Split PDF Guide: Page Ranges, Custom Groups and Odd/Even Output
Admin ·
Split PDF creates new documents from selected pages of one supported local PDF. ToolMellow offers one file per selected page, fixed page-count chunks, custom semicolon-separated groups, or original odd/even page groups. You can inspect the output counts, preview the source or results, and download individual PDFs or a ZIP.
The splitter copies page content without rasterizing it. That does not guarantee preservation of every document-level feature, smaller file sizes or original bytes. Source limits, repeated-page accounting, empty parity groups and preview limits matter, so this guide follows the actual controls rather than assuming every PDF splitting service behaves alike.
Choose one PDF and inspect the planned parts
Use Choose PDF files or drop one local PDF into the workspace. Wait for its page count to be read. The default mode is One file per selected page, with the file’s Pages field blank to include all source pages. Choose the mode and selection, check Output files and selected pages, then press Split PDF and inspect the generated parts.
This workspace accepts one source at a time. Clear or remove it before choosing a replacement; selecting a second PDF does not create a multi-source split. There is no content editor, page-arrangement grid, rotation selector, password box or target-byte control here. Keep the original separately.
- Choose original physical page positions rather than assuming printed labels are range input.
- Inspect both the number of output files and total page occurrences, especially when pages repeat.
- Check generated parts in an appropriate PDF reader before relying on forms, bookmarks or other document features.
MDN: local files and the File APIMDN: Web Worker messaging and lifecycle
Choose a mode by how you want pages grouped
One file per selected page creates a part for each selected occurrence, including repeated pages. Every N selected pages groups successive entries in your selection sequence; it does not require physically consecutive source pages. Pages per file starts at 1 and accepts whole values from 1 to 300. A final chunk can contain fewer pages.
Custom groups separated by semicolons lets each group define one part. Separate original odd and even pages partitions the chosen sequence by original page numbers, preserves order within each partition and drops empty partitions. A page-count chunk is not a fixed number of output files or an exact MB-size split.
| Mode | Grouping rule |
|---|---|
| One file per selected page | One output for every selected occurrence |
| Every N selected pages | Consecutive entries in the selected sequence; shorter final chunk possible |
| Custom groups separated by semicolons | One output for each nonempty group |
| Separate original odd and even pages | Original-number parity; only nonempty groups are output |
Range syntax uses original physical page positions
The file’s Pages field uses ASCII decimal numbers, commas and ascending hyphen ranges. Blank or whitespace-only text selects all source pages. Entries retain their entered order and repetitions: 3,1-2,3 selects original pages 3,1,2,3. A range cannot run backward, refer to page 0, exceed the source count or contain a missing comma item.
Printed page numbers and PDF page labels can differ from physical positions. Roman front-matter labels, named appendices and numbers drawn on a page are not this input grammar. Use positions from 1 through the source page count and inspect the source preview. The splitter does not rewrite printed page numbers or certify preservation of document-level page-label dictionaries.
W3C PDF17: page labels and page positionspdf-lib: PDFDocument.copyPages
Custom groups replace the file’s ordinary Pages selection
In custom mode, Page groups defines every output and replaces the ordinary Pages field. Even malformed text left in that ordinary field is not used for this mode. Separate groups with semicolons and entries within each group with commas or ascending ranges. For example, 3,1;2-3;3 creates three ordered parts with 3,1, 2,3 and 3.
Repeated pages can appear within a part or across different parts. Each exported occurrence counts toward the combined 300-page limit; the same source page is not counted only once. Page groups accepts at most 5000 UTF-16 code units and cannot contain an empty group. A trailing semicolon such as 1-2; is an error, not an empty extra PDF.
Worked examples show actual original-page order
The examples below use a six-page source unless the row specifies three pages. Bracketed sequences list original physical page positions in each generated part, not newly printed page numbers. Chunking follows the selected sequence, while parity follows original page numbers.
