Table of Contents
You converted a PDF to EPUB. The file opens, the pages look right — and then you try to bump up the font size and nothing happens. Search finds nothing. On your e-reader the text is a fixed, slightly blurry block that ignores every display setting.
Congratulations: you own a photo album with an .epub extension. Every “page” is a picture of a page.
This is one of the most common PDF-to-EPUB complaints, and it has a specific, fixable cause. Let’s diagnose it properly.
First, Confirm That’s What Happened
Three checks, any reader app will do:
- Change the font size. Real EPUB text reflows; images don’t budge.
- Search for a word you can see on the page. Image-only EPUBs return zero results.
- Check the file size. A text-based novel is typically 1–5 MB. If your 300-page EPUB is 80–200 MB, it’s images.
All three fail → image-based EPUB, keep reading. Text reflows but looks wrong → different problem, see formatting broken.
Why Your Converter Did This
A converter produces an image-only EPUB for exactly one reason: it couldn’t get text out of your PDF, so it fell back to embedding the page images. Some tools warn you about this fallback; many (including Calibre in some configurations) just quietly do it.
And the reason it couldn’t get text is almost always one of these:
Your PDF is image-based. A scan, a photo-to-PDF export, or a flattened file — there’s no text layer inside, only pictures of text. This is the cause in the overwhelming majority of cases. Quick test: try to select a sentence in the PDF itself. Nothing selects? That’s your problem, and the converter never had a chance.
Your PDF’s text layer is broken. Text selects, but copies out as gibberish (font-encoding damage, common in older PDFs of non-Latin scripts). Some converters detect the garbage and fall back to images rather than emit garbled text.
You told it to do this. Some tools have a “preserve original layout” or “fixed layout” option that sounds appealing and means exactly “embed pages as images.” Check your settings before blaming the tool.
Why You Don’t Want an Image-Based EPUB (Even a Pretty One)
It can seem acceptable at first — the pages do look exactly like the book. But you give up everything EPUB exists for:
- No reflow: on a 6-inch e-reader you’re panning and zooming around a fixed page, the precise misery that made you abandon the PDF
- No font/size/margin control, no dark mode that actually works
- No search, no dictionary lookup, no highlights/annotations on the text
- Huge files that sync slowly and eat your device storage
- No working table of contents, beyond whatever page-level bookmarks got copied over
If you truly only need fixed pages on a large tablet, keeping the original PDF is strictly better. The only good reason to convert to EPUB is to get real, reflowable text — which means the images must go through OCR.
The Fix: Put OCR in the Pipeline
Since the root cause is “no readable text layer,” the fix is to create one, properly. Three routes:
Route 1 — Free, local: OCRmyPDF + Calibre. OCRmyPDF adds a Tesseract-generated text layer to your PDF; Calibre then has actual text to convert. Fine for clean, single-column prose in one language. Multi-column layouts, footnotes, and tables will come out scrambled — that’s a structural limitation of the whole toolchain, not something Calibre settings can fix.
Route 2 — Paid desktop: ABBYY FineReader. Stronger recognition than Tesseract, exports EPUB directly. Still weak at structure — expect manual cleanup on anything book-length. Detailed comparison here.
Route 3 — AI-native: PDF2EPUB. Built for exactly this failure case: modern OCR for recognition, a vision-language model for structure (reading order, headings, footnotes, tables and formulas), and a finished EPUB with a working table of contents at the end. Convert a free sample and run the three checks from the top of this article on the result.
For a deeper comparison of all the options, see the full scanned-PDF-to-EPUB guide.
Verify Before You Batch
Whichever route you pick, convert one book and check:
- Font size change reflows the text ✓
- Search finds a rare word ✓
- File size dropped dramatically (images-of-text are 10–50× larger than text) ✓
- Table of contents lists real chapters ✓
- A footnote-heavy page reads cleanly ✓
Five for five? Now batch the rest of your library.
FAQ
Calibre converted my PDF and I got images. Is Calibre broken? No — Calibre simply has no OCR. Given an image-based PDF it can only embed the images. Add an OCR step in front (Route 1) or use a tool with OCR built in.
My EPUB is partly images — some pages are text, some are pictures. Your source PDF is mixed: some pages have a text layer, some don’t (common when a digital PDF had scanned pages inserted, or when a bad OCR pass gave up on hard pages). A full re-OCR of the whole document is usually cleaner than patching pages.
The images in my EPUB are blurry on my e-reader. Can I at least fix that? Converters often downscale images to “reader-friendly” resolutions, which turns 300 DPI scans into mush. Some tools let you raise the image resolution limit. But you’re polishing the wrong thing — the fix is real text, not sharper photos of text.
Will OCR keep my book’s actual images — figures, photos, illustrations? A good pipeline distinguishes pictures-of-text (OCR them) from actual pictures (keep them as images, placed correctly in the flow). That distinction is exactly where cheap pipelines fail — check figures specifically when you verify your sample.
Stuck with a photo-album EPUB? Re-convert a free sample with PDF2EPUB — real OCR, real structure, and you can verify the text reflows before paying a cent.