Extract PDF Pages Guide: Order, Repetitions, Thumbnails and Rotations
Admin ·
Extract PDF pages creates one new document from the planned pages of a supported local PDF. ToolMellow accepts ordered page ranges and repetitions, with optional thumbnail arrangement, duplication, removal and added quarter-turn rotations. Inspect the plan and result before downloading the generated PDF.
The tool copies supported page content without rasterizing it. It is not a text or embedded-image extractor, a redaction service or a guarantee that every document-level feature survives. Marked cards, the twelve-card display window and preview detail have specific meanings that are separate from which pages export.
Choose one PDF and create one new document
Use Choose PDF files or drop one supported local PDF into the workspace. Wait for its page count, then enter the desired physical page positions in the file’s Pages field, or leave it blank for all source pages. Press Extract pages and inspect the result before downloading it.
Extraction creates one output, not one file per page or several split groups. The workspace accepts one source at a time; clear or remove the current file before replacing it. The thumbnail editor starts hidden and is optional; use Arrange selected pages to open it. The source file is not overwritten by page selection or download.
- Use ranges or removal to define the output plan; a marked checkbox alone does not filter export.
- Check original page positions and added rotations before extracting, especially with repeated pages.
- Review generated document features in the reader and workflow where you will use them.
MDN: local files and the File APIMDN: Web Worker messaging and lifecycle
Range input preserves order and repeated physical pages
The Pages field uses ASCII decimal numbers, commas and ascending hyphen ranges. Blank or whitespace-only text means all source pages. 3,1,3 requests original pages 3,1,3, including the repeated third page. 5,1-3 requests 5,1,2,3. Entries are not sorted or deduplicated.
These are physical one-based positions within the source, not printed page numbers, Roman front matter or PDF page labels. Use the source preview to identify the intended positions. Page 0, descending ranges, out-of-range references, local-script numeral syntax or a missing comma item can be rejected. Selecting a page does not rewrite the text printed on it.
W3C PDF17: labels and physical page positionspdf-lib: PDFDocument.copyPages
Marked cards target bulk edits rather than export inclusion
Arrange selected pages shows the current output plan. Every planned card exports unless removed; checking a card marks it for bulk operations and does not make unmarked cards disappear from the output. Use the range field or remove unwanted cards when you want a smaller page selection.
Mark all targets the whole plan, including cards outside the visible window. Unmark all clears those marks. Rotate marked 90° and Remove marked act on the marked entries, while Reverse pages reverses the entire plan. Marking or unmarking alone does not edit the PDF or automatically re-extract it.
| Control or state | Effect |
|---|---|
| Page range or current plan | Defines all exported occurrences |
| Mark or unmark a card | Chooses a bulk-edit target; does not filter export |
| Mark all | Marks the whole plan, including offscreen cards |
| Rotate marked 90° / Remove marked | Changes only marked planned entries |
| Reverse pages | Reverses the entire planned sequence |
Twelve visible cards are a display window
The editor displays up to 12 planned cards at a time. Previous pages and Next pages move through that window; they do not truncate the exported plan. Card labels distinguish the current output position from the original source page and added rotation. A move button can take a page across a window boundary.
Drag cards or use the earlier/later buttons to change order. Duplicate inserts a new occurrence immediately after the chosen card and copies its current added rotation; later edits target each occurrence independently. Duplication is disabled at 300 planned occurrences. Removing a card changes the plan, while removing all entries disables Extract until a valid plan is restored or selected again.
Added rotation is relative to the source page
Each card’s added rotation starts at zero. Its Rotate button or the marked-page rotation action adds a clockwise 90-degree turn, cycling through 0, 90, 180 and 270. Export combines that added angle with the original page’s rotation. A source at 90° with an added 90° exports at 180°; adding 270° to that same source wraps to 0°.
Supported media/crop bounds, resources and the source’s declared physical page scale follow the copy path. This tool does not crop, resize, deskew at arbitrary angles or render selected pages into screenshots. The source and its printed page numbers remain separate from output sequence and added rotations.
pdf-lib: PDFDocument.copyPagespdf-lib 1.17.1: pinned PDF object copierpdf-lib 1.17.1: page attributes and inheritancepdf-lib: setting page rotation
Worked plans distinguish copies and rotations
These examples name original physical page positions in the output sequence. The source’s original page 1 already has a 90-degree rotation; pages 3 and 4 start at zero. Added angles belong to individual planned occurrences, including duplicates, rather than changing the original source page globally.
