Convert PDF to Kindle Format (KFX) — Academic Papers, Footnotes, and Multi-Column Layouts

KFX is Amazon’s proprietary Kindle Format 10 — the native ebook format shipped on every Kindle device sold since 2018. Where EPUB and MOBI treat text as a simple flow of paragraphs, KFX adds a layout layer: it supports custom font embedding, advanced hyphenation, enhanced paragraph spacing, and page-turn animation that respects the document’s intended typographic weight.

The difference is visible in reading comfort. A KFX file rendered on a Kindle Paperwhite uses the device’s built-in Bookerly or Amazon Ember font with proper optical sizing. An EPUB or MOBI from the same source document will use the same fonts but without the Kindle typography stack — line spacing is less refined, hyphenation is cruder, and margin control is reduced.

For academic and technical readers who spend hours on dense text — legal briefs, research papers, technical manuals — the KFX difference accumulates into noticeably less eye strain. KFX is also the format that Kindle’s enhanced table of contents feature relies on: chapter navigation via swipe gestures and the reading progress bar at the bottom of the screen are more accurate in KFX than in legacy formats. If you are going to convert a PDF to read on Kindle, KFX is the format worth targeting.

Every converter that produces KFX from a PDF works in two steps: PDF → EPUB, then EPUB → KFX. The first step is where the quality is won or lost. Calibre, the most widely used open-source converter, handles the EPUB → KFX step well — but its PDF extraction is the weak link.

Calibre’s PDF extractor reads text in document order, which roughly corresponds to left-to-right, top-to-bottom page scanning. On a single-column PDF this is adequate. On a two-column academic paper, Calibre reads the first line of the left column, then the first line of the right column, then the second line of the left column — interleaving both columns into an unreadable stream. Heading detection is equally unreliable: Calibre applies heuristics based on font name rather than rendered font size, so a section title set in 14pt Bold Helvetica is treated identically to 14pt Bold body text.

Footnotes compound the problem. A PDF footnote is a positional annotation at the bottom of a physical page. When Calibre converts to EPUB and strips the page model, footnotes either get appended to the end of the document with no navigation link, or they disappear entirely. The EPUB → KFX step then faithfully converts that broken EPUB — and the broken structure ends up in your Kindle library.

PDF is the primary input format. The pipeline reads text objects and their x/y coordinates directly from the PDF’s internal structure using PyMuPDF, rather than rendering pages to pixels. Text-based PDFs — academic papers, technical manuals, legal documents, books exported from Word or InDesign — convert with full structure detection. Scanned PDFs (image-only pages) are also accepted; the pipeline triggers OCR via Gemini when no text layer is found.

DRM-free EPUB files are accepted as input for conversion to KFX. Files from the Kindle or Apple Books store that carry DRM cannot be processed — the converter requires files you own outright or files without digital rights restrictions.

Formats leafbind does not accept directly: MOBI (Amazon removed MOBI from Send-to-Kindle in 2022; leafbind follows the same policy), CBZ comic archives, ODT, or Apple Pages files. For those formats, export to PDF first, then convert.

Multi-column PDFs are the hardest class of document to convert correctly to Kindle format, and the most common source of garbled output. Calibre’s own documentation is direct about the limitation: “Complex, multi-column, and image-based documents are not supported.” This refers specifically to Calibre’s PDF input processing — the EPUB → KFX step works fine; PDF extraction is where multi-column layouts break down.

Converting a two-column academic PDF to Kindle format requires a converter that understands the physical layout of the page, not just the document’s text flow. Standard PDF-to-Kindle tools read text in document order, which roughly follows left-to-right, top-to-bottom page scanning. On a single-column PDF this produces correct output. On a two-column document such as an IEEE Transactions paper or a Nature article, document order interleaves the two columns: line one of column A, line one of column B, line two of column A, and so on. The result on your Kindle reads as alternating fragments from each column, making the text unreadable. The correct approach is coordinate-aware extraction: determine the x and y position of each text block, group blocks into columns by their horizontal bounding boxes, then read each complete column top-to-bottom before moving to the next. This is how leafbind’s pipeline works.

