Skip to content

PDF tools: split, merge, compress, encrypt, crop, and edit bookmarks

Open Utilities (实用工具) → PDF tools (PDF 工具). This guide uses a synthetic six-page course PDF: a cover, contents page, two handout pages, and two exercise pages, visibly numbered 1–6. Keep the original PDF and check page count, order, text, and bookmarks after downloading.

The tools can be opened from the web and desktop apps. Their processing pages must load successfully. On narrow phone screens, configuration and downloads are usually behind the purple menu button.

ToolInput and purposeOutput
Split (切分)One PDF; divide it or extract pagesIndividual PDFs, ZIP
Merge (合并)Multiple PDFs; concatenate or interleavePDF
Compress (压缩)One or more PDFs; trade file size against qualityPDF, ZIP
Encrypt / Decrypt (加密 / 解密)One PDF; set an opening password or decrypt with its existing passwordPDF
Crop (裁剪)One PDF; adjust visible page areasPDF
Bookmark editing (书签编辑)One PDF; edit bookmark hierarchy and destinationsBookmarked PDF, bookmark text/JSON

To combine images into PDF, use the scan workspace: add images, arrange pages, check previews, and download the image-based PDF. This list has no separate image-to-PDF tool. For Word or Markdown, parse the document and export from the parsing workspace.

Split: extract pages or divide a document

  1. Open Split, upload the sample PDF, and wait for the page count and preview.
  2. Choose a mode and verify it with a small example:
ModeExampleCheck
Page rangesEnter 1-2,3-4,5-6Produces three parts; use actual PDF page positions
Equal divisionChoose the number of parts, or two pages per partThe last part may contain fewer pages
File sizeSet a target sizeUses average page-size estimation; output is not guaranteed to match or stay below the target
BookmarksChoose an existing bookmark levelCorrect source bookmarks are required; check each complete output range and its local bookmark destinations
  1. Adjust the output naming pattern and start splitting.
  2. Preview each result and verify start/end pages, count, and order. For an exact chapter, find its actual start and end pages in the source PDF and enter that range.
  3. Download an individual PDF or Download all as ZIP. Extract and inspect every part.

Do not treat tolerance, output-order, or preserve-bookmarks settings as guarantees that the downloaded result satisfies those conditions. Check important bookmarks in a PDF reader and repair them with the bookmark editor if necessary.

Size-based and equal splitting: results may still exceed the limit

By size estimates ranges from average page size; it does not guarantee that every PDF stays below the target. Equal splitting groups pages and does not guarantee equal byte sizes. Extract the downloaded ZIP and check each PDF, rather than checking only the ZIP or the rounded MB display.

On October 13, 2026, we authored a separate eight-page PDF: seven small text pages and a large simulated scan on file page 3. Its artificial texture deliberately creates uneven page sizes. This is not a customer scan or a representative compression benchmark. The input is 1,941,743 bytes. These operations were actually performed in the current tools:

SettingsOriginal pages in each outputActual bytes per file
By size: 1 MB target, 10% tolerance1–4; 5–81,936,730; 3,521
Equal: by number of parts, 3 parts1–3; 4–6; 7–81,936,020; 2,850; 2,100
Page range: 331,934,664
Standard compression (75% quality) of that extracted pageStill original page 3772,682

Actual size-splitting interface with a 1 MB target and 10% tolerance; the first result is still displayed as 1.85 MB, and the second as 0.00 MB.

Each complete split ZIP covers original pages 1–8 without missing or duplicate pages. Decoded page streams, text, dimensions, rotation, and extracted image bytes match their corresponding source pages. 0.00 MB does not necessarily mean an empty file: the second result contains 3,521 bytes, four pages, and the end marker. Its MB display was rounded. Check file properties and reopen the actual PDF.

