Home › PDF tools › Watermark PDF

Watermark PDF

Stamp a PDF with text or an image — font size, colour, opacity, rotation and layout are all yours to set — on every page or only on the pages you name. The mark is drawn into the page by pdf-lib inside your own browser and the new file is saved straight back to your device. This tool adds watermarks only: it cannot take one out of a file that already has it.

Choose a PDF file

FileNo file loaded

Pages–

File size–

No PDF loaded yet.

Watermark type

TextReal text drawn with the standard Helvetica font, so the mark itself stays selectable and searchable in the output. Latin-1 characters only. ImageA PNG or JPG you choose — a logo, a stamp or a scan of a signature — placed on the page at the size and opacity you set.

Watermark text

Latin letters, digits and common punctuation. Non-Latin scripts are rejected — see the FAQ.

Font size: 48 pt

The size is in points and is the same on every page, whatever the page size.

Colour

A mid red or grey at low opacity reads as a watermark without hiding the page text.

Layout

Rotation

Pages to stamp

Type all, or a list like 1-3,7 or 2,5,9.

Opacity: 30%

A watermark is a visual mark, not a lock. It does not stop anyone copying, editing or re-using the page, and a PDF editor can remove it. This tool also adds watermarks only — it cannot remove one that is already in your file.

Choose a PDF file to begin. Nothing is uploaded — the file is opened in this browser tab.

After a run the tool re-opens the file it has just written and compares the page count, the page boxes and the page rotation with the input, so the “unchanged” line you see is a measurement of the output, not a promise. Everything happens in your browser: no upload, no server, no account.

What is a PDF watermarker?

A PDF watermarker draws a visible mark onto the pages of a document you already have — a word such as DRAFT or CONFIDENTIAL, a file reference, a client name, a company logo or a scanned stamp. The mark is part of the page from then on: it prints with the page, travels with it when the file is forwarded, and shows up in every viewer that opens the PDF.

This page does the whole job in your browser. You pick a PDF, pick text or an image, set the size, colour, opacity, angle and layout, choose which pages to stamp, and press one button. The watermark is drawn into the page with the pdf-lib library in the tab you are looking at, the new document is assembled in memory, and it is handed back to you as a download. Your original file is never modified, and the bytes never leave your device.

Two limits are worth knowing before you start, and they are repeated next to the controls they affect. First, a watermark is a visual mark: it does not prevent copying, it does not prevent editing, and it can be removed by a PDF editor — this is not digital rights management and nothing here pretends otherwise. Second, this tool adds watermarks only. If your file already carries a watermark, the tool cannot take it out; it will only draw your new mark alongside it.

What the mark is is real, though. Text watermarks are written as genuine text in the page content stream using the standard Helvetica font, so the wording stays selectable, copyable and searchable. Image watermarks are embedded as an image object, so the picture keeps its own pixels and is not re-drawn into the page background.

How to use this PDF watermarker

Press Choose a PDF file and pick the document. The page shows the file name, the page count and the size, and it tells you if the file cannot be read as a PDF — an encrypted file, for example, is refused instead of half-read.
Choose the watermark type. Text is the default: type the wording you want, for example CONFIDENTIAL or Draft — 14 Sep 2026. Image swaps the text field for a PNG or JPG picker and a scale slider.
Set the look: the font size in points, the colour, the opacity slider, the rotation preset and the layout. The orange box under the controls is there to remind you what a watermark can and cannot do.
Choose the pages. Leave the field as all to stamp every page, or name the ones you want such as 1-3,7. Pages are stamped in ascending order whatever order you type them in.
Press Add watermark and build the PDF. The tool reads the file, draws the mark on each selected page, writes the new document and then opens that new document again to check the page count, the page boxes and the rotation against the input.
Read the report, then press Download to save the result under a new name: your file report.pdf comes back as report-watermarked.pdf, and the original stays exactly as it was.

Worked example

The test file is wm-6page.pdf, built for this page: six pages with a different width on each one (400, 401, 402, 403, 404 and 405 points, each 700 points tall), a line of text on every page, and a rotation set on two of them (page 3 turned 90°, page 5 turned 180°). The uneven widths are deliberate — they make it impossible for a rewrite to shuffle or resize the pages without the check noticing. The file is 2,009 bytes (1.96 KB). Every number and every quoted line below came out of that file on this page, in a browser.

