calibre Page Breaks Wrong After Conversion: How to Fix It

When calibre page breaks are wrong after conversion, the visible result usually falls into one of three categories: chapters continue on the same page, blank pages appear between sections, or breaks are inserted before headings that are not real chapter openings. These symptoms are normally caused by structure detection, overlapping chapter and page-break rules, unsuitable source headings, inherited CSS, or the way a reading app paginates the finished file. The safest approach is to test one copy of the book, identify which layer introduces the problem, and change only the relevant setting.

1. Confirm the Symptom With a Small Safe Test

Before changing conversion options, create a controlled test. Work with one book and preserve its original format. Do not delete the library, reinstall calibre, or overwrite the only copy of the source file.

1. Select the affected book in calibre.
2. Open Edit metadata and confirm which formats are attached to the record.
3. Note the source format you intend to convert, such as EPUB, DOCX, HTML, RTF, or PDF.
4. Convert to a reflowable test format, preferably EPUB when appropriate.
5. Open the new output in the calibre E-book viewer before sending it anywhere.

Test at least three chapter transitions. Record whether each chapter begins on a new page, whether the preceding page is unexpectedly blank, and whether nonchapter headings also trigger breaks. A single unusual transition may be caused by local markup, while the same problem at every chapter usually points to a conversion rule.

1.1 Compare Paged Mode and Flow Mode

The calibre viewer supports paged and flow layouts. Switch between them with the viewer controls or the Ctrl+M shortcut. On macOS, calibre documents some shortcuts using Ctrl even though other application actions commonly use Command.

In flow mode, pages are not displayed as fixed screens, but structural separation should still be visible. If a blank page appears only in paged mode, pagination, margins, or a forced break near an almost-full page may be responsible. If the chapter is misplaced in both modes, the output structure or CSS is more likely at fault.

Success means you can describe the problem precisely, such as “every h2 heading starts a new page, including subheadings” or “a blank page appears before chapters that are already separate HTML files.” Once you have a repeatable description, stop testing unrelated settings.

2. Check Structure Detection and Chapter Mark Settings

The most important conversion area for this symptom is Structure detection. calibre can detect chapter boundaries and can separately insert page breaks before selected elements. If both mechanisms match the same heading, the output may contain redundant break instructions. If neither mechanism matches the real chapter heading, chapters may run together.

2.1 Review Chapter Detection

Open the conversion dialog and select Structure detection. Examine the chapter detection expression. calibre uses XPath expressions to identify elements in the intermediate XHTML created during conversion. The correct expression depends on how the source represents its headings.

For example, if every real chapter title is an h1 element and no other h1 elements occur in the body, a simple expression such as //h:h1 may be suitable. If chapters are h2 elements, //h:h2 may work. However, broad expressions are unsafe when the same heading level is used for chapter titles, section labels, acknowledgments, and decorative text.

A more selective expression can match text or a source class. Examples include:

• //h:h2[re:test(., 'chapter|prologue|epilogue', 'i')] for matching heading text.
• //*[@class='chapter'] when the input consistently marks chapters with that class.
• //h:h1 | //h:h2[@class='chapter-title'] when two reliable patterns are required.

Class names must come from the input structure used during conversion. Do not assume a generated class such as calibre1 in the converted output existed when structure detection ran. When uncertain, matching tag names and visible heading text is often easier to verify.

2.2 Set Chapter Mark to Page Break

The Chapter mark setting controls what calibre does with a detected chapter. To start detected chapters on new pages, choose the page-break option. A rule inserts a visible divider instead. Both adds a rule and a page break, while none detects chapters without marking them with either treatment.

If chapters are detected correctly but continue directly after the previous section, confirm that Chapter mark is not set to none or rule. Change it to page break, reconvert, and inspect the same transitions.

Success means every element matched as a chapter starts on a new page and ordinary subheadings remain in the normal text flow. Stop adjusting the XPath once it selects exactly the intended chapter titles.

2.3 Check the Separate Page Break Expression

The page-break expression is independent of chapter detection. A broad rule that matches every h1 and h2 can create breaks before subtitles, internal section headings, or headings already separated by file boundaries.

If Chapter mark already inserts breaks at correctly detected chapters, simplify or disable the additional page-break expression for a test. In calibre conversion options, the expression / disables matching for options that accept it. Alternatively, narrow the page-break rule so it targets only elements that truly need a break but are not handled as chapters.

Success means each intended chapter receives one effective structural break rather than several overlapping instructions. If removing the separate page-break rule fixes blank pages without causing chapters to run together, leave it disabled.

3. Inspect Source Headings, CSS, and Output Structure

Conversion settings cannot reliably identify chapters when the source does not distinguish them. A title that merely looks large may be an ordinary paragraph with direct formatting. Conversely, a publisher may use heading tags for decorative text that should not start a new page.

