Book cover design for indie authors: routes and real costs

There is a 2-inch by 1.2-inch rectangle in the lower right corner of the back cover of nearly every print-on-demand paperback, and it is not yours. Amazon's print specs reserve it for the ISBN barcode. Put your best review quote there and it gets printed over.

Most writing about book cover design for indie authors argues about taste. Taste is the smaller half. The bigger half is a budget decision with a break-even you can compute, and there are four routes to a finished cover, each of which fails somewhere specific and nameable.

What a cover has to do, in order

Three jobs, and they are not equally weighted.

• Signal the genre at thumbnail width. A browse row on a phone puts your cover a few hundred pixels tall beside nine competitors. The reader sorts it rather than evaluates it, and if it does not sort onto a shelf they already buy from, nothing else about the cover gets a turn.
• Stay legible after compression. The title has to survive being shrunk, re-encoded as a JPEG, and covered in one corner by a badge. Thin strokes, low contrast between type and background, and titles running past about six words all break here first.
• Clear the file requirements without an argument. This is the only job with published numbers, which makes it the only one you can be certain you got right.

Those numbers, from KDP's ebook cover image requirements page: 2,560 pixels tall by 1,600 wide is the ideal, 1,000 by 625 is the floor, the height-to-width ratio should be at least 1.6:1, and the file is a JPEG or TIFF in RGB, under 50MB, no more than 10,000 pixels in either direction. The same page asks for a narrow 3-to-4-pixel medium-gray border on covers with white or light backgrounds, because without it the cover has no edge against Amazon's white page and appears to float.

Keep that order in mind, because it inverts the order most authors work in. Job three is the one they research first, since it is the one with answers. Five years of doing this by hand taught me that the file requirements are a gate you pass once and then forget. Genre fit and legibility are the part you can get wrong forever without anyone telling you, because nobody writes to say your cover looked like women's fiction and they bought something else.

Route one: hire a designer

Reedsy's cover design cost page, published December 23, 2025, puts the average professional book cover at $880, with most projects on its marketplace falling between $625 and $1,250. The genre medians run lower and tighter than that average: $700 for thriller, crime and mystery, $700 for general nonfiction, $800 for romance, $850 for science fiction, $900 for young adult, $910 for fantasy. The same page splits the budget by method, suggesting $700 to $800 for a stock-photo-and-typography cover and $900 to $1,500 or more for custom illustration.

What does $880 actually buy?

Execution, plus a modest amount of art direction. It does not buy the brief: the decision about what the cover should promise, which subgenre shelf it belongs on, and which of your book's two or three plausible pitches goes on the front.

I priced this work myself for five years, so I know what sits inside numbers like those. In 2019 I charged $450 for an ebook-only cover with two rounds of revisions, and $1,200 for a full jacket with spine, back and flaps. Almost all of the gap between those two figures is production rather than art. The jacket needs a final page count before the spine width can be set, the spine type has to hold at a fraction of an inch, and every element near the trim gets checked against the printer's tolerances. It is not four times the creativity. It is four times the fussing.

Where this route fails is upstream of the designer. They execute the brief you gave them. If your brief says literary and the book is a domestic thriller, you get a beautifully made literary cover and a book that sells to neither audience. A good designer will push back on an obviously wrong brief. Nobody can catch a subtly wrong one, because they have not read your manuscript and are not going to.

Route two: templates and DIY tools

You can spot a template cover in about a second, and it is almost never the image that gives it away.

The tell is the typography. Template covers carry the type the template shipped with: default letter-spacing, a font pairing chosen to look acceptable on every possible title, and a title block sized for a title of a different length than yours. Set a four-word title into a frame built for two words and the type shrinks to fit, which lands your title at exactly the size the layout was designed to avoid.

The second tell is stock photography other people also bought. Cheap and free libraries have hits, and the hits get used, so the same model in the same coat turns up on covers across three genres in a single browse page.

There is a third tell that is harder to describe and easier to see: the cover has no focal hierarchy. Everything on it sits at roughly the same weight, because a template gives every slot equal prominence and filling all the slots feels like using the product properly. Leave slots empty. A cover with a title, a name and one image beats a cover with a title, a subtitle, a tagline, a series badge and a review quote at every size a reader will actually see it.

This is still the right route for some books. Nonfiction with one strong word for a title, a short run of workbooks, anything where the cover works as a label rather than a pitch. Those cases are real, and a template serves them at the cost of a subscription you probably already pay for.

Route three: AI generation

I moved into generation after those hand-design years, and the split between what carried over and what broke was sharper than I expected. Composition carried over. Contrast, hierarchy, the habit of checking everything at thumbnail size, knowing what a romantasy cover is promising: all of it still applies, and it is why two people typing similar prompts get very different covers.

Letterforms broke. Type generated as part of an image goes wrong past about six words, and it goes wrong in a consistent way. The first two words are clean, the middle is plausible, the end collapses into letter-shaped marks. I do not have the model internals. I have several hundred generations and a failure line that sits in the same place every time. The fix is not a better prompt, it is setting the type as real type after the image exists, which is what a designer does anyway.

