Build log #1 · October 23, 2026

A coloring book in three days

On Wednesday I decided to try selling coloring books on Amazon. By Friday afternoon I had a finished 86-page book and a print-ready cover. Here's how it went, what it cost, and the one mistake that almost sent me down the wrong path.

Picking the niche: my first idea was too late

When I scrolled Amazon's best-seller list for adult coloring books, Halloween and fall books were all over the top of the list. That looked like the obvious move until I checked the calendar. It was October 22. A Halloween book would go live after Halloween and then sit for ten months. Seasonal books need to be on Amazon six to eight weeks before the holiday.

Winter was the window still open. I wanted a New York angle, since that's where I live, so I compared how New York books and cozy winter books actually sell, using each book's Best Sellers Rank (lower means more sales):

Pulled from Amazon.com on October 22. Ranks move every hour, and winter titles are near their seasonal peak in October.

The answer was clear: winter is what sells, and New York is the flavor. So the title leads with winter, City Snow Days, and the pages are cozy New York scenes: a cat on a radiator, a brownstone stoop, the corner shop, a pretzel cart, skaters, pigeons on a snowy ledge.

Making 40 pages with Midjourney

I used Midjourney's $10-a-month Basic plan. My first prompt asked for a "cozy coloring book page for adults" and came back with thin, detailed, almost realistic drawings, with knit textures and every brick drawn. Beautiful, and completely wrong for this market. The books that sell right now are "bold and easy": thick outlines, chunky cartoon shapes, big open spaces.

This is the recipe that worked. Only the scene in the middle changes from page to page:

simple kawaii coloring page, bold and easy, [SCENE], very thick bold black marker outlines, chunky rounded cartoon shapes, minimal detail, big empty areas to color, flat black and white line art on pure white --ar 1:1 --no fine lines, hatching, crosshatching, small details, bricks, texture, shading, gray, color, text, words

Two changes made the difference: dropping Midjourney's "raw" style setting, and listing what I didn't want (fine lines, hatching, texture). Each prompt returns four options, so 40 scenes gave me 160 drafts to choose from. The whole book used about 50 of the plan's roughly 200 monthly generations.

Getting the pages ready for print

Midjourney images are 1,024 pixels square with soft gray edges. A printed 8.5-inch page needs about 2,550 pixels of pure black and white. Instead of paying for upscales, I ran every page through a cleanup step: convert it to pure black and white, trace the lines into smooth vector curves, then render them at full print size. That also removed stray dots and the gray haze. One taxi had "TAXI" written on its roof sign, so I erased the word; the book has no text inside the pictures.

Each design sits on its own right-hand page with a blank back, so markers can't ruin the next picture. Bleed-through is a common complaint in coloring book reviews. Up front there's a title page, a "this book belongs to" page and a color-test page.

The images are AI-generated, and I'm disclosing that to Amazon when I publish, as KDP requires.

The money

The book is 8.5 by 8.5 inches with 86 pages in black ink. Amazon prints each copy when it's ordered, so there's no inventory.

The price matters more than I expected. At $8.99 the royalty rate drops to 50% and I'd earn $1.66 a copy, about half as much for a dollar off. You can check any book with the KDP royalty calculator I built for this.

Spent so far: $10 for one month of Midjourney. I break even at four copies.

What's next

1. Upload to KDP and order a printed proof to check line weight and margins on real paper.
2. Publish by November 15 so it's live for the holidays.
3. Run a small, capped Amazon Ads test.
4. Report actual sales, ad spend and royalties in the next log.