EPUB File Size Analyzer

This tool answers one question: where is the weight in your EPUB? An EPUB is a ZIP archive of HTML chapters, stylesheets, images, fonts and metadata, so the file you ship is the sum of those parts, and usually one or two kinds of file dominate. Size has real consequences at the stores: on KDP's 70% royalty option Amazon deducts a delivery cost per megabyte (US $0.15 per MB on Amazon.com), and Kobo Writing Life does not accept book files over 100 MB. Seeing the real breakdown tells you whether trimming is worth it and where to start.

It also catches build problems: a cover scan, an embedded font family or a duplicated stylesheet that has quietly inflated every export from a template or Word pipeline.

It is a diagnostic: it tells you which image to re-export or font to drop, so a later compression pass is targeted.

Drop or select an EPUB and everything runs in your browser. The tool reads the ZIP's directory, which records every file's size, and sorts each entry by extension into images (JPG, PNG, GIF, WebP, SVG), HTML/XHTML, stylesheets, fonts (TTF, OTF, WOFF, WOFF2), audio, video or other. Nothing is uploaded, and nothing remains once you close the tab.

You get three headline numbers (the compressed size of the files in the ZIP, their total uncompressed size, and the file count), then a bar for each type with its size, file count and share of the total. Below that, the 50 largest files are listed with their sizes and percentages, and the list can be re-sorted by size, name or type. A plain-text report listing every file by size, with the compression ratio, is available to download.

Two limits: it reads .epub files only, and it is strictly read-only; it does not compress, edit or repackage anything. It reports uncompressed sizes per file, which is the right number for deciding what to shrink, but it means the per-file figures add up to more than the EPUB's size on disk.

The most common surprise is the gap between uncompressed and ZIP size. Text, HTML and CSS compress heavily inside the archive, so a category that looks large in the breakdown may add almost nothing to the file you actually ship. Images, fonts and audio are usually already compressed and barely shrink further, so when the goal is a smaller download, weigh categories by how compressible they are, not only by their share of the bar chart.

Classification is by file extension, so an oddly named or extensionless asset lands in Other, and a misnamed image counted as Other understates your image total. A DRM-protected book is flagged, but its sizes are still listed: the ZIP directory records them, so they can be measured without the encrypted contents being read. The screen shows the 50 largest files; the downloaded report lists every file.

The tool flags nothing as a problem: a large font total may be intentional, and only your CSS can say whether a font is unused.

Start with the headline numbers: confirm the file count matches what you expect and that the ZIP size roughly matches the file on disk. Then scan the bars for anything disproportionate; fonts taking a large share often means a full family was embedded when one weight was needed, and a dominant image total points you toward re-exporting or resizing the heaviest files in the list.

Cross-check the top entries by name: a 4 MB cover or a stray full-resolution photo is obvious once sorted by size. After fixing, re-run the analyzer and keep the report as a before-and-after record.

For a store or client delivery, re-check the finished file's size against the destination's own limits after any change.