A single populated group produces one output regardless of the selected mode. Odd-only or even-only selections do not create placeholder files for the absent group. The custom-mode row with an invalid ordinary field illustrates replacement of that field rather than acceptance of invalid group grammar.
| Mode and source selection | Additional settings | Output groups of original pages |
|---|---|---|
One per page: 3,1 | Six-page source | [3], [1] |
Chunks: 6,2-4,1 | N = 2 | [6,2], [3,4], [1] |
Custom: ordinary field invalid | Groups 3,1;2-3;3 | [3,1], [2,3], [3] |
Odd/even: 6,5,3,2,1 | Six-page source | [5,3,1], [6,2] |
Odd/even: 2,4 | Even-only selection | [2,4] |
Odd/even: 5,1,5 | Odd-only with repetition | [5,1,5] |
| One per page: blank | Three-page source | [1], [2], [3] |
Custom: ordinary field 999 | Groups 1-3;4,6;5 | [1,2,3], [4,6], [5] |
The full source and exported copies have separate limits
Choose one nonempty source PDF within the byte ceiling and containing 1–300 total source pages. Selecting a few pages from a document with more than 300 source pages does not bypass the reader limit. Across every output part, exported page occurrences must total 1–300, including all repetitions.
Input and output MB are decimal, not MiB. The encoded byte budget sums generated PDFs before ZIP packaging. A source within the ceiling does not guarantee successful decoding, memory use or output size; repeating pages or resources across parts can increase encoded bytes. A failed group or exceeded output budget prevents a partial completed result list.
| Boundary | Actual limit |
|---|---|
| Source files | 1 nonempty PDF |
| Entire source file | 50000000 bytes, inclusive |
| Entire source page count | 300 pages |
| Exported page occurrences across all parts | 300, including repetitions |
| Combined encoded PDF outputs | 100000000 bytes, inclusive |
| Custom Page groups text | 5000 UTF-16 code units |
Header, parsing and signature-field checks are distinct
The reader checks for a PDF header hint within the first 1024 bytes, then uses the PDF library with invalid-object handling and source-page limits. A .pdf filename does not make arbitrary bytes a PDF. Encrypted or damaged inputs can be refused, and this splitter has no password or unlock control.
Recognized signature fields or markers also cause refusal. The check includes detected /ByteRange, /FT /Sig or /Type /Sig dictionaries and can reject an unused or empty signature field. It is not cryptographic validation of signer identity or a complete PDF-conformance audit. Keep the original and use an appropriate supported source rather than expecting rewriting to preserve a signature.
Copying retains supported page geometry without a rotation editor
Selected pages are copied into new PDF documents rather than rendered into images. Supported page resources, media and crop bounds, existing rotation and the source’s declared physical page scale follow the copy path. The worked checks use different page widths, a nonzero crop box, existing rotation and a declared physical unit to distinguish copying from a generic screenshot.
Split PDF supplies no additional page rotation, resizing or crop operation. Changing order with range text does not rotate content or rewrite printed numbers. When you need thumbnail arrangement, rotation or other transformations, choose the relevant separate tool and review its limits; preservation of supported page geometry is not a universal promise about every PDF feature.
pdf-lib: PDFDocument.copyPagespdf-lib 1.17.1: pinned PDF object copierpdf-lib 1.17.1: page attributes and inheritance
Page content and document-level features need different expectations
Copying keeps supported existing page text, vector drawing and image content in PDF form without rasterizing it. A scanned page remains scanned and does not acquire selectable text or OCR results. Ordinary PDF serialization creates new bytes and may write output metadata; there is no original-byte shortcut or document-metadata preservation control.
