One page per file
Every page creates a separate PDF. A scanned page remains an image; this does not recognize its words or repair its quality.
Split PDF
Separate one PDF into smaller files, keeping every original page in order.
No PDF selected
One PDF · 2–200 pages · up to 50 MiB · no forms, signatures or internal PDF links
Choose one PDF to start.
Review original pages and add cut points
Adding a cut switches to “After chosen pages.” The last page cannot be a cut point.
Your PDF parts
In your browserNo uploadsNo account
How to split a PDF
- Choose one document
Select a PDF and wait for its local check. Use the original-page preview when deciding where a section ends.
- Set the cut points
Separate every page, use a fixed number of pages per part, or enter positions after which to split. All original pages stay in order.
- Review and download
Preview each generated part. Download the PDF you need or save all parts in one ZIP, then open the files before sharing.
Split PDF document pages
Regular groups or chosen cuts
Every few pages makes equal-sized groups, with a shorter final part if needed. After chosen pages makes contiguous sections: cuts 2, 4 in a five-page source produce pages 1–2, 3–4 and 5.
PDF splitter, free to use
Use this PDF splitter free of charge. There is no account, daily conversion quota or added watermark. This page splits one document into multiple files; it does not merge sources, reorder pages or select a disjoint subset to discard the rest.
Support and limits
Files and resource guards
One unencrypted PDF, 2–200 pages, up to 50 MiB and 20,000 PDF objects. There must be at least two output parts. Generated PDFs combined and the ZIP are each limited to 100 MiB. Checking a file and creating its parts each have a 60-second timer, including their initial preview. Cancel stops the current operation and removes its unfinished downloads. Background browser throttling can delay timer delivery. Device memory can impose lower practical limits.
Page content and links
The tool copies pages into new PDFs rather than retyping their text or making screenshots. Review text, image placement, page sizes and rotation in the downloads. HTTP, HTTPS and email link destinations are retained in the checked output. Opening an email link still depends on your PDF reader and configured email application; this tool does not send email.
Unsupported document features
Encrypted or damaged files, forms, signatures or certification restrictions, comments, non-link annotations, internal navigation, layers, portfolios, attachments and document actions stop conversion. Bookmarks, page labels, document metadata and accessibility tags are not rebuilt. Keep originals that rely on these features.
Split points, not file-size targets
Page numbers are physical positions starting at 1, not printed page labels. Cuts must increase without duplicates and cannot include the final page. Splitting does not promise compression, a particular file size or OCR. It does not certify PDF/A archival or PDF/UA accessibility compliance, rebuild accessibility tags or preserve digital signatures.
Test files and results
On October 27, 2026 we split one original English five-page PDF containing two Letter report pages, a rotated/cropped A4 appendix, an image-only receipt and a memo. The table links to actual ZIP downloads, each containing PDF parts.
Five-page source PDFPage-by-page check recordsSource notes
| Setting | Actual output | Observed result |
|---|---|---|
| Every page | Five one-page PDFs (ZIP) | All five original pages present once, in order. The rotated/cropped appendix and image-only receipt stayed as their source pages. |
| Every few pages: 2 | Three parts (ZIP) | Groups 1–2, 3–4 and 5. The shorter last group retained the memo. A separately downloaded first part matched its ZIP entry byte for byte. |
| After chosen pages: 2, 3 | Three custom parts (ZIP) | Groups 1–2, 3 and 4–5. Cut points were added using the original-page preview. The receipt and memo remained together in the final part. |
What we checked
All 11 PDF parts and 15 output pages were rendered and viewed. Text, page drawing commands, media/crop boxes, rotation, HTTPS link targets, embedded image data and rendered pixels at 90 dpi matched their original pages. The receipt stayed image-only; no OCR was added.
A 390-pixel browser-window run selected the same real file, entered cuts, created parts and downloaded a ZIP. Its page content, boxes, rotation and images matched the desktop result. This was not a physical-phone test.
Rejected inputs and settings
Eight file checks rejected encryption, forms, internal links, damaged structure, 201 pages, an input above 50 MiB, a non-PDF extension and a single-page PDF. Eight invalid cut/interval settings offered no completed output. Changing a method invalidated old downloads.
An additional synthetic PDF with a cryptographically verified, self-signed digital signature was rejected. This checks the refusal behavior; it does not certify trusted signer identity or preserve signatures. Download the signed refusal sample.
Page and input boundaries
A 200-page English source produced 200 one-page PDFs. Every page kept its original numbered text, drawing commands and page size; all rendered pages matched the source pixels.
A valid two-page PDF padded with an unused stream to exactly 50 MiB completed; one byte above was rejected. This checks the byte limit, not the performance of a complex 50 MiB document. Separate processing checks accepted 20,000 PDF objects and rejected 20,001.
Output limits and interrupted work
A source with a shared image exceeded the 100 MiB combined-PDF limit after copying that image into separate parts and was stopped. A separate source stayed below that combined limit but exceeded 100 MiB after ZIP packaging and was also stopped. Neither case offered a completed download.
Controlled stalls exercised Cancel during background processing, source-preview loading and result-preview loading. The 60-second timer was also exercised on stalled checking, splitting and source-preview loading. Controls recovered and unfinished downloads were removed. These are intentional interruption checks, not timing measurements for ordinary documents.
External link destinations
The three-link source PDF and its downloaded parts contain HTTP, HTTPS and email URI annotations. Their complete destination strings, including the email subject, matched after splitting.
This checks PDF link data. It does not establish that every reader will launch a browser or email application, and no email was sent.
Device scope of these checks
The resource and interruption checks ran on this Windows desktop; a 390-pixel check examined the narrow layout. On October 27, 2026, the site owner separately reported success on iPhone 14 with Safari and Redmi Note 10 Pro with Chrome: select the five-page source, use Every page, download the ZIP and open an output PDF. These are owner-reported checks of that fixed file, not measurements of phone speed, memory or battery use.
Page counts, file bytes and object counts are separate limits; a file below them can still be too demanding for a device. Download the boundary check records for the exact files and test scope.
Local processing and privacy
Local documents and downloads
Your PDF, its name, the generated parts and ZIP are handled in this browser, not uploaded to a conversion server. Components load from this site. Clear the file or close the page to release its working state; downloads remain on your device.
Page requests and analytics
Page delivery and analytics have separate requests, described in our privacy policy. File names, document contents and conversion results are not sent as analytics events. There is no account or saved conversion history.
Frequently asked questions
Why did the downloads disappear after I changed a setting?
The old parts represented the previous cuts. Split again so the visible downloads match the current settings.
Why is the last part shorter than the others?
A fixed interval includes the remaining pages as the final part. For example, five pages split every two pages become groups of two, two and one.
Can I split a signed or fillable PDF?
No. These documents are rejected because copying their pages does not preserve the original digital signature or interactive form behavior. Keep the original document.
Can I split into files below a chosen number of megabytes?
No. Splits follow page positions, not output byte targets. Images and shared resources can make part sizes uneven, and the sum of part sizes can exceed the original.
Page updated: