About PediaPDF
PediaPDF is a place to read Wikipedia, in any language, as a properly typeset PDF you can keep or print. Everything happens inside your browser, and you can change the fonts, margins, paper size and layout while you are looking at the result.
What kind of tool this is
Most ways of turning a Wikipedia article into a PDF are converters. You hand over a URL, a machine somewhere produces a file, and it comes back. If the result is not what you wanted, you start again.
PediaPDF is not that. It is a place to read Wikipedia that renders what you are reading as a PDF. You search, open an article, drop the sections you do not want, change the type size or the paper or the margins — and the document re-sets in front of you. Articles you have made stay in a list you can come back to. Nothing is queued and nothing is uploaded, because the typesetter is running in the page you are looking at.
That last part is what a converter cannot offer. By the time the file reaches you, the machine that made it has forgotten you exist.
Wikipedia’s own “Download as PDF” is worth a word, since you have probably used it. It runs on Wikimedia’s servers and it does the job, but it gives you one layout and no say in it.
Where your data goes
There is no account, no sign-in and no advertising, and the article text is never uploaded anywhere: the parsing, the layout and the PDF itself are produced on your own machine. That is worth stating precisely rather than as a slogan, so here is everything your browser actually talks to.
Wikipedia and Wikimedia
Your browser requests the article and its images directly from Wikipedia and Wikimedia Commons, exactly as it would if you were reading the page normally. Those requests are covered by the Wikimedia Foundation privacy policy, not by anything I control.
Maps
Articles about places sometimes carry a map. Drawing it needs map tiles from OpenStreetMap and feature data from the Overpass API, so your browser contacts those services for those articles only.
Fonts
When an article uses a writing system your chosen font does not cover, the app needs a cut-down font that does. It asks this site for one, sending the set of distinct characters it needs — not the text, not the order, not the article. If that request fails, or the service is down, your browser builds the same font itself and the PDF is unaffected.
If you choose a font installed on your own computer, your browser will ask permission to see your installed fonts. That prompt comes from the browser, not from this site; the font is read locally and embedded in your PDF, and no list of your fonts is sent anywhere.
A usage count
When a PDF starts and again when it finishes, the app sends one short message to this site recording the article title, the language, which viewer was used, whether it worked and if not what went wrong, how long it took, how many pages it came to and how quickly they were produced, how large the file came out, whether any characters came out unreadable, and whether you chose only part of the article rather than exporting all of it. For a book it records how many chapters there were but never which ones — a list of articles says far more about a reader than any single article does. The pair is how I catch a generation that never finished at all — on its own, that failure is silent. This is how I know the thing is being used and where it fails.
If you save or share the finished PDF, that is recorded too — which of the two, and for which article. It is counted once per document, and it is the only thing here that tells me whether a PDF was worth keeping rather than merely produced.
There is a further message, which only happens on a phone or tablet. If the built-in preview cannot draw the document and the app falls back to simply handing you the file to save or share, that is recorded too: the title and the language again, and a short code for whether the preview failed to start or started and could not read the document. The error text shown on the screen is never sent — only its type.
No IP address is recorded, and there is no cookie or identifier of any kind, so these are counts rather than a reading history — there is nothing in the log that ties one article to the next, or to you. But the title is in there, and you should know that rather than have me call the site analytics-free.
On your device
Your saved options, your recent articles and any book you are assembling are kept in your browser’s local storage. They never leave the device, and clearing your browser data removes them.
What you may do with the PDFs
Wikipedia’s text is published under the Creative Commons Attribution-ShareAlike 4.0 licence, and that licence travels with anything you make here. In practice this means you may read, print, copy, hand out, and build on these documents — including for teaching and including commercially — as long as you credit the source and pass on the same freedom to anyone you give them to.
By default every PDF ends with a “Sources & license” page crediting the article and stating the licence, so a document that leaves your hands stays properly attributed. You can switch that page off in PDF Options — but the licence still applies to the text, so if you are passing the document on, leaving it in is the easy way to meet it.
Images can differ: many are under their own licences, and a few are used on Wikipedia under narrower terms, so check the file’s page on Wikimedia Commons before reusing an image on its own.
PediaPDF is not affiliated with or endorsed by the Wikimedia Foundation.