Original page 3 alone still exceeds 1 MB. Page splitting cannot remove its image bytes while keeping that full page. We used this follow-up workflow:

  1. Enter 3 in Page range, run Start splitting again, and download the single page.
  2. Open Compress, select that unchanged native download, and retain the original.
  3. Select Standard (75% quality) and All pages. All three structure/title options remained enabled in this check.
  4. Run Start compression and download the result. Check byte size, page count, and the chart; a completion notice is not a content review.

Actual standard-compression result: the 1.85 MB single-page input becomes 754.57 KB and remains one page; its exact size is 772,682 bytes.

Local renders of the actual extracted and compressed PDFs. Values 18, 28, 20 and total 66 remain visible; the image was re-encoded and is not lossless.

The compressed PDF still contains one 595 × 842 point page. Extracted title/footer text differs only by the final newline. The 1400 × 1800 image is re-encoded as JPEG and its pixels change. We visually checked 18, 28, 20, total 66, the bars, and labels. The chart text remains inside an image; this did not add selectable text or run OCR. Check your own small text, formulas, and thin lines separately.

Download the eight-page input, two native split ZIPs, extracted page, compressed file, and check notes. The outer bundle was assembled locally; native downloads inside it are unchanged. Equal splitting and page extraction retained the earlier size-1mb filename pattern. Identify operations by the enclosing ZIP, visible FILE PAGE labels, and the byte table instead of inferring settings from filenames.

This check does not cover bookmarks, forms, digital signatures, passwords, or large batches. Tolerance must not be treated as a verified hard limit. If the receiving platform allows only one file, first check whether splitting meets its requirements. If a smaller file and sufficient clarity cannot both be achieved, follow the recipient's requirements for another delivery method.

Reorder chapters: check bookmarks after splitting and merging

On 2026-10-13, we checked the current online tools with a self-authored eight-page handout: two front-matter pages, chapter starts at file pages 3, 5 and 7, and child bookmarks at pages 4, 6 and 8. Bookmarks → Level 1 (L1) with preserved bookmarks produced three complete two-page chapters through actual downloads. The captured tool UI is in Chinese.

  1. Check the split files. They contain original pages 3–4, 5–6 and 7–8. Each chapter's bookmarks target local pages 1 and 2. The individually downloaded part-001.pdf is byte-identical to its member in the native Download all ZIP from this run.
  2. Choose the chapter order. Add part-003.pdf first and part-001.pdf second in the merge tool, then select sequential merging with preserved bookmarks. The four-page result contains chapter 3 before chapter 1.
  3. Reopen the actual download. In the bookmark tool, all four bookmarks were clicked and reached page indicators 1 / 4, 2 / 4, 3 / 4 and 4 / 4. Compare the chapter and body, not only the count.
BookmarkOriginal eight-page fileSplit chapterReordered mergeRetained printed number
Chapter 3: Review7115
Review: Answers8226
Chapter 1: Area3131
Area: Worked example4242

Current online bookmark split of the authored eight-page handout, showing page 1/2 of chapter 1 and three output files.

The first chapter's actual two-page preview retains its original FILE PAGE 03/08 label and printed page 1.

Actual reordered download reopened in the bookmark tool, with chapter 3 and child before chapter 1 and child; selecting chapter 1 reaches 3/4.

File position and printed numbers are different information. The merge preserved page content streams, text and dimensions. It did not rename chapter 3 to chapter 1 or print new footer numbers. Chapter 3 moved from original page 7 to new page 1, while chapter 1 moved from original page 3 to new page 3. A single offset cannot describe both changes.

Download the source, native split ZIP, individual first chapter, reordered PDF and checks. This review bundle was assembled locally; the contained native output bytes are unchanged. No OCR, translation or AI bookmark generation was used. These small text examples do not establish results for damaged, encrypted, signed or large files.