In practice, this handles IEEE, arXiv, ACM, and Nature two-column layouts correctly. Three-column layouts (common in newspaper-style PDFs) are also supported. Layouts where columns are unevenly sized or where sidebars break the column grid may require a manual review of the output.

For a detailed walkthrough of the column-detection pipeline, see Multi-column PDF to Kindle →

Academic and legal documents rely on footnotes for citations, cross-references, and annotations. In the source PDF, a footnote is a positional annotation at the bottom of the physical page — the PDF format has no explicit structural link between the in-text superscript marker and the footnote body. That relationship is implied by position, not encoded.

Most converters that produce EPUB from a PDF either append footnotes at the end of the document (losing the marker relationship) or drop them entirely. When Calibre then converts that EPUB to KFX, the broken structure carries through — you end up with a wall of unnumbered annotations at the back of the file with no way to navigate to them while reading.

leafbind’s pipeline detects superscript markers in body text — numeric superscripts, symbolic markers (*, †, ‡), and parenthetical citation numbers — and locates the corresponding footnote text by searching the bottom of the page region in which the marker appears. The pipeline then generates linked anchor pairs: one at the in-text marker, one at the footnote body. In the KFX output, those links become Kindle’s native footnote popup overlays — tap the superscript, the footnote slides up without losing your reading position.

For a full explanation with before/after rendering examples, see PDF footnotes to Kindle →

Calibre is excellent — it is the industry-standard open-source ebook manager and converter, and for most ebook formats it works well. If you have a simple single-column PDF and time to install a desktop application plus the KFX Output plugin, Calibre is a completely viable path to KFX at no cost.

Where Calibre falls short is specifically on complex PDFs:

• Multi-column extraction — Calibre reads in document order and interleaves columns. Its own manual acknowledges that multi-column documents are not supported.
• Heading detection— Calibre’s heading heuristics use font name rather than rendered font size. A title in Bold Helvetica is treated the same as bold body text. Table of contents in Calibre-converted academic papers is frequently flat or missing.
• Footnote backreferences— The EPUB Calibre produces from a PDF typically appends footnotes at the end of the document with no navigation link from the in-text marker. The EPUB → KFX step then converts this faithfully, preserving the broken structure in your Kindle library.

The honest tradeoff: Calibre is free and handles the full ebook ecosystem. leafbind is a paid web service that solves a specific problem well. For a novel exported to PDF, a Word document, or any single-column text PDF, Calibre’s output is adequate and there’s no reason to pay. For academic papers, technical manuals, and legal documents with multi-column layouts, footnotes, and structured headings, leafbind produces noticeably better KFX output.

One more thing worth knowing: leafbind’s pipeline uses Calibre’s own KFX Output plugin for the final EPUB → KFX step. Calibre’s KFX conversion step is good — the weakness is in PDF extraction, not in the KFX assembly itself.

Send-to-Kindle is convenient and free. For simple documents — a plain-text ebook, a DRM-free EPUB, a clean single-column PDF — it works well and is the easiest path to getting a file onto your Kindle.

The problems arise when layout fidelity matters:

• Amazon re-converts your PDF on delivery. When you send a PDF via Send-to-Kindle, Amazon converts it on their servers. The conversion strips multi-column structure, drops footnote links, and flattens heading hierarchy. What arrives on your Kindle looks like a reflowed plain-text document, not the original PDF layout.
• File size limits.The Send-to-Kindle web uploader accepts files up to 200 MB. For larger files, USB transfer is the alternative — it bypasses conversion entirely and loads the PDF into Kindle’s native PDF reader as-is.
• Format restrictions. Send-to-Kindle no longer accepts MOBI (removed in 2022). Accepted formats include PDF, DOC, DOCX, TXT, RTF, HTML, EPUB, and common image formats. For other formats, convert to PDF first.

For PDFs where formatting matters, the better alternatives are USB transfer (loads the PDF as-is into Kindle’s PDF viewer — no conversion, all layout preserved) or leafbind (converts PDF to KFX with correct column order, tappable footnotes, and navigable headings).