I made the book in one weekend. It took two days of focused work, cost $10 in Midjourney credits, and produced something my niece has read approximately forty times since her birthday.

This is the complete guide to how I did it, and how you can too.

Weekend Overview: The Plan

Friday evening: Story outline and character design Saturday: Generate all illustrations Sunday: Layout, formatting, and print preparation

Total active work time: approximately 10–12 hours.

Step 1: Write the Story First (Friday Evening)

The most important step happens before you open any AI image tool: write the complete story.

This sounds obvious, but many people start generating images and then try to write story pages around them. This creates two problems: inconsistent image style and a story structure that serves the images rather than the other way around.

Write the story completely first. For a standard 32-page picture book:

• 14–16 spreads (double-page layouts)
• 1–2 sentences of text per spread
• Clear story arc: introduction → challenge → adventure → resolution
• One main character, 1–2 supporting characters

My story: A girl named Maya who can talk to plants. She discovers that the old oak tree in her garden is sad and goes on a journey to find what the tree misses. Total: 14 spreads, around 300 words of text.

Character Design Notes

Write detailed character descriptions before generating anything:

• Age, approximate appearance
• Key visual signature elements (what makes them recognizable — a distinctive hat, a specific color scheme, an unusual hair style)
• The simpler the character design, the more consistent your AI generations will be

Step 2: Create Your Character Reference (Saturday Morning)

Before generating any story illustrations, create a detailed character reference sheet. This is the single image that will anchor all your subsequent generations.

My Midjourney prompt for Maya: children's book illustration, young girl approximately 5 years old, curly auburn hair with two small braids, green eyes, wearing a yellow sundress with white collar, round face with freckles, friendly expression, simple clean illustration style, white background, character reference sheet

Generate 4–6 options. Select the one that best matches your vision. This becomes your "character bible" — the reference image for all future generations.

Using style reference: Save this image and use it with --sref [image URL] in all subsequent Midjourney prompts. This anchors the visual style and character appearance across generations.

Step 3: Generate Scene Illustrations (Saturday)

Work through your story page by page. For each spread:

1. Describe the scene in detail
2. Include your character reference (--sref)
3. Maintain consistent environment and art style descriptors across all prompts
4. Generate 3–4 options, select the best

Prompt template I used: children's book illustration, [scene description], [character description with key visual elements], warm watercolor style, soft colors, [time of day/lighting], --sref [character reference URL] --ar 4:3 --style raw

Example (spread 5): children's book illustration, young girl with curly auburn hair and yellow sundress sitting beside a large old oak tree with a sad expression, whispering to the tree trunk, golden afternoon light through leaves, warm watercolor style, soft colors, --sref [URL] --ar 4:3 --style raw

Dealing with Character Consistency Issues

AI consistency across a 14-spread book is imperfect. My strategies for maintaining consistency:

• Keep the character description in every prompt (don't abbreviate)
• Use the same --sref reference throughout
• If a character looks wrong in a generated image, regenerate until the character matches rather than accepting a near-miss
• Use Photoshop or Canva to make minor adjustments (adjust hair color, change dress color to match) on otherwise good compositions

By spread 14, I had accepted some variation in minor details (freckle density, exact hair arrangement) while maintaining the essential character: right age, curly auburn hair, yellow dress, round friendly face.

Cover Image

Generate the cover last — once you've developed a consistent character appearance. The cover should feature your best-looking character image with space for title text.

For text on the cover, I used Ideogram AI to generate the title treatment, since it handles text in images more accurately than Midjourney.

Step 4: Layout and Formatting (Sunday)

Tools for Layout

• Canva — easiest, good templates for book layout
• Adobe InDesign — professional, steeper learning curve
• Google Slides — surprisingly adequate for simple layouts
• Affinity Publisher — professional, one-time purchase

I used Canva. The book template library includes picture book formats.

Layout Process

1. Create a new Canva project with your book's target size (typically 8.5×8.5" for square picture books or 8.5×11" for landscape)
2. Create a spread template with image area and text area
3. Import each illustration
4. Add text for each spread in a clear, readable font (avoid overly decorative fonts for the main text)
5. Create front cover with title and author name
6. Create back cover with brief description and barcode placeholder

Typography Considerations

• Font size: minimum 18pt for body text (children's books are read aloud with the child looking at pictures)
• Avoid serif fonts for picture books — use clear sans-serif
• High contrast text on backgrounds — dark text on light, or light text on dark

Step 5: Print Preparation

For Print-on-Demand (Amazon KDP, Ingram Spark)

File specifications:

• PDF format
• 300 DPI resolution for all images
• CMYK color profile (KDP also accepts sRGB)
• Bleed: 0.125" on all sides that extend to the edge
• Interior margins: minimum 0.5" inner margin (for spine), 0.25" outer

Midjourney upscaled images are typically sufficient for 300 DPI at picture book size — the SDXL high-resolution upscale produces images suitable for up to 8×10" print without quality loss.

For Personal Printing (What I Did)

Export as PDF and order through a local print shop or an online service like Shutterfly, Snapfish, or Artifact Uprising for high-quality photo books. This is faster and cheaper for single copies.

My niece's book was printed at a local Staples for $18 with spiral binding. It's not a professional children's book — it's better in some ways, because it's the only one exactly like it.

Further Reading

• How to Turn Photos Into Paintings With AI (Van Gogh, Picasso, More)
• 5 AI Image Generators Specialized in Anime Style (2026)
• 6 AI Image Generators That Are Safe to Use Commercially (2026)
• Best Free AI Image Upscalers for 4x Lossless Enhancement (2026)
• Free AI Tools for Isometric Room Illustrations (Cozy Spaces 2026)
• Using ChatGPT to Build a Dropshipping Business from Scratch
• How QuillBot Saved My Academic Career (A Graduate Student's Story)
• Claude AI Free Tier: Everything You Can Do Without Paying