How I convert PDFs to EPUB semi-automatically
Sometimes e-books come only in PDF format. Almost always PDFs are a pain to read on e-book readers. You can use Calibre to automatically convert it, but the results are okay-ish at best. If the PDF has footnotes, forget it. Unfortunately, the type of books that most often come only as PDFs are science books and these usually have a lot of footnotes.
One option is to use Calibre to convert and then fix the result, but I have found that I get better results in less time when I create a new EPUB, copy the PDF’s content into Emacs, clean it up there and then copy it over to Calibre. This process is what I want to share with you here. You will need Calibre, Emacs or another editor with keyboard macros and some knowledge of XHTML and CSS to follow this guide. It will take long and is boring, but the result is a clean and enjoyable book.
Note
|
I will use the Emacs key notation throughout this guide. |
Create a new book in Calibre
Click on
. Then fill in the metadata and select “EPUB” as format. You can add more metadata and a cover image by right-clicking the book and then selecting Edit metadata. Open Calibre’s editor by right clicking on the book and selecting Edit book.Calibre creates EPUB 2 books by default. Convert the book to EPUB
3[1] by clicking . This
will, among other things, convert toc.ncx
to nav.xhtml
. To
support e-book readers which can’t handle EPUB 3
yet, re-create toc.ncx
as an empty file. It is filled automatically when you
create the table of contents. Open metadata.opf
and replace <spine>
with
<spine toc="id1">
(id1
is the ID of toc.ncx
, defined a few lines above).
You start with a single XHTML file, start.xhtml
. I always use that for the
title page, the copyright notice and so on. You can force a page break to
separate the title and the copyright notice with CSS: Add
style="page-break-after: always;"
to the last element of the virtual “page” or
use a CSS class. To add a CSS file click and enter a
filename ending with .css
. Add the CSS file to the document by right clicking
on start.xhtml
in the file browser and selecting Link stylesheets….
Note
|
The built-in preview does not show page breaks. |
Your files should look similar to this:
start.xhtml
<?xml version='1.0' encoding='utf-8'?>
<html xmlns="http://www.w3.org/1999/xhtml" xmlns:epub="http://www.idpf.org/2007/ops" lang="de">
<head>
<title>Meine zwei Jahre in Russland</title>
<link rel="stylesheet" type="text/css" href="style.css"/>
</head>
<body>
<section class="pagebreak center" id="title" aria-label="Title">
<p>Emma Goldman</p>
<h1>Meine zwei Jahre in Russland</h1>
</section>
<section aria-label="Meta data">
<p>1. Auflage<br/>
München, Januar 2020</p>
<p id="copyright">Anti-Copyright (siehe S. 362)</p>
<p>Die englische Originalausgabe erschien 1921 und 1925 in den
USA aufgrund eines Versehens in zwei Teilen unter den Titeln
<em>My Disillusionment in Russia</em> und <em>My Further Disillusionment in
Russia.</em></p>
</section>
</body>
</html>
style.css
.pagebreak {
page-break-after: always;
}
.center {
text-align: center;
}
I added the IDs “title” and “copyright” to add semantic links to them later.
Some styling advice
Please refrain from using CSS too much. Most people have configured their e-book readers the way they like, with the right font, font-size, margins and so on. If you override their settings, they will be annoyed. I usually only style the title page and headlines.
Do not use <i>
or <b>
tags to emphasize, do not use font-style: italic
or
font-style: bold
. Use <em>
for emphasis and <strong>
for
importance so screen readers will be able to pronounce the text differently.
Make your books as accessible as possible. Using the right tags and not just
<div>
and <span>
for everything is a good start, using epub:type
and/or
role
as well as aria-label
or aria-labelledby
is even better. Read more at
AccessiblePublishing: EPUB Semantics, ARIA Roles, &
Metadata.
Add text to the book
Add a new .xhtml
file in Calibre and write in the heading of the first
chapter. Then switch to Emacs and copy the first paragraph from the PDF into a
text-mode
buffer. The emphasis will not be copied over, so you’ll have to
re-add it. We ignore (but keep) the footnote numbers for now. Repeat with the
rest of the paragraphs of the chapter, leaving 2 blank lines between each
paragraph. The paragraphs will be broken and likely be full of hyphens at the
end of the lines.