A duplicate of a rotated card starts with the same added rotation. Moving it retains that rotation, and editing another occurrence does not change it. These are examples of the supported copy path, not a guarantee that arbitrary document-level forms, annotations or metadata will behave identically in every reader.
| Planned original pages | Added angles in plan order | Exported page rotations |
|---|---|---|
[3,1,3] | [0°,0°,0°] | [0°,90°,0°] |
[3,1,4] | [0°,90°,0°] | [0°,180°,0°] |
[3,1,1,4] | [0°,90°,90°,0°] | [0°,180°,180°,0°] |
[3,1,4,1] | [0°,90°,0°,90°] | [0°,180°,0°,180°] |
[1] | [270°] | [0°] |
Source pages and exported occurrences have separate ceilings
The complete source must be one nonempty PDF within the byte limit and containing 1–300 total pages. Choosing a few pages from a source with more than 300 pages does not bypass the reader limit. The output plan separately accepts 1–300 occurrences, including range repetitions and card duplicates.
The output is one newly encoded PDF with its own byte ceiling. MB is decimal, not MiB. Page extraction is not an exact-size target or a guarantee of smaller bytes; serialization, copied resources and document structure affect the result. A failure prevents a completed result from appearing, and the source ceilings do not guarantee every document below them can be processed.
| Boundary | Actual limit |
|---|---|
| Source files | 1 nonempty PDF |
| Entire source bytes | 50000000, inclusive |
| Entire source page count | 300 |
| Output occurrences, including repeats and duplicates | 300 |
| New encoded PDF bytes | 100000000, inclusive |
Hiding cards preserves the plan; changing ranges resets it
Hide page thumbnails changes visibility while preserving the current plan, marks and results. Showing the editor again does not restore removed pages or forget added rotations. Restore pages from file ranges is a different control: it discards arrangement edits, duplicates added through cards and added rotations, rebuilding the plan from the current range text.
Changing the source or editing its range field also resets the plan, marks, display window and results. It does not keep card edits as a second hidden selection layer. Restoration retains the range text rather than replacing it with all pages, and malformed range text still needs correction. Read the plan again before extracting after a reset.
Copying page content is not whole-document preservation
Supported page text, vector drawing and image content is copied into a new PDF without rasterizing it. Existing raster images remain images; a scanned page does not acquire OCR or newly selectable text merely through extraction. The operation does not offer a text export, embedded-font/image extraction or content editor.
Document-level bookmarks, interactive forms, accessibility tags, attachments and cross-page links are not guaranteed to survive. New PDF serialization can also write output metadata rather than preserve original bytes or every source document property. Page selection is not a redaction or hidden-object sanitization certificate. Review the output features your actual reader and workflow require.
| Requirement | What to expect |
|---|---|
| Supported existing page text/vector/image content | Copied without rendering pages into screenshots |
| Selectable text from a scan | No new OCR text is added |
| Document-level forms/bookmarks/tags/attachments/links | Preservation is not guaranteed |
| Exact original bytes, metadata or sanitized hidden objects | No such identity or sanitization guarantee |
pdf-lib: PDFDocument.copyPagespdf-lib 1.17.1: pinned PDF object copierpdf-lib 1.17.1: page attributes and inheritance
Source checks are separate from page selection
The reader looks for a PDF header hint within the first 1024 bytes, then parses the document with invalid-object handling and the whole-source page limit. A .pdf suffix cannot turn arbitrary bytes into a PDF. Encrypted or damaged files can be refused; this workspace has no password or unlock field.
Detected signature fields or markers also cause refusal, including a recognized empty or unused field. These include /ByteRange, /FT /Sig or /Type /Sig dictionaries. The check is not cryptographic verification of signer identity or a complete PDF-conformance audit. Use a supported source and retain the original rather than expecting page extraction to preserve a signature.
Preview displays are independent of the exported document
The file row’s Preview control browses the whole source independently of the range or arrangement. The result’s Show PDF preview browses the new output document with its own page count; its page 1 can be original source page 3. Arrangement thumbnails are a separate view of planned entries and their added rotations.
Preview detail offers 800, 1600 and 2400 pixel longest-side display limits; card thumbnails use up to 260 pixels. These display settings do not resize the copied PDF or change its output quality. Source/result previews use PDF.js parsing support and main-thread canvas rendering, while the inspection/copy operation runs as a local tool-worker job.
PDF.js: viewport and canvas rendering examplesMDN: Web Worker messaging and lifecycle
Preview failure does not always prevent copying
Large embedded images above 24 megapixels or 12000 pixels on either side can prevent previews. Rendered-page display has separate 12-megapixel/8192-side guards. When preview rendering is unavailable but the source is otherwise supported, the plan’s page labels and copy path can remain usable; thumbnail absence does not necessarily mean extraction is impossible.
Preview reading has a 50 MB file ceiling, even for a generated PDF, while the one-output copy ceiling is 100 MB. A successfully downloaded large output can therefore be unavailable in the preview. Source-header, encryption, detected-signature-field or source-page-count rejection is different and can block extraction itself. These are app policy limits rather than universal PDF.js constraints.
Download extracted.pdf rather than a split-parts ZIP
Successful extraction produces one file named extracted.pdf, containing the planned page occurrences in their output order. Download saves that generated file. Since this tool returns one result, Download all as ZIP is not offered here. Use Split PDF when the task requires independent page/group files rather than one assembled document.