Text watermark, default settings: CONFIDENTIAL, Helvetica at 48 pt, red #ff0000, 30% opacity, 45° rotation, a single mark centred, all six pages. The tool reported:

Added a “CONFIDENTIAL” text watermark in Helvetica at 48 pt, #ff0000, 30% opacity, 45°, single mark centred, to 6 pages (all 6 pages). Original 1.96 KB → download 3.61 KB. Press Download to save wm-6page-watermarked.pdf.

Output check on the file the tool just wrote: same page count (6), same page boxes (400×700, 401×700, 402×700, 403×700, 404×700, 405×700) and the same page rotation (0°, 0°, 90°, 0°, 180°, 0°). 6 marks drawn in total.

The download was wm-6page-watermarked.pdf, 3,699 bytes. Decompressing its content streams and reading them directly shows exactly what was added to the document:

Six text-showing operations, one per page, each the string (CONFIDENTIAL) Tj — the mark is real text, not a picture of text.
Six graphics states carrying /ca 0.3, the fill alpha that makes each mark 30% opaque — and none in the original file, which is how you can tell the transparency is the tool's own doing.
The original six (Original page N text) Tj operations still there, untouched, and /Rotate 90 still on page 3's page object.
The same document information dictionary as the input — title, author, creator, producer and both dates — unchanged.

Rendering page 1 with Ghostscript confirms the geometry rather than just asserting it: the mark is red, sits across the middle of the page rising from bottom-left to top-right at 45°, is visibly transparent, and leaves the original line of blue text at the top of the page exactly where it was.

Page ranges. With the range 1-3,7 the tool reported “…to 3 pages (pages 1-3). 1 page number you named (7) does not exist in this 6-page file, so it was left out.” — three marks drawn, 2,954 bytes written, and no seventh page invented. The range 9 produced no download at all and answered “Nothing was saved: None of those pages exist in this file, which has 6 pages. Use numbers between 1 and 6, or type all to stamp every page.”; the range abc answered “Nothing was saved: “abc” is not a page or a page range.” In both refusals the download link was removed and the file on disk was never touched.

Tiled layout. Tiled repeats the mark in a grid spaced at six times the font size, which is a spacing that wants small text. At 24 pt on this file the tool drew 48 marks — 8 per page on average — and the rendered page shows eight whole copies with none clipped by the page edge; at the default 48 pt the same file takes 12 marks, 2 per page, because the spacing grows with the size. The grid is computed from each page's own width and height, so the 400 pt page and the 405 pt page each get their own arrangement rather than one page's layout copied to the rest.

Image watermark. The same six pages with a 96 × 32 px PNG (164 bytes) at 40% of each page's width, no rotation, 30% opacity: the output was 3,636 bytes. Inspecting it shows one image object — /Subtype /Image with /Width 96 — drawn six times inside six /ca 0.3 graphics states. The picture is stored once and referenced six times and is not re-encoded on the way in, so a small logo stays a small logo. Tilted to 45° with everything else unchanged the output was 3,858 bytes, with the same page count, page boxes and rotation.

What was refused on purpose. Watermark text in Devanagari was refused with “Could not build that PDF: that wording cannot be written with the standard Helvetica font…” instead of producing a file with missing glyphs. An empty text field, a file named .png whose bytes are plain text rather than a PNG, and a PDF saved with an open password were each refused with a plain reason and no download in every case. The browser console reported no JavaScript error during any of these runs.

Who is this PDF watermarker for?

Anyone sending a draft out: put DRAFT or NOT FOR DISTRIBUTION diagonally across every page of a proposal before it leaves your desk, so a copy that turns up later in someone else's inbox is labelled with where it came from.
Offices dealing with confidential paperwork: stamp contracts, payslips, invoices and medical letters with CONFIDENTIAL or a client file number, on a work laptop, without an upload and without installing a desktop editor.
People handing out samples: mark a preview or a partial document with a company name and a date, so the recipient can see at a glance that what they hold is not the final version.
Teachers, trainers and students: add a class name, a roll number or a SPECIMEN mark to a worksheet packet, a specimen paper or a submission the examiner asked to be tagged.
Photographers, designers and studios: drop a logo or a wordmark across a low-resolution proof before a client sees it, at an opacity that leaves the artwork readable.

