Repository navigation
Help
PDF Extractor detects and crops images, tables, text, titles, formulas, and other page elements. This guide covers the web app. For local, offline document processing, use the Desktop application guide.
- The web app accepts PDF files. Export other file types to PDF first.
- A free account is required. Create an account or log in.
- Plan allowances, retention, and prices can change. Your account and checkout screens show the current terms.
- Open Extract after signing in.
- Drag a PDF onto the upload area or use the file picker.
- Choose one or more element categories. Image is selected by default.
- Open Advanced Settings only if you need a page range, different output resolution, OCR, or another specialized option.
- Select Extract Elements.
- Review the Results page, then download individual elements or a selected group.
For a password-protected PDF, enter its password in Advanced Settings before starting. The password is used to open the document for the requested extraction.
The web app exposes these category names:
Image, Table, Text, Title, Section Title, Caption, Footnote, Expression, Entry, Header, and Footer.
Select only what you need. A focused category set can reduce noise and processing time. If an expected element is missing, first confirm that its category was selected.
| Option | What it controls | Useful starting point |
|---|---|---|
| Page range | Pages to process, such as 1-10 or 1,2,5
|
Use a short range while testing settings |
| Image resolution | Rendering and output DPI | 200 DPI is the default; raise it for small or low-resolution content |
| Output format | Original/default, JPG, PNG, or TIFF | Keep the default unless a downstream tool needs a specific format |
| Minimum dimensions | Small detections to ignore | Lower the values if small figures or marks are missing |
| Confidence threshold | Minimum detection confidence | Lower it slightly for more results; expect more false positives |
| IoU threshold | Removal of overlapping detections | Keep the default unless overlapping results are a problem |
| Timeout | Maximum processing time | Use a smaller page range before increasing it |
| Store file | Whether results remain available in extraction history | Disable it when you only need the immediate download |
| OCR | Text recognition at element or whole-page level | Enable it for scanned or image-only text |
Use OCR when text in the PDF cannot be selected or copied.
- Standard OCR (Tesseract) is suited to straightforward character recognition. It supports page-level or element-level OCR and text output.
- Advanced OCR is a layout-aware vision model for complex pages, multiple columns, irregular tables, and handwriting. It runs in page mode and can produce text, Markdown, or structured JSON.
-
Structured table extraction transcribes detected Table regions to Markdown, HTML, or CSV when the option is available. Include
Tablein the selected categories.
OCR takes longer than detection alone. Advanced OCR and structured table extraction depend on the current plan and service availability.
The Results page lets you filter by category, switch between PDF pages, preview crops and recognized text, and download individual or selected elements.
Feature callout: Markdown previews appear as cards alongside the detected crops. In advanced Markdown output, figure markers can bind to extracted Image cards from the same page.
Feature callout: Table results remain separate cards, which makes them easy to filter, inspect, and download without exporting every other element type.
If Store File was enabled, the extraction remains available in history for the retention period shown by your plan. If storage was disabled, download the immediate result before leaving the flow.
Business accounts can upload several PDFs together. Each PDF is processed as a separate extraction and results remain grouped by source file. If a large batch fails, retry fewer files or shorter page ranges.
Recent jobs appear on the Extract page. You can filter by date, reopen stored results, delete jobs, and show expired or deleted entries. An expired entry can remain in the list even though its stored files are no longer downloadable.
Plans differ by monthly page allowance, storage retention, OCR access, and batch support. Use the Subscription page for current plan details, billing management, plan changes, and cancellation.
Report a reproducible bug or request an enhancement in the public issue tracker. Search existing issues first.
Issues are public. Do not include PDFs, document content, personal information, passwords, API tokens, license keys, or billing details. Use private support for account, billing, licensing, privacy, security, and document-specific questions.
Last reviewed: August 21, 2026.