Merge: concatenate files or interleave pages

  1. Open Merge and select the PDFs. The current upload interface states a maximum of 20 files; this run used two. Check the current page's conditions for your own batch sizes and document sizes.
  2. Drag files into reading order. Put the handout before the exercises in the example. Remove incorrect files and add replacements.
  3. Under Merge method (合并方式), choose Sequential merge (顺序合并) to append each complete document. Interleaved merge (交错合并) offers Pages per file (每个文件取页数) for alternating one or several pages at a time. Check each source's order before combining separately scanned fronts and backs.
  4. Enter Output filename (输出文件名), start merging, and wait for a downloadable result.
  5. Download the PDF. Check that its total equals the sum of the source pages, and inspect the beginning, file boundaries, and ending.

After choosing the count, start a new merge and download that result. When a shorter source runs out, remaining pages from other sources are retained. The tool does not automatically add blanks or correct reversed scans. Check Preserve bookmarks (保留书签) against new-file destinations, not just the switch state.

English merge: page groups and bookmark targets

The English PDF merge guide follows the current English merge tool and provides the original PDFs and two actual downloads from October 13, 2026. File A has three portrait pages; file B has two landscape pages.

  1. Select A and B with Select Files, check their order and enable Keep Bookmarks.
  2. Choose Sequential Merge, set the output name, click Start Merge, then Download. The saved order is A1, A2, A3, B1, B2.
  3. Choose Interleaved Merge and set Pages per File to 2. Change the output name and process again. The new download contains A1, A2, B1, B2, A3.
  4. Open each result. Compare its page boundaries and bookmark destinations; changing settings does not update an earlier download.
Existing bookmarkSequential outputTwo-page interleaved output
A - ReportPage 1Page 1
A - Methods, under ReportPage 2Page 2
A - Findings, under ReportPage 3Page 5
B - AppendixPage 4Page 3
B - Table, under AppendixPage 5Page 4

Actual English tool after processing: Interleaved Merge, two pages per file, Keep Bookmarks enabled and Completed status.

Both five-page downloads preserve each source page's text, decoded drawing instructions, page boxes and rotation. The bookmark titles and nesting levels also match; their destinations follow the new positions. The bookmark tree stays grouped by source, even when pages alternate. The printed source labels are unchanged.

Local rendering of the actual five-page interleaved download: A1, A2, B1, B2 and A3, with mixed portrait and landscape pages.

Download the source PDFs, unchanged outputs and verification record. This comparison ZIP and preview image were assembled locally from the actual downloads. The two-file operation required no sign-in or payment. The interface shows a 20-file limit, but this example does not test that maximum, large PDFs, encryption, signatures or forms. Combining pages adds neither OCR nor translation.

Japanese PDFs: English controls and actual merged files

On October 13, 2026, we selected two self-authored Japanese PDFs in the English merge tool. A contains two pages about area and a worked example; B contains one appendix page. We selected Sequential Merge with Keep Bookmarks, downloaded A→B, then moved B before A using its list handle with Space → ArrowUp → Space. After changing the output name, we ran Start Merge again and downloaded B→A.

Source bookmarkDestination in A→BDestination in B→A
第1章:面積の確認12
例題:三角形の面積23
付録:確認問題31

Actual English merge interface with B before A, Sequential Merge and Keep Bookmarks selected, and the completed result ready to download.

Both native files have three pages. Text, decoded page contents, embedded font data, dimensions, and rotation match the corresponding source pages. Rendering both with the same local engine gives identical pixels for each corresponding page. Japanese kanji/kana, both area formulas, and the original ORIGINAL PAGE labels remain.

Local rendering of the unchanged B→A output: appendix B1, main text A1, and worked example A2.

For an additional navigation check, we reopened B→A in the Chinese bookmark tool and clicked all three Japanese bookmarks, reaching 1 / 3, 2 / 3, and 3 / 3. No bookmark edit or new export was made. This does not establish an English bookmark interface; check your own result in the PDF reader you use.

