creatorvalet Search

Delete pages from PDF

Click the pages you want gone and take back the rest. Runs in your browser — no upload, no account.

  • SECURE
  • NO UPLOADS
  • NO SIGNUP
  • BROWSER BASED
  • FREE
  • FOREVER.

Support us with a link or a share

Drop your PDF here

Nothing is uploaded. The pages come off here, from bytes that never leave this tab.

How to delete pages from PDF files without handing over the document

Think about why anybody reaches for this. A scanned stack came back with a blank sheet between every invoice. The last page of a lease is somebody else's signature block. A cover memo carries a home address that has no business traveling with the rest. In almost every case the page being deleted is the reason the file is sensitive at all — and the usual advice, repeated by search engines and by the tools themselves, is to upload the whole document to a server so that one sheet can be taken off it. The page you wanted gone is the first thing that leaves the building.

Here the document stays where it is. Underneath its appearance a PDF is a numbered catalog of objects rather than an image, so the library working inside this tab walks that catalog, gathers the sheets you left alone, and assembles them into a new document — arithmetic on references, essentially, which is why a forty-page report is finished before you have registered that it began. Routing that through a data center would buy you a queue and somebody else's disk holding your paperwork. Every site can type the phrase in your browser, so distrust ours and go looking: developer tools, network view, mark a sheet, press the button. Better still, run it once so the PDF writer is fetched — that first press is the last thing this address is asked for — then cut the machine off the network and trim the rest of the pile anyway. Checking a promise like that takes about two minutes on any site, and ours deserves it as much as anyone's.

Two ways to point at a page, and why both are on screen

The thumbnails come first because a page is recognizable and a page number is not. Nobody remembers that the duplicated receipt was sheet nineteen; they remember that it was the crooked one. Click it, it is marked with a cross and the word removing, click it again and it comes back. Marks survive everything except handing in a different file, so changing your mind never costs a restart.

A thumbnail is small on purpose, because forty of them on one screen is the point. When a page is too close to call at that size, the button in its corner opens it at whatever size your window allows — several times taller than the tile, with arrows to step through the document and a button to mark or unmark the page you are looking at. It is a way to read a page, not a second place where decisions are stored: whatever you do there shows up in the grid behind it immediately.

The box above the grid is not a fallback for people who dislike clicking. Past thirty or so sheets, hunting for three of them by eye is worse than typing, and anybody clearing the blank separators out of a long scan already has the numbers in their head. What it takes is the notation people reach for anyway — 1-3 for a run, 12- for everything from twelve onward, -4 for everything up to four, last when you would rather not count, and the longer dash a word processor slips in behind your back. Grid and box are two views of one selection, so typing lights up tiles and clicking rewrites the string. No third hidden state exists for them to disagree about.

Previews are drawn as they come into view rather than all at once. That is a deliberate answer to the size question: nothing is ever refused for being too big, so a four-hundred-page file has to behave, and rendering four hundred images the moment it lands is precisely how a browser tab stops responding. The counter under the grid reports previews genuinely finished. A page that will not draw is still a page you can remove — the tile keeps its number and its checkbox and says nothing reassuring that is untrue.

When our splitter is the shorter route

It would be dishonest to sell this as something the rest of the site cannot do. Splitting a PDF cuts the document into consecutive pieces at the page breaks you mark, which is the same machinery seen from the other end. If you want several documents out rather than one, that page asks the question in the form you are already thinking in and you will finish faster there.

This page exists because the opposite intention is awkward to express as a keep-list. Somebody with a six-page scan who wants the fourth sheet gone is not composing 1-3, 5-6 in their head; they are pointing at something and saying no. At six pages that difference is a matter of taste. At two hundred it is the difference between a gesture and a spreadsheet exercise, and getting the inverted range slightly wrong is a mistake you will not notice until somebody else does.

What comes through unchanged, and what does not

No sheet is re-encoded on the way through. Type, artwork and measurements survive untouched, which means a Letter sheet inside a file that also holds A4 comes back as Letter, and words remain words rather than a picture of words. That second property is the one destroyed when a tool re-renders each sheet as an image, and its loss tends to surface months later, when somebody tries to search the file and finds nothing. If you want pictures rather than a document, PDF to PNG draws them.

Three document-level structures do not survive. Fillable form fields and bookmarks hang off the document as separate indexes rather than belonging to the sheets, and the source Info fields and XMP metadata describe the old document as a whole. The file you get back is a new document those structures do not travel to. We measured the alternative before picking a path: deleting pages from the original in place does preserve the form, but a three-page form with its middle sheet removed came back reporting three fields across two pages, one of them attached to a sheet that no longer existed. On an encrypted source that same route left the encryption dictionary behind and produced a file no reader would open at all. A count of any fields your document carries appears before you run anything, together with bookmark, metadata and internal-link counts, so the trade is visible in advance rather than discovered afterwards. Whatever somebody typed into a field stays legible, because it was painted onto the page as well. An internal page link is different: it is rewritten to the new page reference when its target stays in the file, and dropped when that target is removed.

The things that go wrong, and what is said about them

Mark every sheet and the tool stops and explains itself instead of quietly disabling a button. Zero pages is not a very short PDF, it is not a PDF: the library will cheerfully write a document with an empty page tree, and the result opens as a blank window in some readers and not at all in others. At least one page has to stay.

