Remove specific pages from any PDF with visual previews. Select the pages you do not need and download a clean file — processed privately in your browser.
You have a PDF that is almost right. There is a blank page the scanner produced, or an internal cover sheet with a reviewer's name on it, or three pages of an appendix that belong to a different client. Everything else in the document is correct and you do not want to touch it. What you want is the same file, minus those pages, and nothing else changed.
That is the whole job this tool does. You load the PDF, click the numbers of the pages you want gone, and download a new PDF containing the pages you kept, in their original order. The file never leaves your browser: the PDF is read with the File API into memory, rebuilt with the pdf-lib library that loads into the page itself, and handed back to you as a local download link. There is no upload step and no server that sees your document, which matters when the reason you are deleting a page is that it contains something you do not want other people to read.
This is the first thing worth being precise about, because most tools in this category ask you to type something like 2,5-8 and this one does not. After the file loads, the tool reads the page count and draws one small numbered button per page, from 1 to the last page. Clicking a number selects it for removal and it turns red. Clicking it again deselects it. A running counter above the grid shows how many pages are selected and how many will remain, and it updates on every click, so you can confirm the arithmetic before you commit to anything.
There is no text box for range syntax. Two shortcut buttons sit under the grid — Select All and Deselect All — and Select All is genuinely useful in an inverted workflow: select every page, then click back the handful you want to keep. On a 60-page document where you only need pages 1, 2 and 47, that is three clicks after Select All instead of fifty-seven.
Two consequences of the button-grid design are worth knowing before you start:
3.These two tasks get confused constantly, and choosing the wrong one wastes a round trip. The distinction is simply how many files you end up with.
Removing produces exactly one output file: your original document, shorter. The pages you did not select stay together in one PDF, in the same order, and the pages you selected are gone. Use it when the document is still one document and you are correcting it.
Splitting produces several output files, one per range or per page. Use it when you want the pages separated — pulling a signed signature page out to send on its own, or breaking a scanned batch back into the individual documents it came from. Our PDF splitter handles that side, including typed page ranges.
A quick test: if your sentence is “I need to send them the deck without the pricing slides,” you want removal. If it is “I need the pricing slides as their own file,” you want splitting. If you want both — the deck without pricing, and pricing on its own — do them as two separate operations on the same original.
A 24-page contract comes back from a document feeder that scanned both sides of every sheet, and the reverse of each page was blank. The blanks landed on the even page positions. In the grid you click 2, 4, 6 and so on through 24 — twelve buttons. The counter reads 12 selected · 12 remaining. You press Remove, and the download is a 12-page PDF with the original odd pages in order.
The output filename is derived from the input: a file called contract.pdf comes back as contract_pages_removed.pdf. That naming is deliberate and worth leaving alone until you have opened the result — it stops you overwriting the original in your downloads folder while you are still checking the work.
A second example in the other direction. A 60-page internal review has a cover memo on page 1 that names the reviewers, and the client should not see it. You click 1. The counter reads 1 selected · 59 remaining. That is the entire operation. The point of the tool is that removing one page out of sixty does not require re-exporting the whole document from the application that made it, which is where formatting tends to shift.
The tool does not edit your original file in place. It creates a brand-new empty PDF and copies the pages you kept into it, one by one, in their original sequence. That rebuild is what makes the operation safe — the source bytes are never modified — but it also explains precisely what carries across and what is left behind.