The other thing that broke is identity across a series. If book two needs the same character with the same scar and the same coat, generation will hand you a cousin. You can get close with a reference image and a long afternoon, and you will still be explaining to yourself why the jaw changed. Object-led series art, a symbol or a weapon or a doorway, holds up across six books. Face-led art does not, and choosing it on book one commits you for the whole run.

The cover generator I build takes a title, an author name, a genre and one line about the book, then returns three variations with the type set as type rather than pasted on top of a photograph. That is the whole scope of it, and being plain about the scope matters more than the pitch.

One thing to handle before you upload. KDP's content guidelines require you to inform Amazon of AI-generated content when you publish a new book or republish an edited one, and they state that AI-generated images include cover and interior images and artwork. Content a tool created counts as generated even if you substantially edited it afterward. A cover you made yourself and refined with an AI tool is AI-assisted, and that needs no disclosure. Answer the question honestly in the form. It is a checkbox, not a penalty.

Route four: learn a design app yourself

Almost nobody should take this route. Affinity Publisher and InDesign will both do everything KDP asks of a print file, and the software is the cheap part. It took me a year to stop making beginner mistakes with type and four more to get fast. You are trying to write books. Buy the skill or rent it.

The number that decides this for you

Do the arithmetic before you pick a route. Convert the price into copies sold, because copies are a figure you can hold against your honest expectation for book one.

Check the table instead of trusting it. KDP's ebook royalty page says the 70% option pays 70% of list price less delivery costs, and puts average delivery cost at $0.06 per unit sold, varying with file size. So you keep 70% of list, minus six cents. KDP's list price requirements page sets the 70% band on Amazon.com at $2.99 to $12.99, a ceiling that moved up from $9.99 on July 7, 2026. Copies are rounded up to a whole copy and worked out from the unrounded royalty, not from the rounded cents in the second column.

The $880 column is Reedsy's stated average. The $450 column is my own 2019 ebook-only price, in for a sense of how the arithmetic behaves at a lower number.

The table also settles the price question people ask separately from the cover question. At $4.99 the same $880 cover repays in 257 copies; at $9.99 it repays in 127. Price a debut at $2.99 to be safe and you have chosen a price that makes every dollar of production cost more than four times as many copies to recover as the top of the band does.

Read the top-left corner of that table. A debut priced at $2.99 needs 433 copies to repay an average cover, before a cent has gone to editing, proofreading or ads. Selling 433 copies of a first book with no mailing list and no series behind it is not a rounding error. For most debuts that is the entire outcome they are hoping for.

Why an $880 first cover is the wrong call

Paying $880 for the cover of your first book is a bad allocation.

The strongest argument against that is one I would have made myself for most of my career. A cheap cover caps the book permanently. The cover is the highest-return dollar in indie publishing, because it is the only asset every potential reader sees before deciding, and cheap covers are exactly why indie titles still read as indie on a browse page. All of that is true. It still does not get you to $880 on book one.

That money buys execution, and execution is not where a first book fails. A first-time author knows less about their own positioning than they ever will again: which subgenre the book actually sits in, which of its plausible promises the market answers, whether the series look should be object-led or figure-led. A wrong brief executed beautifully is a wrong cover with better kerning, and you will have spent the production budget to find that out.

Because the ceiling argument is the serious one, it deserves a real answer. Spend it on book three. By then you know the shelf, you have a series look worth locking down, the cover has a backlist to sell into rather than one title, and the brief comes from evidence instead of hope. A $1,200 jacket that lifts four books is a different purchase from an $880 cover that lifts one.

On book one, generate or template your way to something genre-correct, publish it, watch for ninety days, and treat the cover as a hypothesis you are allowed to replace. Swapping a cover on KDP costs nothing but the re-upload.

What none of this fixes

AthenaCover cannot produce a print-ready paperback cover. It makes an ebook-proportioned image. A paperback cover is a single PDF holding back cover, spine and front as one image, sized to your trim size and page count, at 300 DPI or better. Bleed is 0.125 inches on the top, bottom and outside edges. Text stays at least 0.125 inches inside the trim lines, spine text at least 0.0625 inches clear of the spine edge, and nothing sets below 7 point. KDP also asks you to allow 0.0625 inches of variance either side of the fold lines. And the barcode area stays empty: 2 inches wide by 1.2 inches high at a minimum of 300 DPI, positioned at least 0.76 inches up from the bottom of the cover and at least 0.25 inches clear of the spine hinge. Supply no barcode of your own and Amazon prints an ISBN barcode in a white box of exactly those dimensions in the lower right corner of the back cover. KDP's paperback and hardcover cover pages publish every one of those numbers. It does none of that. If the paperback is the plan, use a designer or KDP's own cover template.

The other limit is text volume. The layouts have room for a title and a name. A title plus a subtitle plus a series line plus a tagline is four blocks of type in a space built for two, and it will crowd.

So decide in this order: print or ebook, then the number of copies the cover has to earn back, then the route. If book one is ebook-only, generate three covers and hold them against the shelf you are trying to join, at thumbnail size, before you spend anything. And if your title is a full sentence, hire a person.