How the mark is written into the page — and what a watermark cannot do

The tool appends a drawing to the content stream of each selected page. Nothing is rasterised, nothing is re-rendered and no page is replaced: the stream that was already there stays byte-for-byte, and the mark is added after it, which is why the original wording on the page keeps its fonts, its positions and its selectable text.

Text marks are text. The wording is written with the standard Helvetica font as a text-showing operation, so it is selectable, copyable and searchable in the output — the same way any other text in the file behaves, and it is also plain to see in the file's content stream.
Image marks are embedded images. A PNG is placed as a PNG and a JPG as a JPG, at the width you asked for as a percentage of that page's own width. The image is not converted on the way in, so a small logo does not turn into a large one.
Opacity is real transparency. The slider is written into the page's graphics state as a fill alpha — /ca 0.3 for 30% — so the page text shows through the mark instead of being hidden by it. That is the ordinary PDF transparency mechanism, and it was confirmed by rendering the output: the red mark comes out pale and the text under it stays readable.
Rotation happens about the mark, not the page. The mark is rotated about its own drawing point and then positioned from the four corners of its rotated box, so a 45° mark is centred on the page rather than hanging off one edge, and a 90° mark runs up the page instead of across it.
The page structure is left alone. Page count, page boxes and page rotation are copied straight through, and the tool re-opens its own output and compares all three with the input before it offers you the download — that is the “Output check” line you see after every run, and it is a measurement of the file, not a promise.
A page that carries its own rotation is left alone. If a page in your file already has a stored rotation of 90° or 180° — common in scans and in exports from some apps — the mark is placed in that page's own unrotated coordinate box and the viewer turns it along with the page. The page keeps its rotation exactly as it was, but the mark can end up a little off-centre in the view you see, rather than perfectly centred. Rewriting the page geometry would change the page box too, and that trade is not worth making for a watermark.
Metadata is not rewritten. The document information dictionary that was in your file — title, author, subject, creator, producer and the two dates — is carried into the output unchanged, so the tool adds no fingerprint of its own.

What a watermark cannot do. It cannot stop a copy being made: the page content under the mark is unchanged and fully available, and anyone who opens the file in a PDF editor can delete the added drawing, cover it with a white rectangle, or simply crop it away. It is not encryption, it is not a signature and it is not a usage restriction — the PDF format's own permission flags are a different tool, and even those are advisory. Treat a watermark as a label that follows the document around and makes casual re-use obvious, which is exactly what it is good at.

One hard limit on text. Standard PDF fonts cover the Latin-1 character set, and the browser cannot add a font it does not have; so Devanagari, Arabic, Chinese, Cyrillic, Greek and emoji are refused with an explanation instead of being dropped or mangled. Two honest ways round it: transliterate into Latin letters, or render the wording elsewhere and place it as an image watermark, where the wording is a picture and can be any script you like.

Common use cases

Marking a document DRAFT or SAMPLE so nobody mistakes an early copy for the final one.
Putting a company name, a client reference or a date across a PDF that is going to a third party.
Labelling a confidential payslip, contract or medical letter before it is printed or filed.
Stamping a scan batch with the date it was received, so the pages are traceable after they are copied.
Placing a studio logo diagonally across a low-resolution proof before the client sees it.
Adding a watermark to the pages of a PDF without touching the pages that do not need it.
Watermarking a file on a locked-down work machine where installing a desktop PDF editor is not an option.