| Part of the file | What happens |
|---|---|
| Page content (text, images, vectors) | Copied intact, at full quality. Nothing is re-encoded or re-compressed, so a kept page is byte-for-byte the same visual result as before. |
| Page order | Preserved. Kept pages stay in their original relative sequence; you cannot reorder them here. |
| Page size and orientation | Preserved per page. A mixed portrait/landscape document keeps its mix. |
| Printed page numbers | Not renumbered. A footer reading “Page 7 of 24” is drawn onto the page as artwork, so it still says that after the document becomes 12 pages long. |
| Bookmarks / outline | Not carried into the new document. The rebuild copies pages, not the document-level outline tree. |
| Interactive form fields | Field appearances travel with the page, but the document-level form structure does not, so filled forms are best flattened before you remove pages. |
| Internal links and cross-references | A link that pointed at a page you deleted has nothing to point at any more. Links between two kept pages may also not survive the rebuild — check them if your document relies on them. |
| Document metadata (title, author, subject) | Not copied. The output carries fresh metadata rather than the original's, which is often an advantage when the author field is the thing you were trying to get rid of. |
| Encryption / passwords | Cannot be handled. A password-protected PDF will not load at all. |
The two rows people get caught by are the printed page numbers and the bookmarks. Neither is a defect in the tool; both follow from what a PDF actually is. A page number in a footer is ink on the page, not a live field, and no page-removal tool can update it without re-running the layout engine that drew it. If the renumbering matters — a document going to a court or a regulator, say — remove the pages here and then add a fresh Bates or page-number stamp afterwards, or go back to the source document and re-export.
Usually yes, and roughly in proportion to what the removed pages contained. Deleting twelve blank scanner pages from a 24-page scan removes twelve page images, and the file shrinks accordingly. Deleting twelve pages of plain text from a document whose bulk is one large embedded photograph on page 3 will barely move the number, because the photograph is still there.
The tool shows you the exact size of the result next to the download button, so you never have to guess. It is worth saying plainly what this operation is not: it is a page-level edit, not a compression pass. Nothing inside the pages you keep is downsampled, re-encoded or optimised. If the file is still too large after removal, that is a separate problem with a separate tool.
| Message | What it means |
|---|---|
| Please upload a PDF file. | The file you dropped is not identified by the browser as application/pdf. A .docx, an image, or a PDF whose extension was renamed will all trigger this. Convert it to PDF first. |
| Failed to read PDF. The file may be corrupted or encrypted. | The parser could not open the document. In practice this is nearly always a password-protected file. It can also be a genuinely damaged PDF, or one written by an unusual generator with a malformed cross-reference table. |
| Cannot remove all pages. At least one page must remain. | You selected every page. Deselect at least one. |
| Failed to remove pages from PDF. | The document opened but the rebuild failed part-way. Rare, and usually a sign of unusual internal structure. Re-saving the file from a PDF reader and trying again generally clears it. |
The encrypted case deserves a note because it surprises people. If a PDF asks for a password when you open it, this tool cannot read it either — the page contents are encrypted and there is nowhere to type the password. Open it in your reader with the password, save an unprotected copy, then remove pages from that copy.
Really gone, in the sense that matters: the output file is built from scratch and the deleted pages were never copied into it, so there is no hidden layer to recover them from. That is a stronger guarantee than you get from tools that mark pages as invisible. It is not the same thing as redaction, though. Removing a page removes the whole page. If the sensitive material is a paragraph in the middle of a page you are keeping, deleting pages does not help you — you need to redact that page properly, and drawing a black box over text in most editors leaves the text underneath it.
You do not need to. Your original file is untouched on disk, and the result arrives as a separate download with _pages_removed in the name. If you selected the wrong pages, go back, adjust the selection and press the button again; a new result replaces the old one.
There is no fixed limit written into the tool, but there is a real one: your browser tab's memory. The whole file is held in memory and a second copy is built alongside it, so peak usage is roughly twice the file size plus overhead. Ordinary documents of a few hundred pages are unremarkable. Very large scanned archives — hundreds of megabytes — are where a tab runs out of room, and the symptom is the tab becoming unresponsive rather than a tidy error message. On a file that big, work on a desktop application.
No. The tool takes one file at a time, which is a consequence of the page grid: the selection is specific to one document's page count. Load, remove, download, then press Reset and load the next.
Once the page has loaded, the processing itself needs no network. Everything — parsing, copying pages, writing the new file — happens in JavaScript inside your tab. This is worth stating because it is the honest answer to the question people actually have, which is whether a document they are cleaning up before sending it out has quietly been posted to somebody's server first. It has not.