3.1 Verify the Source Headings

For DOCX sources, use real heading styles consistently instead of enlarging normal paragraphs manually. For HTML or EPUB sources, inspect whether chapter titles use meaningful h1 or h2 elements or a stable chapter class. For plain text, configure the text input processing so recognizable chapter lines become headings before structure detection runs.

PDF is especially difficult because it describes page appearance rather than clean document structure. A PDF heading may be reconstructed as several spans or paragraphs, and existing physical page boundaries may produce unwanted gaps. If you have access to a legitimate EPUB, DOCX, or HTML source, it will usually provide a better basis for a reflowable conversion.

Success means the source has a consistent, inspectable pattern for real chapter openings. If the source itself is inconsistent, correct that structure before repeatedly changing XPath expressions.

3.2 Look for CSS That Forces Extra Breaks

Open the output with Edit book and inspect the stylesheet rules applied to chapter headings and their containers. Relevant properties may include page-break-before, page-break-after, break-before, and break-after. Large top margins, fixed heights, empty blocks, and rules that prevent content from breaking can also make a mostly empty screen look like a blank page.

A common pattern is a chapter heading that already begins in a new HTML file while CSS also forces a page break before it. Another is an empty title or separator element with both a fixed height and a break rule. Remove only the redundant declaration, save a test copy, and review it again.

Do not delete all page-break CSS globally unless you have inspected where it is used. Front matter, parts, appendices, and title pages may legitimately need different behavior.

3.3 Split Long HTML Files at Chapter Boundaries

Some source books place the entire manuscript in one large HTML file. calibre can insert CSS page breaks within that file, but separate files at chapter boundaries often provide cleaner navigation and more predictable behavior across reading systems.

In Edit book, open the long text file and use the split controls in the preview area, or choose Split at multiple locations. The multiple-location tool can split at matching headings or another consistent structural marker. calibre updates affected links and references during the split.

After splitting, check the file order in the editor’s File browser. This reading order, sometimes called the spine, should place the chapters in sequence. Save the book, run Check book if available in your workflow, and preview the result.

Success means each chapter begins in its own logical content file, navigation remains intact, and no text is duplicated or omitted. Once the split structure reads correctly in the viewer, do not add another forced break merely because each chapter is now a separate file.

4. Separate Conversion Problems From Viewer and Device Behavior

Metadata downloads, email delivery accounts, Content server options, library naming templates, and most device connection settings do not determine internal chapter breaks. Investigate them only when the correct file is not reaching the intended destination. The first question should be whether the converted file is already wrong before transfer.

4.1 Preview the Exact Output in calibre

Open the newly converted format directly from the book record. If several formats are attached, verify that you are viewing the output you just created. An older EPUB or AZW3 file may remain attached to the same record and cause confusion.

If the book looks correct in the calibre viewer but wrong on a device, send the exact tested file again. Remove the previous copy from the device first when its reading software may retain an older version. Eject or disconnect the device safely after the transfer completes.

Success means the same output file produces predictable chapter transitions in calibre. At that point, stop changing conversion detection and investigate the destination reader.

4.2 Consider Reader Pagination Differences

Reading devices calculate screens according to font size, margins, line spacing, orientation, and available display area. A forced break near the bottom of a screen can leave a short page before the next chapter. That is normal reflow behavior and is different from a completely empty page caused by duplicate breaks or empty content.

Test with the device’s default theme, moderate font size, and standard margins. If only one reading app shows blank pages, compare the same file in another standards-compatible viewer. This helps identify a reader-specific interpretation without damaging a file that works elsewhere.

Fixed-layout books and formats based on physical pages may not behave like ordinary reflowable EPUB files. In those cases, changing chapter-detection XPath may not alter the device’s page model.

4.3 Rule Out Plugins and Viewer Styles

Conversion plugins can modify input or output behavior, while custom viewer styles can change what appears on screen. If the issue began after installing a plugin or adding custom CSS, temporarily repeat the test without that customization.

Use a copy of the book and change one factor at a time. If disabling a plugin fixes the output, confirm the result with a second conversion before removing or reconfiguring anything permanently. If custom viewer CSS is responsible, the underlying e-book may not require editing at all.

5. Check the Operating Environment Only When Relevant

Operating system permissions, antivirus software, cloud synchronization, USB mode, and network settings rarely create a page break inside a successfully generated e-book. They matter when calibre cannot write the output, when the converted file is incomplete, or when the wrong copy is transferred.

On Windows, macOS, and Linux, use a local writable folder for conversion debug output and temporary tests. Avoid placing the test library inside an actively synchronized folder while diagnosing unexplained file conflicts. Pause synchronization briefly if files are being replaced or restored during conversion.

If a device receives an old file, confirm that it is connected in the mode expected by calibre and that the transfer job finishes without errors. For Content server testing, download the converted file to the local computer and compare it with the library copy. Firewall settings may block server access, but they do not normally rewrite chapter structure.