The result offers Compress this PDF, Review text for Word and Export page images, each passing the generated file through one-time same-tab memory to a separate tool under that destination’s limits. Extraction itself is not automatic PDF compression, an exact KB/MB target or a guarantee of smaller bytes. Compare the actual result before using a strict upload threshold.
Cancel keeps the source; Clear resets the extraction plan
Editing ranges or planned order/rotation/removal invalidates previous results and requires another Extract pages action. The whole copy job has a 60-second timeout. Cancel terminates processing while retaining the input; it does not expose a partial completed document or delete files already downloaded.
Clear files removes the source, range text, manual plan, marks, window position, results and source preview. It retains the editor’s shown/hidden arrangement preference; a new source uses that preference but not the old page edits. Original device files are not deleted. Refreshing or leaving discards this transient workspace, so keep originals and downloads separately.
MDN: Web Worker messaging and lifecycleMDN: stopping a worker
Embed the tool with a visible ToolMellow backlink
Use Add this tool to your website on the localized page to copy or download the supplied HTML snippet. Retain the iframe, sizing behavior and visible ToolMellow backlink with its existing link attributes. Choose the intended language and enough frame space for page selection, arrangement controls and results.
Visitors choose their own supported local PDF under the same one-source and page/budget rules. The snippet is not an API for extracting arbitrary files hosted by your website. Ordinary hosting, asset and analytics requests are separate from local PDF-byte processing, without an anonymity or zero-logging guarantee. Test the intended layout and browsers, keeping the credit link visible.
MDN: local files and the File APIMDN: Web Worker messaging and lifecycle
Common questions
How do I extract pages into one PDF?
Choose one supported local PDF, wait for its page count, enter ordered ranges or leave the field blank for all pages, optionally arrange the plan, then press Extract pages. Inspect the generated document and download extracted.pdf.
Can I repeat pages or change their order?
Yes. Comma entries and ascending ranges preserve entered order and repetitions. Card moves, drag, duplication and removal change the output plan. Every planned occurrence counts toward the 300-page output limit.
Does marking a card select it for export?
No. Marking targets bulk rotation or removal. Unmarked planned cards still export. Use the range field or remove unwanted entries to define the exported page sequence.
Does the twelve-card window limit output?
No. Twelve cards are a display window, not the export limit. Previous/Next pages change the visible window; the full valid plan exports, up to 300 occurrences. Mark all also includes offscreen planned cards.
How do added rotations interact with source rotation?
The card angle is added to the original source rotation and normalized to quarter turns. A source at 90° plus 90° exports at 180°; plus 270° wraps to 0°. Duplicating a card copies its current added rotation, while later edits are independent.
What happens when I change ranges or restore the plan?
Changing source/range text or restoring from ranges discards manual order/rotation/card edits and rebuilds the plan. Restore keeps the current range text. Hiding the thumbnail editor only changes visibility and preserves its plan.
Can I choose a few pages from a PDF with more than 300 pages?
No. The complete source reader accepts 1–300 source pages. Selecting a smaller subset cannot bypass that cap. The output plan separately counts repeated and duplicated occurrences toward its own 300-page limit.
Does extraction change the original PDF?
Your input file stays unchanged. Extraction creates a new PDF containing the planned visible page sequence, with new serialization rather than an original-byte shortcut. This is not a redaction or hidden-object sanitization guarantee.
Will forms, bookmarks and accessibility tags survive?
Document-level forms, bookmarks, accessibility tags, attachments and cross-page links are not guaranteed to survive page copying. Review generated features in the PDF reader and workflow where they matter.
Why can preview fail while extraction works?
Preview has separate file/render/image guards. Supported page copying can remain available when thumbnails or source/result previews fail. A generated PDF may also fit the 100 MB output cap but exceed the preview reader’s 50 MB ceiling.
Can I extract text or images with this tool?
This tool extracts whole PDF pages into one new document. It does not export text, fonts or embedded images or perform OCR. Result handoffs deliberately open separate tools with their own capabilities and limits.
Can I embed the PDF page extractor?
Yes. Copy or download the supplied HTML iframe snippet from Add this tool to your website, retaining its visible ToolMellow backlink and sizing behavior. Visitors select their own supported local files; it is not a hosted-file extraction API.
Can this save one file per page or a ZIP of parts?
Extraction produces one extracted.pdf containing the current plan and has no multi-part ZIP action. Choose the separate Split PDF tool when you need independent files per page or group.
Can I supply a password or a PDF with an unused signature field?
This workspace has no password/unlock control. Encrypted or damaged sources and detected signature fields/markers can be refused, including a recognized unused field. The rule is not cryptographic signature verification.
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-lib: setting page rotation
- PDF.js: viewport and canvas rendering examples
- W3C PDF17: labels and physical page positions
- MDN: local files and the File API
- MDN: Web Worker messaging and lifecycle
- MDN: stopping a worker