Convert PDFs to HTML with pdftohtml

Copy-pasting out of a PDF viewer mangles the layout, but pdftohtml can turn the whole document into real HTML you can actually parse. You will finish with a repeatable way to convert a PDF, limit the job to a page range, and check which files actually got written. This guide uses pdftohtml from Poppler 24.02.0, supplied by Ubuntu's poppler-utils package.

Allow about fifteen minutes. You need a readable PDF and enough free space for the HTML and any extracted images. The examples write into a new working directory, so the source PDF is never touched. No command here needs sudo; reach for elevated privileges only if your chosen input or destination is genuinely inaccessible to your account.

1. Check the installed command

Start with a read-only version check. This catches a common distraction: options and output naming vary between PDF tools, so do not substitute a similarly named converter.

$ command -v pdftohtml
/usr/bin/pdftohtml
$ pdftohtml -v
pdftohtml version 24.02.0
Copyright 2005-2024 The Poppler Developers - http://poppler.freedesktop.org

The package version is useful when comparing output on another machine:

$ dpkg-query -W -f='${Package} ${Version}\n' poppler-utils
poppler-utils 24.02.0-1ubuntu9.9

Your package revision may differ even when the Poppler program version matches.

2. Make an isolated destination

Work in a directory that holds only this conversion. The command normally creates several files, and its positional HTML name is a naming stem, not a guaranteed single output file:

$ mkdir pdf-html-test
$ cd pdf-html-test
$ cp /path/to/input.pdf source.pdf

Replace /path/to/input.pdf with a real path, and keep the original PDF until you have checked the result. Before converting, confirm it is readable:

$ file source.pdf
source.pdf: PDF document, version 1.x, ...
$ test -r source.pdf && echo 'input is readable'
input is readable

3. Convert a page range to ordinary HTML

For a first pass, convert pages 1 through 3 with a clear output stem. The -f and -l options select the first and last page:

$ pdftohtml -f 1 -l 3 source.pdf converted.html
Page-1
Page-2
Page-3

The progress lines are normal. On Poppler 24.02.0, this creates a frames file, one HTML file per converted page, and extracted images where the document needs them. The exact names depend on the supplied stem and the document itself:

$ find . -maxdepth 1 -type f -printf '%f %s bytes\n' | sort
converted-html.html  ... bytes
converted001.png     ... bytes
converted002.png     ... bytes
converted003.png     ... bytes
converteds.html      ... bytes
source.pdf           ... bytes
$ file converted*.html
converted-html.html: HTML document, UTF-8 Unicode text
converteds.html: HTML document, UTF-8 Unicode text

Do not assume a requested name of converted.html means that exact file will exist. Inspect the directory first, then open the generated frames HTML in a browser or hand it to your next processing step.

4. Choose a single HTML document when needed

Use -s when you want one HTML document containing every selected page, instead of the normal frames-style set:

$ pdftohtml -s -f 1 -l 3 source.pdf single.html
Page-1
Page-2
Page-3
$ find . -maxdepth 1 -type f -name 'single*' -printf '%f %s bytes\n' | sort
single.html ... bytes

Output details such as image count, whitespace and generated CSS vary with the PDF. Check both that the file is non-empty and that its first lines look like HTML:

$ test -s single.html && echo 'single HTML exists'
single HTML exists
$ sed -n '1,5p' single.html
<!DOCTYPE html>
<html ...>

If your PDF runs to a large number of pages, a single document can be less convenient to process than the default per-page output. Pick a smaller range with -f and -l while you are testing.

5. Send HTML to standard output

Use -stdout when another command should receive the HTML through a pipe. This avoids choosing generated filenames, but it does not automatically make the output safe to store as a single complete document in every mode:

$ pdftohtml -s -f 1 -l 1 -stdout source.pdf > page.html
$ test -s page.html && echo 'page.html exists'
page.html exists
$ sed -n '1,4p' page.html
<!DOCTYPE html>
<html xmlns="http://www.w3.org/1999/xhtml" ...>

Shell redirection truncates page.html before pdftohtml even runs. For an output you cannot easily recreate, write to a new name and swap it in only after checking it:

$ pdftohtml -s -stdout source.pdf > page.html.new
$ test -s page.html.new && mv page.html.new page.html

If conversion fails, leave the existing page.html alone and inspect the error. Remove the incomplete page.html.new only once you have confirmed it is the failed temporary output.

6. Produce XML for position-aware processing

Use -xml when a later program needs text and coordinates rather than presentation HTML:

$ pdftohtml -xml -f 1 -l 1 source.pdf page.xml
Page-1
$ file page.xml
page.xml: XML 1.0 document, Unicode text, UTF-8 text

Do not treat XML output as HTML. It describes extracted page content and its positions, which usually makes it a better input for a layout-aware parser. Add -noroundcoord only when keeping unrounded coordinates matters; the manual limits that option to XML output.

7. Handle images, passwords and warnings safely

Images are normally written as separate files. -i tells the converter to ignore them, while -dataurls asks for data URLs instead of external images where the build supports it. If you need predictable downstream files, check the result rather than assuming either choice was applied.

The -nodrm option overrides document DRM settings. That is a rights and policy boundary, not a routine troubleshooting switch: use it only when you are authorised to. Likewise, -q hides messages and errors; avoid it until a conversion works, because silence can conceal a bad input or incomplete output.

8. Diagnose the likely failures

A non-zero exit status or an empty output needs investigation. Run the same command without -q, check the input and destination permissions, and verify the selected page range actually exists. A successful exit status is not a visual-quality check, so inspect representative pages and images before you delete the source.

$ printf 'exit status: %s\n' "$?"
exit status: 0
$ find . -maxdepth 1 -type f -size 0 -print

The second command should print nothing. If it prints an output file, rerun into a fresh directory and check the PDF itself. If pages come out blank, try the documented -hidden option for PDFs whose useful text is marked hidden, or compare the visual PDF with the extracted result. Do not assume hidden text is trustworthy, or that extraction preserves the original layout.