A new document assembled from pages is different from retaining an entire source document’s root structure. Bookmarks, interactive forms, accessibility tags, attachments and cross-page links are not guaranteed to survive. Check the result in the reader and workflow where those features are needed; copying page content alone is not an accessibility or full-fidelity certificate.
| Requirement | What to expect |
|---|---|
| Existing supported page text/vector/image content | Copied as page content without rasterizing |
| Selectable text from a scanned page | No new OCR text is created |
| Bookmarks, forms, tags, attachments and cross-page links | Document-level preservation is not guaranteed |
| Exact original bytes or original metadata | New PDF serialization; no original-byte or metadata-preservation mode |
pdf-lib: PDFDocument.copyPagespdf-lib 1.17.1: pinned PDF object copierpdf-lib 1.17.1: page attributes and inheritance
Preview detail changes display, not copied output
The file row’s Preview button opens a preview of the whole source, independently of the chosen split selection. Preview page uses that document’s physical positions. Each generated part has its own Show PDF preview control and its own page count, so output preview page 1 may be original source page 3.
Preview detail offers longest-side limits of 800, 1600 or 2400 pixels for display. These options do not change the copied PDF’s page geometry, fonts, image resolution or output quality. The visible preview is a raster display of a PDF; the downloaded split document is produced by the separate page-copy path.
PDF.js: page viewport and canvas rendering examplespdf-lib: PDFDocument.copyPages
A preview limitation is not always a copying failure
Preview and image rendering have separate resource guards. Embedded images above 24 megapixels or 12000 pixels on either side can prevent rendering, while supported page copying remains available. Rendered preview pages are also bounded to 12 megapixels and 8192 pixels on either side; these are this app’s limits, not universal PDF.js limits.
Preview reading uses a 50 MB source ceiling even for a generated part, while the combined copied-output ceiling is 100 MB. A large generated PDF can therefore download successfully but fail its preview. Header, encryption, signature-field or source-page-count rejection is different: those acceptance failures block the copy path itself. A missing preview alone does not prove the output is corrupt or that copying cannot work.
Download individual parts or a ZIP of actual PDF buffers
If the split produces one populated group, its filename is pages.pdf. Multiple groups produce indexed names such as split-001.pdf, split-002.pdf and later numbered parts in group order. These names depend on the number of outputs rather than the original filename or the name of the selected mode.
More than one result enables Download all as ZIP, named toolmellow-results.zip. It packages the generated PDF buffers with stored entries; ZIP packaging is not automatic PDF compression. Separating pages does not promise a smaller combined size or an exact KB/MB target. Parts can duplicate resources and encoding overhead, so compare actual sizes before using a strict upload threshold.
Changes invalidate results while Clear retains split settings
Editing page selections, split mode, chunk size or custom groups clears existing results and requires another Split PDF action. Invalid ranges or groups can be corrected before retrying. The whole copy job has a 60-second timeout; Cancel terminates processing while retaining the selected source, without exposing a partial completed output list.
Clear files removes the selected source, its ordinary Pages field, results and source preview. It retains split mode, chunk size and custom-group text, so a new source inherits those choices and old groups can be out of range for a smaller replacement. Clearing the workspace does not delete source files or downloads on the device. Keep originals separately.
MDN: Web Worker messaging and lifecycleMDN: stopping a worker with terminate()
Local copying and preview rendering use different browser components
You choose files on the device; inspection and page copying run as local tool-worker jobs. PDF previews use PDF.js parsing support with main-thread DOM-canvas rendering. This is not a claim that every preview operation runs in the same tool worker. Ordinary hosting, asset and analytics requests remain separate from processing the PDF bytes, without an anonymity or zero-logging guarantee.
The splitter has no server PDF-upload endpoint, remote-URL or cloud-drive importer, saved project, text editor or automatic OCR step. Each result deliberately offers Compress this PDF, Review text for Word and Export page images. Those buttons transfer a generated file through one-time same-tab memory to another tool with its own limits. Leaving or refreshing discards this workspace.
