Why your ebook is enormous, and what to do
Nobody sets out to make a large ebook. You add a map, a few photographs, an author portrait, a font you like, and the file quietly grows by a factor of thirty. Nothing warns you, because nothing on your own machine is slow.
It surfaces later, in two places: a store's delivery charge coming out of your earnings on every copy sold, and a reader on a phone watching a progress bar and deciding this can wait.
What is actually taking up the space
Three things, in descending order of likelihood.
- Images. Nearly always the answer. A handful of photographs saved at print resolution will outweigh the entire text of a novel many times over.
- Embedded fonts. Each face is a file, and each weight and style is another file. A body face plus its italic, bold and bold italic is four, and a display face for chapter openings is a fifth.
- A fixed-layout structure. Occasionally the cause. Fixed layout stores full-page images for pages that would otherwise be a few kilobytes of text, and it multiplies by the page count. It is the right choice for a picture book and the wrong one for a novel — see Fixed layout: when the page is the design.
An EPUB is a zip archive. If you want to know where your megabytes went, unzip a copy and sort the contents by size. The answer is usually visible in five seconds.
The arithmetic, which is not subtle
Text compresses extraordinarily well. A 90,000-word novel is roughly half a million characters; compressed inside the EPUB, the whole book runs to a few hundred kilobytes. That is the complete text, every chapter, cover to cover.
Now one photograph sized for print. A 6 × 9 page with bleed at 300 DPI is 1,838 × 2,775 pixels — about 5.1 million pixels — and as a JPEG at the quality setting a print export uses, that is commonly several megabytes.
So a single full-page photograph at print resolution can be ten times the size of the novel it sits inside. Eight of them and the text is a rounding error. This is why "it's only a few pictures" is never only a few pictures.
In the Interior Formatter. It checks every image against the size it will actually print or display and flags art that is too low in resolution for the space it fills — which is the same measurement that tells you when an image is far larger than it needs to be. Start a book → · about the Interior Formatter
The fixes, in the order worth doing them
Size images for screens, not for a press. This is the whole game. An image never needs more pixels than the widest device will display, and a screen is not a printing press. For a sense of scale, the cover slot stores ask for is 1600 × 2560 — an interior image that will be shown at most full-width rarely needs more than that on its long edge, and many need far less. Resize the file itself. Do not place a print-resolution image and set it to display small: the full file still ships, and the reader downloads every pixel of it.
Choose the format that suits the content. Photographs and anything with soft gradients belong in a lossy format like JPEG, where the compression is designed for exactly that. Line art, diagrams, charts, logos and anything with flat areas and hard edges belong in a lossless format like PNG, which keeps the edges crisp and stays small because there is little detail to store. Using it the other way around costs you size and quality at once. Newer formats compress better but support across reading systems varies, so check before you rely on one.
Compress on purpose. An export default is chosen to be safe, which means large. Drop the quality setting deliberately and look at the result at the size it will be seen. The point where you can see the difference on a screen is usually far below where the file stops shrinking.
Cut decorative images. A flourish under every chapter number, a photographic texture behind the part titles, a logo repeated forty times. Each one is a file, and repetition across chapters multiplies it. A scene-break ornament drawn as type or vector art costs nothing by comparison and looks better at any size.
Embed fonts only when the book needs them. In a reflowable ebook the reader can override your typeface anyway, and many do. Embed when the face carries meaning the book depends on; otherwise let the device use its own. If you do embed, subset to the characters actually used, and resist adding a fifth file for one page of chapter titles.
The two costs of getting it wrong
A delivery fee. Some stores deduct a charge based on the delivered file size from each sale, so the same fat file costs you again on every copy for the life of the book. The rate, the threshold and whether it applies at all vary by store and by territory, and they change. Check the current terms in the store's own help pages before you assume it does not matter.
A reader who stops. The other cost has no invoice. Someone buys your book on a phone, on mobile data, in a queue. A file that downloads in two seconds gets read. A file that sits there gets abandoned, and the sample they never finished downloading is the sale you never made.
A well-made novel with a dozen illustrations has no reason to be large. Get the images right and the problem does not come back.
Make yours
Everything here is built into the tools, and they are free to use: name your book, bring your manuscript or your artwork, and work as long as you like. You only pay when you want the files without the preview watermark.
Next: What makes an EPUB pass validation · More: Turning a finished print book into an ebook · Putting images inside a book · Ebook cover size, and why it is not your print front