(defun my/html-surround-with-tag (beg end)
"Surround region with HTML tag."
(interactive "*r")
(if (region-active-p)
(let ((tag (completing-read "Tag: "
'("blockquote" "em" "strong"))))
(insert (concat "<" tag ">"
(delete-and-extract-region beg end)
"</" tag ">")))
(message "No active region")))
Make sure that auto-fill-mode
is disabled. Position the cursor at the start of
the buffer and press <f3> to start recording a macro. Press <end>
<deletechar> SPC (space bar) and then <f4> to stop
recording. If there is a hyphen at the end of the current line, press
<backspace> 2 times. Press <f4> to call the macro and repeat until
you are at the end of the paragraph. Move the cursor to the first line of the
next paragraph and repeat…
Now you should have a text file with 1 paragraph per line. We need to wrap all
lines in <p>
tags, except block quotes and sub-headlines. Either use another
macro (<p>
<end> </p>
<down> <down> <home>) or this
elisp function:
(defun my/html-paragraphify-buffer ()
"Wrap every line not beginning with < or a newline in <p> tags."
(interactive)
(goto-char (point-min))
(while (re-search-forward "^\\([^<
].+\\)$" nil t)
(replace-match "<p>\\1</p>")))
Once you are done, copy the result into Calibre.
Add footnotes
Use the method from above to copy the footnotes into the now empty Emacs buffer
and clean them up until you have 1 paragraph per line. Footnotes need to be
hyperlink-able, so we can’t just wrap them in plain <p>
tags, they need IDs. I
like to use <li epub:type="endnote" id="fn1">[…]</li>
if the footnote numbers
are increasing throughout the book or <li epub:type="endnote"
id="fn1_1">[…]</li>
if they start with 1 in each chapter. We are going to use a
macro with a counter to generate consecutively numbered IDs. First, set the
counter to 1 with C-x C-k C-c 1
. Then, record this macro:
<li epub:type="endnote" id="fn1_
C-x C-k <tab> ">
<end> </li>
<down> <down> <home>
Tip
|
Use M-x describe-key (mapped to C-h k by default) to find
out what a key combination does.
|
Call the macro until every footnote is wrapped and copy them to the end of the
chapter or the end of the book in Calibre. Wrap them in <section
epub:type="endnotes" role="doc-endnotes"><ol>[…]</ol></section>
Note
|
If the notes appear at the bottom of each page, they are called “footnotes”
(epub:type="footnote" ). If they appear at the end of each chapter or the end
of the book, they are called “endnotes”. (epub:type="endnote" ). Some e-book
readers display footnotes on the same page where the link to them is even if you
put all footnotes at the end of the chapter / book.
|
Add references to footnotes
The footnotes are probably superscript numbers in the PDF but normal numbers in
the EPUB right now. I found that the footnote-numbers are usually preceded by a
space and followed by a space or <
. I use the find & replace function in
Calibre in Regex-mode to convert them to hyperlinks.
Find: ([0-9]{1,3})([ <])
(note the leading space)
Replace: <sup><a epub:type="noteref" role="doc-noteref" href="#fn1_\1">\1</a></sup>\2
Press <f3> to search through the text and C-r to replace.
Finishing touches
Click
, remove the existing entry and click Generate ToC from major headings or Generate ToC from all headings.Click
and set the location of the title page, copyright page, beginning of text and so on.Select
and fix the errors.Use EPUBCheck (command-line) and/or The Ace App (GUI) for more thorough checks.
You’re done! Enjoy your cleanly formatted book. 😊
Updates
-
[2021-03-21] Added
epub:type
s to examples. See Helicon Books: Footnotes in EPUB3, the EPUB specification and AccessiblePublishing: EPUB Semantics, ARIA Roles, & Metadata for more information. -
[2021-03-25] Clarified difference between footnotes and endnotes, replaced
<span><aside>
with<li>
. -
[2021-03-27] Added conversion to EPUB 3, use HTML 5 and ARIA attributes in example, added accessibility-advice.
-
[2021-03-29] Removed
role="doc-footnote"
from<li>
s because it breaks the list semantics for Assistive Technology users.