PDF to EPUB That Actually Reflows: Settings and Limits

You converted a 300-page reference PDF to EPUB, side-loaded it onto your e-reader, tapped the font-size button — and nothing happened. The page just sits there, tiny and unreadable, exactly like the PDF did. That's the single most common complaint about pdf to epub text reflow, and it almost always has the same cause: the converter didn't convert anything. It wrapped each page in a picture and put an .epub extension on the box.

Below is how to tell the difference in ten seconds, which settings actually produce flowing text, and where PDF genuinely cannot be rescued — so you stop burning an hour on a file that was never going to work.

Why your EPUB still looks like PDF pages

A PDF describes positions: "draw the glyph T at x=142, y=688 in 9.5pt Minion." There are no paragraphs, no headings, no reading order — just coordinates. EPUB describes structure: this is a paragraph, this is an h2, let the device decide where lines break.

Converting between them means rebuilding structure that was thrown away at print time. That's real work, so lazy tools skip it in one of two ways:

• Image wrapping. Each PDF page is rendered to a PNG or JPEG and placed in an XHTML file. The EPUB opens, but font size, margins and hyphenation controls do nothing. File size is usually a giveaway: a text-only 14 MB PDF that becomes a 30 MB EPUB is images.
• Fixed-layout EPUB. A legitimate format for comic books and children's picture books, where absolute positioning is intentional. For a reference manual it's the wrong choice — it pins text to a page-sized canvas, so it cannot reflow by design.

The ten-second test: open the EPUB on your reader, increase the font two steps and rotate the device. If line breaks move, you have real text. If the whole page scales like a photo, you have a screenshot album.

Check the PDF first: does it even have text?

Before blaming the converter, confirm the source has a text layer. Open the PDF in any viewer, select a sentence with the mouse and copy it. Paste it into a text editor.

• Clean text appears — great, a reflowable conversion is possible.
• Nothing selects — it's a scan. You need OCR first; no converter can reflow pixels.
• Garbled characters appear (\ue003, random accents, missing spaces) — the PDF uses broken font encoding. Conversion will produce the same pdf to epub scrambled text, and OCR is usually the faster fix than fighting the encoding.

This one check explains the majority of "pdf to epub reflow broken" reports. If copy-paste fails in Acrobat, it will fail in Calibre too.

The three things that break reflow

Assuming you have a real text layer, three artefacts of print layout cause almost all of the ugliness.

Running heads and page numbers

Every page has a header like Chapter 4 · Routing Protocols and a footer like Page 118 of 412. In a PDF those sit in the margin. In a reflowed EPUB they land mid-paragraph, 400 times, chopping sentences in half. Good converters detect repeated text at the same coordinates and drop it; the rest need you to say so explicitly with a search-and-replace rule.

Hyphenation at line ends

Justified print text breaks words across lines: imple- / mentation. Once lines reflow at a different width, you get "imple- mentation" scattered through the book. The fix is dehyphenation: rejoin the fragments, but only when the result is a real word — "co-" / "operative" should stay hyphenated in some styles, and "e-" / "mail" definitely should.

Two- and three-column pages

This is the hardest one. A naive extractor walks the page top to bottom in the order glyphs were drawn, so on a two-column academic paper it reads the first line of column 1, then the first line of column 2, then back again. The output is grammatically shredded — every sentence interleaved with a different sentence. Column-aware extraction has to cluster text blocks by x-position first, then read each column fully before moving right.

Two-column text with clear gutters usually survives. Three columns, sidebars, pull quotes and figure captions floating between columns usually don't.

Hands-on: Calibre with --enable-heuristics

Calibre ships a command-line tool, ebook-convert, that does the structural guessing. The key flag is --enable-heuristics, which turns on a bundle of repairs: unwrapping hard-wrapped lines, dehyphenating, deleting blank paragraphs, detecting chapter headings and normalising scene breaks.