Download the two inputs, two unchanged native outputs, and verification notes. The outer ZIP was assembled locally. The Japanese PDF merge guide explains the workflow in Japanese while explicitly identifying the English tool controls.

These three simple pages use embedded fonts. They do not verify every font, vertical Japanese text, forms, signatures, encrypted files, or damaged PDFs. Merging does not add OCR text or translate content; text that was inside a source image remains an image.

Two-page interleaving: check page order and bookmark order separately

The same day's run used self-authored source A with three pages and source B with two. Interleaved merge → Pages per file 2, with preserved bookmarks, produced A1, A2, B1, B2, A3. All five pages remained, with text, dimensions and rotation matching their corresponding inputs.

The bookmark tree still groups by source: A's destinations are new pages 1, 2, 5 and B's are 3, 4. The tree's page numbers are therefore not ascending. That alone does not mean the targets are wrong; compare each title with its corresponding source page.

Current Chinese merge tool using three-page A and two-page B, interleaving two pages per file with preserved bookmarks and a completed result.

The review ZIP above also contains both A/B inputs and this actual output. The screenshot establishes the chosen settings; page order and bookmark destinations were checked in the downloaded PDF. This does not establish automatic recovery of arbitrary front/back scan order.

Split by explicit page ranges, or arrange files before merging. (workflow diagram)

Split by explicit page ranges, or arrange files before merging. (workflow diagram). Open the image in a new tab to enlarge it.

Compress: compare size and readability

  1. Open the PDF compression tool and select up to 10 PDFs per batch. Files are processed in a queue.
  2. Start with High quality, Standard, or High compression. Custom settings control image quality, target DPI and the resampling threshold. An image below the threshold is not downsampled merely because the target DPI is lower.
  3. Choose All Pages or Custom Range, using actual file positions such as 1-2,5. A custom range exports only those pages, deduplicated and kept in document order. A range outside a file's page count produces an error for that file.
  4. Review Clear title/author information, Fast web view, and Optimize Structure, then click Start Compress.
  5. Check the actual size and output/original page counts. Size reduction for a custom range includes page extraction. Download an individual PDF or use Download All for a ZIP of successful results, then reopen the files and inspect text, formulas and charts.

Choose a compression preset

PresetImage quality settingTarget DPIResampling threshold
High quality85220300 DPI
Standard75150300 DPI
High compression60120200 DPI

These values are processing settings, not percentages of visual fidelity retained. Higher compression usually removes more image detail. Generate each candidate from the original and inspect small print, subscripts, minus signs and thin lines; do not repeatedly compress an already blurry copy.

When the file barely gets smaller

Compression cannot guarantee an arbitrary target such as 500 KB or 1 MB, and the current tool has no target-size input. Text, vector and previously compressed PDFs may offer little room for further reduction. When compression provides no benefit, the tool keeps the version with your page selection and document information settings applied. Fast web view can increase file size. Check the actual downloaded bytes against the receiving platform's limit.

If image compression is unavailable, the result uses only page selection and document structure processing; image quality, DPI and fast web view were not applied. Download any successful results and retry from the original after recovery. Large scan-heavy files depend on device memory; reduce the batch or split the file if processing fails.

Selectable text and document information

Compression is not OCR. A scan without a text layer still needs recognition before its text can be searched or copied. For PDFs that already have selectable text, check the text, numbers and reading order again after downloading. Prefer the clear original for any subsequent OCR.

Clear title/author information processes the original document information dictionary and document-level XMP. The output may still contain the processing program and generation dates. This option does not inspect annotations, attachments or page content and is not content redaction. Keep the original of files with bookmarks, forms or signatures and inspect those features in the intended reader.

Choose compression quality, then compare size and readability. (workflow diagram)

Choose compression quality, then compare size and readability. (workflow diagram). Open the image in a new tab to enlarge it.

Encrypt and decrypt: set or remove a known password

Select one PDF at a time. The tool checks whether it is encrypted.

