Cleaning a Watermark Out of a PDF
The file itself is a container rather than a picture — but the pages inside it can be exported, cleaned one at a time and put back.
A PDF is a container that can hold vector text, raster images, annotations and fonts at the same time, and which of those the watermark belongs to decides how you remove it. The Watermark tool here works on pictures, not documents: it reads PNG and JPG. So the route is to export the affected pages as images, clean those images, and rebuild the document from them.
That route has a real cost, and it is worth understanding before you start. Turning a page into an image makes any selectable text on it part of the picture, so it stops being searchable or copyable afterwards. On a scanned document that costs you nothing, because the page was already a picture. On a born-digital text PDF it costs you the text layer.
What it does
Built for this specific job, not a generic editor.
Works on pages, not on the file
The unit of work is one exported page image. The tool reads PNG or JPG, so a PDF has to become one of those before anything can happen.
Clean only the pages that carry the mark
If the mark sits on a cover or repeats in a footer, you decide which pages are worth ten credits each and which can be left exactly as they are.
Scanned documents lose nothing
Where the page is already a raster image, export and replace is close to lossless — the pixels you get back are the pixels you sent, minus the mark.
Resolution is your call
Export at the pixel size the document is actually read at. Under-exporting throws away the detail around the mark that the reconstruction needs.
How it works
Four steps, no timeline to learn.
Find out what kind of mark it is
Open the file in a PDF reader and try selecting or moving the watermark. If it is a separate annotation or object, removing it there is the better path, because it leaves the text layer untouched.
Export the marked pages as PNG or JPG
Most readers and editors can write pages out as image files. Choose a size that matches how the document will be read, and prefer PNG so the export does not add compression noise.
Clean each page image
One page per run, 10 credits each, roughly 15 to 30 seconds. A twenty-page document is twenty runs, so count the cost before you queue the whole thing.
Rebuild and check what changed
Put the cleaned images back into a PDF and confirm the pages still read correctly. Expect the file to be larger, and expect text on those pages no longer to be selectable.
What people use it for
The jobs this actually gets used for.
Scanned contracts and forms
Documents that came off a scanner with a stamp or a review banner burned into the page image rather than layered on top of it.
Proof copies
Drafts circulated with a draft mark or a reviewer name across the page, where the version that goes out needs to be clean.
Exported slide decks
Decks saved out of a template that prints a logo or an event name onto every slide, so the mark repeats identically page after page.
Reports with a marked cover
Documents where only the first page carries the mark, which means one run is enough instead of twenty.
Why a PDF cannot be handled like an image
The container is the reason there is no direct upload, and the reason the honest workflow has three steps.
A raster image is a grid of pixels with one value per cell, which is exactly what an inpainting model works on. A PDF is a set of drawing instructions that may include text objects, vector paths, embedded fonts and placed images. Asking a pixel model to edit that directly means asking it to rebuild every object on the page, and the result would not be the document you started with.
So the work splits by document type, and the split matters. If your PDF is scanned or image-only, exporting a page is a faithful copy and the clean-up is straightforward. If it is a born-digital text document, exporting rasterises it, and you trade a searchable, selectable text layer for a picture of that text.
There is also a cheaper answer in one common case. When the watermark is a separate annotation rather than something burned into the page content, a PDF editor can delete it without touching anything else, and no credits are involved. Check that first — it takes seconds, and it is the only route that leaves the document fully intact.
Credits and plans
Every tool draws on one shared credit balance.
Ten credits per page image. The Free plan gives 30 credits a month with no credit card, so three pages before you pay; Starter at $9 gives 400 credits, Pro at $19 gives 1,000 with three parallel tasks and a priority queue, and Studio at $59 gives 4,000 with six tasks and API access. For a long document this is mostly a credit question rather than a time question.
Related guides
Longer reads on the same workflow, from the OpenFilmo blog.
AI Content Labeling Rules: What You Must Disclose in 2026
What AI content labelling rules require: what to disclose, where human review is mandatory, and how platform rules differ from law…
AI Video for Agencies: A Client Delivery Workflow
A client-ready AI video workflow for agencies: scoping honestly, rights clearance, approval gates, revision economics and how to p…
AI Video Watermarking and C2PA: Provenance Explained
How AI video watermarking, embedded metadata and C2PA content credentials differ, what survives re-encoding and screenshots, and h…
Related tools
Closely related pages on the same problem.
Frequently asked questions
Can I upload the PDF itself?
No. The Watermark tool accepts PNG and JPG only, so the pages have to be exported to one of those formats first. This page describes the workflow around that limit; it is not a claim that PDF files are supported.
How much does a twenty-page document cost?
Two hundred credits — 10 per page image. The Free plan covers three pages a month with no credit card; Starter at $9 covers the whole document and has credits left over.
Will the text still be selectable afterwards?
Not on the pages you rasterised. Once a page is an image, its text is part of the picture. If search and copy have to survive, check whether the mark is a removable annotation before converting anything.
Is it fine to remove a mark from a document someone sent me?
Removing a mark transfers no rights. If you do not hold the rights to the document, cleaning it does not change what you are permitted to do with it, and in most cases the honest answer is to ask for a clean copy.
Only one page has the mark — do I have to do them all?
No. Export and clean only the affected pages and leave the rest of the document as it is. Working page by page is slower than a bulk action but it is also what keeps the untouched pages intact.