MDN: local files and the File APIMDN: Web Worker messaging and lifecyclePDF.js: page viewport and canvas rendering examples
Embed Split PDF with the visible ToolMellow backlink
Use Add this tool to your website on the localized tool page to copy or download the supplied HTML snippet. Retain the interactive iframe, sizing behavior and visible ToolMellow backlink with its existing link attributes. Select the intended language and give visitors enough space for selection fields, results and previews.
Visitors choose their own supported local PDF and use the same single-source, range, grouping and budget rules. The snippet is not an API for automatically splitting files hosted by your website or bypassing source limits. Test its workflow in the intended browsers and layout, and keep the credit link visible.
Common questions
How do I split a PDF into separate files?
Choose one supported local PDF, wait for its page count, set the split mode and page selection or custom groups, then press Split PDF. Inspect the output counts and generated parts before downloading each PDF or the ZIP.
How do custom page groups work?
Custom semicolon groups replace the ordinary file Pages field. For example, 3,1;2-3;3 creates three parts containing 3,1, 2,3 and 3. Each group must be nonempty, and every repeated page occurrence counts toward the combined limit.
Can I repeat pages or change their order?
Yes. Comma entries and ascending ranges retain the entered sequence and allow repeated pages. Chunk mode groups that sequence, while parity mode preserves it within each original-number partition. There is no thumbnail arrangement grid in this splitter.
Are odd and even pages based on original numbers?
Yes. Parity uses original physical page numbers, not a page’s position in your selection or its printed label. If both groups contain pages, odd pages form the first part and even pages the second. Empty groups are omitted.
Why did even-only splitting create one file?
Only populated groups become outputs. Selecting 2,4 in odd/even mode produces one even-only PDF named pages.pdf, with no placeholder odd file or second part.
Can I select pages from a PDF with more than 300 pages?
No. The source reader requires the complete source to contain 1–300 pages. Selecting a smaller subset does not bypass that source limit. Separately, exported occurrences across all parts are limited to 300 including repetitions.
Does splitting reduce PDF file size?
There is no guaranteed size reduction, exact byte target or automatic compression search. Each part is newly encoded, and repeated resources or packaging can increase combined bytes. Compare actual results and use the separate compression tool when appropriate.
Does the PDF become an image?
The output copy path does not rasterize supported page content. Its preview is a separate raster display. A scanned page remains an image inside a PDF and does not gain OCR text merely by splitting.
Will forms, bookmarks and accessibility tags survive?
Preservation of document-level forms, bookmarks, accessibility tags, attachments and cross-page links is not guaranteed. Review the generated parts in the PDF reader and workflow where those features matter.
Can I split encrypted or signature-field PDFs?
Encrypted or damaged documents can be refused. Detected signature fields or markers also cause refusal, including an unused field. There is no password/unlock control or cryptographic signature-validation service here.
Does preview detail change output quality?
No. The 800, 1600 and 2400 pixel options control the displayed preview. They do not change copied PDF page geometry or image resolution; preview-resource limits are separate from the copy budget.
Can I embed the PDF splitter?
Yes. Copy or download the supplied HTML iframe snippet from Add this tool to your website, retaining the visible ToolMellow backlink and sizing behavior. Visitors choose local supported files under the same rules; it is not an automatic hosted-file API.
Why does a Roman or printed page label fail in the range field?
Range input uses ASCII decimal physical positions from 1 through the source count. Printed numbers and PDF page labels can differ from those positions. Use the source preview to identify the intended physical page.
Does Clear files reset my split mode and groups?
It removes files, ordinary per-file page selections, results and preview, but keeps split mode, chunk size and custom-group text. New files inherit those settings; adjust retained groups when the replacement PDF has different pages.
Sources and further reading
- pdf-lib: PDFDocument.copyPages
- pdf-lib 1.17.1: pinned PDF object copier
- pdf-lib 1.17.1: page attributes and inheritance
- PDF.js: page viewport and canvas rendering examples
- W3C PDF17: page labels and page positions
- MDN: local files and the File API
- MDN: Web Worker messaging and lifecycle
- MDN: stopping a worker with terminate()