Set an opening password

  1. Open Encrypt / Decrypt, upload an unencrypted PDF, and choose the encryption tab.
  2. Enter and confirm the opening password. Keep it safely: it is needed to open or decrypt the file later.
  3. Choose permissions shown in the interface, such as printing, modification, copying, annotation, forms, extraction, and assembly.
  4. Click Start encryption (开始加密), then Download file (下载文件) after completion.
  5. Open the downloaded PDF in a reader and verify that the password works.

Decrypt using the existing password

Upload the encrypted PDF, choose the decryption tab, enter its original password, and click Start decryption (开始解密). Download and verify the result. Correct the password and retry if necessary; this tool does not crack unknown passwords. The current flow is for one file, with no batch operation or user-selectable 128/256-bit algorithm step.

Reader software determines how permission restrictions are enforced. An opening password and editing/copying permissions serve different purposes; test the actual receiving application when those restrictions matter.

Two-page Example: Set a Password, Decrypt With It, and Reopen

On 2026-10-12, we selected a previously published, self-authored two-page English PDF in the actual tool. We set the opening password WillowDemo!2026, unchecked Allow Printing and Allow Content Copying, left the other five permissions at their defaults, then used Start Encryption → Download File. This deliberately public password is for the fictional sample only, not a recommendation for real documents.

Actual single-file encryption settings: an opening password and confirmation, print and copy unchecked, with no algorithm or separate owner-password selector

The actual chooser accepts one PDF at a time. There is no batch queue, independent permissions-password field, or algorithm selector. The downloaded encrypted file requires the correct password. Chrome's built-in PDF reader also displayed a password prompt and showed the two pages after the public sample password was entered.

We then selected that encrypted output. The tool marked it Encrypted, selected Decrypt PDF automatically, and disabled encryption. A deliberately incorrect sample password produced the Chinese message 密码错误或无法解析该 PDF (incorrect password or an unreadable PDF). Using the correct existing password for the same file, then Start Decryption → Download File, produced an unlocked copy. The screenshot below records the actual error message.

Actual wrong-password message in the tested Chinese interface, with the public encrypted sample selected

CheckActual result for this sample
Source, encrypted, and decrypted sizes4,976 / 4,548 / 4,717 bytes, retained separately
Opening behaviorThe encrypted file asks for a password; the actual decrypted file opens directly in Chrome
Pages, dimensions, and textTwo A4-sized pages each; extracted text and pixels match under identical rendering settings
Document propertiesThe decrypted file no longer has the original Title and Author fields
Special objectsThis sample has no forms, digital signatures, or attachments; their preservation was not tested

The encrypted output has V=2, R=3, Length=128, corresponding to 128-bit RC4, not AES or a strong confidentiality solution. The print and copy permission bits are disabled, but that does not establish identical enforcement by different readers. The password prompt and successful opening were verified; copy and printing enforcement were not reliably established. See qpdf's explanation of passwords, permissions, and encryption for these distinctions.

Matching page appearance does not establish byte-for-byte, metadata, or special-object preservation. Keep the source, distinguish output filenames, and check the actual reader and document. The three native PDFs and checking notes are available from the public PDF password guide. Forgotten or unknown passwords are outside this verified example.

Crop: adjust visible page bounds

  1. Open Crop, upload a PDF, and select a page in the preview.
  2. Drag a crop frame around the required content. Zoom in to avoid removing page numbers, headers, or questions.
  3. Select Current page (当前页面) or All pages (所有页面). Try the current page first when dimensions, orientation, or margins vary.
  4. Click Start cropping (开始裁剪), inspect the result, then Download crop result (下载裁剪结果).
  5. Open the downloaded PDF and check all affected pages.

Cropping changes the visible page area. Content outside it may remain inside the PDF, so this does not permanently remove sensitive material or provide secure redaction.

Set a PDF password or adjust the visible page area. (workflow diagram)

Set a PDF password or adjust the visible page area. (workflow diagram). Open the image in a new tab to enlarge it.

