PDFSwift › EPUB tools › EPUB Repair
🛠 EPUBFree · runs in your browserRepair Common EPUB Errors Online
Apply the repairs that can be done safely and mechanically: archive layout, container, manifest and spine references, required metadata, navigation documents and undeclared HTML entities. Every change is listed and your original is never modified.
Drop an EPUB here
or choose a file from your device
What this tool does
A broken EPUB usually has a small, specific problem: a chapter was renamed but not updated in the package, a converter wrote — in XHTML where XML requires a numeric reference, or a re-zipping tool put the mimetype entry in the wrong place. Those are solvable by rules, without guessing what the author meant.
The repair tool does not rewrite your book’s content. It works on the structure around it and reports each change by name. The list includes: rewriting the archive with mimetype first and uncompressed; rebuilding container.xml when exactly one package file exists; removing manifest entries whose files do not exist, and the spine and navigation references to them; re-linking chapters that the table of contents still points at; adding unlisted files to the manifest; correcting image media types from their bytes; generating a missing navigation document or NCX; and replacing undeclared named entities with numeric references.
Some required values cannot be recovered, only filled in. A missing language becomes und (undetermined), a missing title becomes “Untitled”, and a missing identifier gets a freshly generated one. The report states each of these so you can set real values in the metadata editor afterwards.
How to use this tool
- Choose the EPUB you want to repair. If you have not run it through the validator, do so first to see what is wrong.
- Press Repair EPUB. There are no options; the tool applies every safe repair it has.
- Read the summary: errors before and after, then the list of changes and any errors that remain.
- Press Download repaired EPUB. The file is named
your-book-repaired.epub. - Open the repaired copy in your reading app and compare it with the original before deleting anything.
Supported formats and limitations
- Input/output: EPUB 2 and EPUB 3; the version is preserved.
- Fixed: mimetype position/content, container.xml, missing-file manifest entries, dangling spine and navigation references, unlisted files, wrong image media types, required metadata, EPUB 3 modification date, missing nav/NCX, undeclared entities.
- Not fixed: malformed XML inside chapters, broken CSS, wrong content, encrypted books, damaged ZIP archives that cannot be read at all.
- Guarantee: the original is untouched; the repaired copy is validated again and the result is shown.
- Input: EPUB 2 and EPUB 3 files (
.epub), up to 100 MB. Anything else is rejected before processing. - Archive safety: files with unsafe paths (such as
../), duplicate entries, more than 8,000 entries or an expanded size above 500 MB are refused. - Not supported: DRM-protected books and books with encrypted or font-obfuscated resources (
META-INF/encryption.xml) can be opened for inspection but cannot be converted or merged. - Memory: the whole book is held in your browser’s memory. Very large illustrated books may fail on a phone; a desktop browser has far more headroom.
Privacy and safety
This tool runs entirely in your browser. When you choose a file it is read by JavaScript on this page; it is not uploaded to PDFSwift or to any other server, and the result is created on your device and handed to you as a download. Closing or reloading the tab discards the book from memory.
The page shows advertising supplied by a third party. Advertising scripts run separately from the tool and never receive your book. If you prefer, you can disconnect from the internet after the page has loaded and the tool will still work.
Frequently asked questions
What can this tool repair?
Structural problems that follow clear rules: archive order and mimetype, container, manifest and spine references, navigation, required metadata and undeclared HTML entities.
Can it fix a corrupted ZIP?
No. If the archive itself is truncated or damaged the book cannot be opened, so there is nothing to repair. Try to obtain the file again.
Will it change my book’s text?
No, except that undeclared named entities such as — are written as numeric references (—), which render identically.
It filled in “Untitled” or “und”. Why?
Those fields are mandatory. The repair uses neutral values and tells you; set real ones with the metadata editor.
Is the book uploaded?
No. The repair happens in your browser.
Why do errors remain after repair?
Remaining errors need human judgement or content-level editing, such as XHTML that is not well-formed. They are listed with their codes.
Is the repaired file guaranteed to be valid EPUB?
It is re-checked with this site’s structural validator and reported. The tool does not claim full conformance; for that, run W3C EPUBCheck.