Pro tips

Opacity 20–35% is the readable zone. Above that the mark competes with the page text; below it the mark is easy to miss in print. Start at 30 and look at a page.
Match the mark to the page shape. A long line of text at 0° usually overflows a portrait page. The 45° preset fits far more wording into the diagonal, which is why it is the default.
Use the bottom-right corner for a quiet label. A small grey bottom-right mark suits a file reference or a date, because it never crosses the body text.
Tiled is for “impossible to miss”, and it wants small text. The spacing is six times the font size, so a 48 pt mark gives only a couple of copies per page while a 24 pt mark tiles a page properly — on the test file, 24 pt produced 8 copies per page against 2 at 48 pt. A tiled mark also makes a page harder to read, so pair it with low opacity.
Stamp only what needs it. On a long pack, a page range like 1,2,50-54 puts the mark on the cover pages and the appendix and leaves the middle clean.
Test one page first. Set the range to 1, run, and open the download to check the size, colour and position before stamping the whole document.
Keep the original. The download is a new file and the original on your disk is untouched, so a watermark you do not like costs one extra run, not a re-download of the source.

FAQ

Are my PDFs uploaded to a server?

No. The file is read from your disk by your own browser, the watermark is drawn with the pdf-lib library inside the tab, and the new file is written back to your device. Nothing is uploaded, nothing is stored on a server, and no account or email address is needed.

Does a watermark stop people copying or editing my PDF?

No, and nothing on this page will tell you otherwise. A watermark is a visual mark drawn on the page: the text underneath is still there, still selectable and still copyable, and anyone with a PDF editor can remove or cover the mark. Use a watermark to label and identify a document, not to control what happens to it.

Can this tool remove a watermark that is already in a PDF?

No. This tool adds watermarks only. If a file already carries a watermark, the mark you add is drawn on top of it and the existing one stays exactly where it was. There is no removal, erasing or covering-up step here.

Will the text in my PDF still be selectable after watermarking?

Yes. The mark is written as real text in the page content stream using the standard Helvetica font, so it is selectable, copyable and searchable like the rest of the page. The wording that was already on the page is untouched and stays selectable too.

Does watermarking change the page count, the page size or the rotation?

No. The number of pages, each page's dimensions and each page's rotation are the ones the original file had. The tool re-opens the file it has just written, compares the page count, the page boxes and the rotation against the input, and reports the result next to the download. Only the page content gains the mark.

What do the four rotation presets do?

They set the angle of the mark in the page's own coordinate system: 0° is horizontal, 45° rises to the right, -45° falls to the right, and 90° runs vertically up the page. The mark is rotated about its own drawing point and then positioned from the corners of its rotated box, so it stays centred on the page, or in the corner, instead of drifting off it.

How does the tiled layout work?

The mark is repeated in a grid across the page, with the spacing taken from the mark's own size — six times the font size for text. Each copy is drawn only when its rotated box fits inside the page, so the grid is built from that page's own width and height and a page of a different size simply gets a different number of copies.

Can I watermark only some pages?

Yes. Leave the page range as all, or type a list such as 1-3,7 or 2,5,9 to stamp only those pages, in ascending order. Numbers that do not exist in the document are ignored rather than guessed at, and the tool reports how many of the pages you named were actually stamped.

Can I use an image or a logo as the watermark?

Yes — switch the type to Image and choose a PNG or JPG. The scale slider sets the image width as a percentage of each page's own width, so one setting looks right on a document with mixed page sizes. The picture is read in the tab like the PDF, and only that image is embedded in the output.

Why was my watermark text rejected?

Because the built-in Helvetica font can only encode the Latin-1 character set: Latin letters with their accents, digits and common punctuation. Text in Devanagari, Arabic, Chinese, Cyrillic, Greek or emoji cannot be written with a standard font, and this page does not embed a font file for every script. Use a Latin transliteration, or put a picture of the wording on the page as an image watermark.

Can I watermark a PDF that is password-protected?

No. An encrypted file cannot be read, and the page says so instead of writing a broken download. Remove the password in a PDF reader, or print the file to a new PDF, then add the watermark to the unlocked copy.

Is it free, and is there a limit?

Yes, it is free: no sign-up, no email gate, no limit on how many files you watermark, and nothing is added to your PDF beyond the mark you asked for. Everything is processed in your browser's memory, so the practical limit is your own device — a few hundred pages is comfortable, and a very large file simply takes longer.