Protection on a PDF covers two situations that behave nothing alike. The weaker sort merely requests that readers refrain from printing or copying; the document itself opens for anyone, so the tool tries the handle first and, finding it turns, never bothers you about a key you were never given. The stronger sort actually seals the contents, and there a box appears and no sheet moves until what you type genuinely unlocks it. The move we decline to make is the one most libraries suggest: switching off the encryption check while every scrambled stream is copied, untouched, into a container that claims to hold plain text. The rebuilt file drops the source author name, title and editing-history metadata rather than carrying it into a new document under false pretences. If metadata is the only thing that needs changing, stripping the metadata from a PDF does that without removing a sheet. When the part that has to go sits on a page that has to stay, redacting the PDF removes the words rather than hiding them, and when the finished document still needs a name at the bottom, signing a PDF is the last step. If sheets from elsewhere need to replace the ones you took out, merging PDF files runs the other way.

And when nothing is wrong with any sheet except how much of it shows — a scanner border on every one, a black surround around an exported slide, a running header nobody outside should read — discarding sheets is the wrong instrument entirely. Cropping a PDF narrows the displayed boundary on the sheets you keep, and will tell you plainly whether what it trimmed is merely out of view or actually gone, since those are two different outcomes that the same word usually covers.

Have an idea for this tool?

Tell us what would make this tool more useful, or suggest another tool you would like us to build.

Questions

Does my PDF get uploaded when I delete pages from it?

It never leaves the machine. Your browser reads the bytes from wherever the document already sits, rebuilds a fresh one around the sheets you did not mark, and hands back a link into memory nothing outside this tab can reach. Rather than take that on trust, go and watch it happen: developer tools, network view, mark a sheet, press the button, and see whether anything at all goes out. There is a small irony behind why this matters. The sheet somebody wants gone is very often the sensitive one — a cover page with a home address, an internal note stapled to a report, a salary column left in an appendix — so a service that starts by copying the whole document to a server has already received the very thing you were trying to be rid of.

How do I remove pages from a PDF?

Hand in the file and the pages appear as thumbnails. Click the ones you want gone and they are marked; click again and they come back. If you would rather type, the box takes 1-3, 7, 12- and the two stay in step in both directions — marking a thumbnail rewrites the box, and typing in the box moves the marks. Then download what is left. On a long document the box is usually faster: nobody scrolls four hundred thumbnails to find the three they want rid of.

What if I mark every page?

The tool says so and does not run. A PDF cannot have zero pages, and the failure would not look like one: the library will happily write a document with an empty page tree, and the result opens as a blank window in some readers and not at all in others. That is the quiet kind of broken this site exists to avoid, so instead of a silently disabled button you get a sentence explaining that at least one page has to stay.

Is this different from splitting a PDF?

It is the same machinery reached from the other end, and it would be dishonest to pretend otherwise. Our splitter cuts a document into consecutive pieces at the page breaks you mark, so if you want several files rather than one, that page is the shorter route. This page exists because the opposite intention is genuinely different to express. Somebody dropping a six-page scan who wants the fourth sheet gone is not thinking in terms of "keep 1-3 and 5-6"; they are pointing at a page. On a forty-page document that difference stops being a matter of taste.

Do bookmarks, metadata and fillable form fields survive?

Those document-level structures do not. The result is a new document around the sheets you keep, so source Info fields and XMP metadata, bookmarks and fillable form controls are not carried across. The form fields, bookmarks and internal links are counted before you run anything, because they change what you get; the metadata count appears on the receipt. Text somebody typed into a field remains visible as drawn ink. Internal page links follow a narrower rule: a link is rewritten when its target remains in the result and dropped when that target page is removed.

What happens to a password-protected PDF?

Two quite different situations wear the same badge, and most tools treat them as one. Protection meant to discourage printing and copying stops nobody opening the document, so we try the handle quietly and, when it turns, get on with the job without troubling you for anything. Protection that genuinely seals the contents is another matter entirely: a box appears, and no sheet is touched until what you type really unlocks the thing. The one maneuver we will not perform is the one nearly every library recommends, which switches off the encryption test without decrypting a single byte. Files made that way report the correct number of sheets, raise no error whatsoever, and hold not one legible character.

Is there a page or file size limit?

Nothing is turned away. Something enormous earns a sentence about how much memory the job is likely to want, and then the job goes ahead, since the computer belongs to you and so does the judgment about what it can cope with. Four hundred sheets is an ordinary Tuesday rather than a mistake. Instead of refusing we simply decline to make it painful: previews appear as you scroll toward them, one at a time, in place of four hundred images being drawn the instant a document lands. The number under the grid counts previews that genuinely finished, not an estimate of how far along we might be.

Will the pages I keep look any different?

Nothing is re-encoded. Each kept sheet arrives carrying the type, the artwork and the measurements it already had, so a Letter sheet living in a file that also holds A4 comes back out as Letter. Words stay words instead of turning into a photograph of words, which is what befalls a document whenever a tool re-renders every sheet as an image — a loss usually discovered much later, by whoever tries to search it. Folio numbers printed on the sheets are ink like everything else on them, so a file that loses its fourth sheet will still read three, then five.