Bookmark editing: create a clickable contents tree

Bookmarks control navigation in a PDF reader's sidebar; editing them does not change body headings. The host app currently opens the Chinese tool interface, so the Chinese button labels below can help you find controls.

Edit manually and save

  1. Open Bookmark editing and upload the PDF. Inspect the PDF on the left and wait for its bookmark tree on the right. An empty tree can be used as a starting point.
  2. Click Add (添加), enter a title and actual PDF page number. Add “Handout” at page 3 and “Exercises” at page 5 in the example.
  3. Edit existing titles and destinations, adjust hierarchy/order, and optionally apply bold, italic, color, and expanded/collapsed state. Select multiple entries for batch styling or deletion.
  4. Click bookmarks to check their destinations. Printed page numbers may differ from actual PDF positions.
  5. Click Save and export PDF (保存并导出 PDF), wait for Bookmarks saved (书签保存完成), then Download file. Open the result in a PDF reader and test the bookmarks.

Import, export, and shift page numbers

Under More actions (更多操作), choose Import from text (从文本导入). Use one bookmark per line, Tab indentation for hierarchy, and a Tab between title and page number. Follow the dialog's example; use the supported schema when importing bookmark JSON. Test two entries before importing a full contents list.

Batch page offset (批量偏移页码) supports page mapping and direct offsets. If printed page 1 is actual PDF page 3, set that mapping and check the first and last bookmarks before applying it. Bookmarks read directly from the PDF usually need no new offset.

Export as text (导出为文本) lets you copy bookmark text or Export JSON file (导出 JSON 文件). These contain bookmark data. Use Save and export PDF separately to write bookmarks into a PDF.

Optional: AI contents recognition and bookmark translation

  1. Choose AI contents recognition (AI 识别目录) from More actions. Configure your model service's API Key, Base URL, and model name.
  2. Enter the actual PDF page range containing the contents list, then choose Start extracting bookmarks (开始提取书签).
  3. Review titles, hierarchy, and destinations. Map or offset printed page numbers to actual PDF positions, and inspect out-of-range entries.
  4. If needed, choose a target language and Translate bookmarks (翻译书签). Recheck terminology and destinations.
  5. Select Append to existing bookmarks (追加到现有书签) or Replace existing bookmarks (替换现有书签), then Import bookmarks (导入书签). Inspect the tree and save/export the PDF.

AI extraction sends images of the selected contents pages to your configured model service. Translation also uses that service, so both require connectivity and applicable service allowance. Without a configured service, you can still edit manually or import text. Always review recognition and translation, especially the offset between printed and PDF page numbers.

Check bookmark destinations before saving a new PDF. (workflow diagram)

Check bookmark destinations before saving a new PDF. (workflow diagram). Open the image in a new tab to enlarge it.

Downloads and troubleshooting

Downloaded PDFs and ZIPs appear in the browser download list, the default download folder, or the location chosen when saving. Follow the actual prompt in the desktop app. A preview is not a saved file; download results before closing the tool.

ProblemRecovery
PDF cannot be readCheck it opens in a reader and is complete; decrypt with its known password before using other tools
Invalid rangeUse actual PDF positions starting at 1, not printed numbering; try 1-2 and stay within the total count
Bookmark splits differ from the intended chapter rangeCheck source destinations and the chosen level, then compare individual files with ZIP contents; use explicit ranges when the boundaries do not suit the task
Merged pages are out of orderRearrange source files and try sequential merge; test interleaving with a small sample
Compression fails or stops respondingPreserve completed results, reduce the batch or split large PDFs, then reopen and retry
Encryption/decryption failsCheck matching confirmation or the correct original password; retain the original file
Cropping cuts required contentStart from the original with a larger frame, or crop the current page first
AI contents extraction failsCheck model configuration and contents-page range, or create bookmarks manually