Linking to an article
If you keep a wiki, a blog, a reading list or a course page, you can send someone straight to an article. It opens here and begins setting itself immediately:
The last part is the Wikipedia language code — /Taxi/pt for
the Portuguese article, /Taxi/ja for the Japanese one. Leave
it off and the reader gets the article in whatever language their own
browser asks for. Spaces become underscores, exactly as in a Wikipedia
address.
One exception worth knowing: a title containing a slash, such as
AC/DC, cannot be written this way. Open the article here and
copy the address out of your browser’s address bar instead — that longer
form always works, for every title.
Following a link like this is no different from searching here yourself: the article is fetched from Wikipedia by the reader’s own browser, and everything described above about where your data goes applies unchanged.
What does not work well yet
Some of this is honest engineering limits rather than bugs, and it seems better to say so here than to let you discover it in a document you were about to print.
- Right-to-left scripts are drawn as pictures. Arabic, Hebrew, Syriac and Thaana are rendered as images so that the letters join and order correctly. They look right, but you cannot select, search or copy that text. Left-to-right complex scripts, including the Indic writing systems, come out as real text.
- Screen readers get the words but not the structure. These PDFs carry no tagging — the hidden markup that tells assistive software what is a heading, what order to read in, and what a picture shows. The words come through; nothing tells the software how they fit together. If you read with a screen reader, the Wikipedia page itself will serve you better than anything made here.
- Very large diagrams are replaced by a link. A few articles embed enormous transcluded diagrams — railway route maps are the usual culprit — that are taller than any page. Rather than print them off the edge of the paper, the PDF explains what is missing and links to the page where you can see it.
- Very wide tables get restructured. A table with too many columns to fit the paper is narrowed, and if that is not enough it is reorganised so it stays readable. The information survives; the shape may not.
- Long articles take a moment. All the work happens on your device, so a very long article with many images is slower on a phone than on a laptop.
- Links are live, not printed. Links in the PDF are clickable, which is right for reading on a screen — but if you print the document, the addresses behind them are simply gone.
And one thing this is deliberately not: a typesetter for archival print. If you want a Wikipedia article set by TeX, with the full apparatus of a printed book, mediawiki2latex does that properly.
Terms of use
PediaPDF is free to use, and using it means accepting these terms. They are short because the service is simple.
What you may do
Use the app at pediapdf.com, in your browser, to read Wikipedia articles and make PDFs of them — for yourself, for teaching, for work, as often as you like. Link to it freely, as described above. What you may do with the PDFs themselves is also covered above: they carry Wikipedia’s licence, not mine.
What you may not do
- Copy the app. Your browser has to download the code in order to run it, but that does not make it open source. PediaPDF’s own code is proprietary — © 2026 Manuel Pereira, all rights reserved — and you may not copy, republish or modify it, or build another product or service from it.
- Use the site’s services from outside the app — for example, calling the font service from your own software.
- Put automated load on the site at a volume that degrades it for everyone else.
None of this takes away any right the law gives you whatever a website says.
No warranty
The app is provided as it is. I work hard to make the PDFs faithful to the articles, but I cannot promise that every document is complete or correct, or that the site will always be available. If something matters, check it against the article itself. As far as the law allows, I am not liable for any loss that comes from using the app.
Changes
If these terms change, the date at the foot of this page changes with them.
Software by others
PediaPDF is built on other people’s work, used under their own licences: Mozilla’s pdf.js (Apache 2.0); HarfBuzz through harfbuzzjs, bidi-js, opentype.js, fonteditor-core, Project Nayuki’s QR Code generator and Matt DesLauriers’ colour dictionary (MIT); the Noto fonts (SIL Open Font License), DejaVu (Bitstream Vera licence) and Latin Modern Math (GUST Font License); and hyphenation patterns from the TeX hyph-utf8 project, each under its authors’ terms. Nothing here limits what those licences allow.
Who made this
I spent my working life on large IT projects in public education and healthcare — fields where how you handle other people’s data is the job, not a footnote. That is most of why this app is built the way it is, and why the section above tells you exactly what leaves your browser instead of claiming nothing does.
My background is a degree in Philosophy and a postgraduate degree in Information Management, which between them turn out to be a reasonable preparation for worrying about how documents are put together. I built it with heavy use of AI coding tools.
I am retired, so PediaPDF is free, has no advertising, and is not selling anything. It is the first thing I have built purely because I wanted to, and it exists because Wikipedia is one of the best things on the internet and deserves to be read on something better than a screenshot of a web page.
— Manuel Pereira
Contact:
Last updated: September 2026.