Success means the converted file has a current modification time, a plausible file size, and identical behavior before and after a normal local copy. Stop investigating permissions or networking when the same local file already contains the unwanted break.

6. Use Conversion Debug Output and Job Details

When the cause is not visible in Edit book, calibre’s conversion debug output can show where the break was introduced. Open the conversion dialog, enable debug output, and choose an empty writable folder.

The generated folders represent stages in the conversion pipeline. The input stage shows what the input plugin produced. The parsed stage is useful for examining the XHTML used by structure detection. The structure stage shows the result after structure detection, before later CSS processing. The processed stage shows content close to what is handed to the output plugin.

1. Find the same chapter heading in the parsed and structure folders.
2. Check whether structure detection added a break or changed the surrounding markup.
3. Compare the structure and processed stages for added styles or empty elements.
4. Use the evidence to adjust one XPath or CSS rule.
5. Run the conversion again and compare the same heading.

Also open calibre’s Jobs panel and double-click the completed conversion job to view its detailed log. Warnings about malformed HTML, unavailable resources, or input processing can explain why the result differs from the visible source.

Advanced users can run the graphical interface from a terminal with calibre-debug -g, but most page-break problems can be solved with conversion debug output and Edit book. Command-line debugging is useful when the graphical job log is insufficient, not as a required first step.

Success means you can identify the first pipeline stage where the wrong break appears. Once identified, fix that stage rather than changing device, metadata, server, and conversion settings together.

7. Run a Clean Temporary Test Before Reinstalling

Reinstalling calibre does not repair bad source markup or an overbroad XPath expression. Before reinstalling or deleting a library, run a minimal test that removes saved conversion choices and unrelated customizations from the investigation.

1. Make a backup copy of the source book outside the calibre library.
2. Add the copy as a temporary test record or use a separate temporary library.
3. Convert with simplified settings and no unnecessary heuristic processing.
4. Set a chapter XPath that matches one verified heading pattern.
5. Use Chapter mark set to page break.
6. Disable the separate page-break expression for the first test if chapter marking is sufficient.
7. Preview the output in both paged and flow modes.
8. Add other options back one at a time only if needed.

If the simplified conversion works, the application installation is probably sound. Compare the successful options with the original conversion choices. Saved per-book conversion settings can survive earlier experiments, so verify the current dialog rather than assuming defaults are in use.

If the clean test still fails, inspect the source and debug pipeline. Reinstall only when calibre itself cannot run, open books, or complete ordinary conversions, not when one title has unusual structure.

8. Quick Fix Checklist

• Preserve the original source and test with one copied book.
• Confirm the problem in the newly converted file, not an older attached format.
• Compare calibre viewer paged mode with flow mode.
• Verify that real chapters use consistent source headings or classes.
• Narrow the chapter-detection XPath to actual chapter titles.
• Set Chapter mark to page break when detected chapters should start separately.
• Disable or narrow the separate page-break XPath if it duplicates chapter marking.
• Inspect CSS for redundant break rules, empty blocks, and excessive fixed spacing.
• Use Edit book to split a single long HTML file at chapter headings.
• Check text-file order after splitting.
• Test the exact output in calibre before sending it to a device or server.
• Temporarily remove relevant plugins or custom viewer CSS.
• Review the conversion job log and generate pipeline debug output when necessary.
• Stop changing settings as soon as the same file works at several chapter transitions.

9. Frequently Asked Questions

9.1 Why do all h2 headings start on new pages?

The page-break expression may be matching every h2 element, even when h2 is also used for subsections. Narrow the expression to a chapter class or recognizable chapter text, or let chapter detection and Chapter mark handle the breaks.

9.2 Why does calibre create a blank page before each chapter?

The chapter may already start in a new content file while CSS or conversion rules add another break. Inspect the chapter file and its stylesheet in Edit book. Remove only the duplicate break or empty block, then preview the edited copy.

9.3 Should every chapter be a separate HTML file?

It is not mandatory, but separate chapter files often improve navigation and compatibility. Use Edit book’s split tool when a long source file has reliable chapter headings. Confirm the reading order and links after splitting.

9.4 Why does the book look correct in calibre but wrong on my e-reader?

The device may use different pagination rules, retain an older copy, or interpret particular CSS differently. Verify the exact file, resend it cleanly, and test ordinary font and margin settings. Do not alter working chapter detection until you confirm the file itself is wrong.

9.5 Can metadata, email, or the Content server cause wrong page breaks?

Ordinarily, no. Those features may affect which file is delivered, but they do not usually create internal chapter breaks. Compare the downloaded or delivered file with the tested library output to make sure you are opening the same version.

9.6 When should I stop troubleshooting?

Stop when real chapters begin on new pages, nonchapter headings remain in normal flow, blank pages are gone, and the same output works across several transitions. Do not continue simplifying XPath or deleting CSS after the intended structure has been restored.