A realistic starting command for a 412-page, 14 MB technical handbook:

ebook-convert "Networking-Handbook.pdf" "Networking-Handbook.epub" \
  --enable-heuristics \
  --html-unwrap-factor 0.4 \
  --sr1-search "(?m)^\s*Page \d+ of 412\s*$" --sr1-replace "" \
  --sr2-search "(?m)^\s*Chapter \d+ · .*$" --sr2-replace "" \
  --remove-paragraph-spacing \
  --smarten-punctuation \
  --output-profile tablet

What each part is doing:

• --html-unwrap-factor controls how aggressively short lines are joined into paragraphs. Lower values (0.3–0.4) merge more; raise it toward 0.6 if paragraphs start swallowing headings and list items.
• --sr1-search / --sr2-search are regular-expression passes that delete the running footer and header. Open the PDF, copy the exact strings, and build the patterns from them — generic patterns miss.
• --remove-paragraph-spacing restores indent-style paragraphs instead of blank-line gaps inherited from the PDF.
• --output-profile sets sensible base font metrics for the target device class.

If one heuristic misbehaves you can keep the rest: --enable-heuristics --disable-dehyphenate leaves hyphens alone, which is what you want for a chemistry text full of compound names. Convert, read five pages from the middle of the book on the actual device, adjust one setting, repeat. Two passes usually settles it; a 14 MB PDF typically lands around 2–4 MB as EPUB once page images are gone.

The online route when you don't want a toolchain

Installing Calibre for one travel guide is overkill, and the command line is not everyone's idea of a good evening. For a straightforward single-column PDF with a real text layer, use our online PDF to EPUB converter — it extracts the text layer and rebuilds paragraph structure rather than pasting page images, so the output resizes on a Kobo, Kindle or phone.

Apply the same ten-second test to whatever comes back: bump the font two steps, rotate, and check that line breaks move. If you're juggling several formats — MOBI, AZW3, DOCX — the rest of the ebook converter hub covers those conversions too.

Use the desktop route when you need regex header stripping, per-heuristic control, or a custom table of contents. Use the online route when the PDF is simple and you want it on the device in a minute.

Where PDF to EPUB genuinely hits a wall

Be honest about the limits so you don't chase an impossible result:

• Scanned books without OCR. Pixels in, pixels out. Run OCR first, then convert, and expect to proofread.
• Wide tables. A 9-column spreadsheet cannot reflow onto a 6-inch screen in any readable way. Keep those pages as images or consult them on a tablet.
• Mathematical notation. Equations set as vector drawings or odd font encodings come out as nonsense. Keep maths-heavy PDFs as PDFs.
• Heavily designed layouts. Magazines, brochures and textbooks with sidebars wrapping around figures lose their spatial meaning the moment text flows.
• Footnotes. They typically land as loose paragraphs at the end of a chunk rather than as linked notes, unless the PDF was generated with tagging.

A repeatable five-minute workflow

Put together, this is the routine that avoids wasted conversions:

1. Copy a sentence out of the PDF. No text, no reflow — send it to OCR instead.
2. Check the layout: single column converts well, two columns usually survive, three columns or heavy design probably won't.
3. Note the exact running header and footer strings so you can strip them.
4. Convert — online for simple files, ebook-convert --enable-heuristics when you need control.
5. Read five pages from the middle of the result on the real device, at your real font size.
6. Adjust one setting (--html-unwrap-factor, or disable a single heuristic) and run again.

The goal isn't a pixel-perfect copy of the print book. It's text that flows at the size your eyes want, with chapters you can jump to — which is the entire reason you left PDF behind.

FAQ

Why won't the font size change in my converted EPUB? Because the EPUB contains page images or is fixed-layout. Reconvert from the original PDF with a tool that extracts the text layer; no reader setting can fix an image-based file.

Can I convert a scanned PDF to a reflowable EPUB? Only after OCR. OCR creates the text layer that conversion needs. Expect typos in anything italic, handwritten or low-resolution.

Why is my converted text scrambled or interleaved? Almost always a multi-column page read line-by-line across both columns, or a PDF with broken font encoding. Test by copy-pasting from the PDF: if the paste is already scrambled, the source is the problem.

Does `--enable-heuristics` ever make things worse? Yes. On poetry, code listings and verse, line unwrapping destroys intentional breaks. Disable individual heuristics (for example --disable-unwrap-lines) rather than turning the whole bundle off.

Do images and tables survive? Images are carried over and scaled to the screen. Simple tables survive; wide ones become unreadable and are better kept as images.

EPUB or just send the PDF to my e-reader? If the PDF is single-column and set at roughly paperback size, reading it directly is fine. For anything A4-sized or academic, a reflowed EPUB is far more comfortable.