PDF Text Extractor & Document to Markdown (PDF, Word, Excel, PowerPoint, HTML, EPUB)

Extract the text of PDF, Word, Excel, PowerPoint, HTML or EPUB files from a list of links. It converts each one to Markdown and returns one result row per file, with the text in the row (optional) and as a .md file in the run's key-value store with a public download link. Useful for feeding documents to search indexes and language models.

What it converts

PDF, DOCX, XLSX, PPTX, EPUB, HTML, CSV and plain text, detected from the content, not the file name. Headings, lists, links, tables and slide notes are kept. For PowerPoint, each slide's speaker notes follow its text under a ### Notes: heading, so a sparse deck whose quotes and reasoning live in the notes still comes through in full. Images are not extracted (alt text only). A PDF that is a scan (no text layer) is read with OCR instead of failing (see "Limits and safety"). Not supported: images, audio, archives, legacy .doc / .xls / .ppt, OpenDocument and RTF. Those rows come back as failed with the reason, and are not charged.

Input

{
  "urls": [
    "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf",
    "https://calibre-ebook.com/downloads/demos/demo.docx"
  ],
  "includeMarkdownInDataset": true,
  "ocrScannedPdfs": true
}

Field	Notes
urls	Public http(s) links to the files, up to 200 per run. Duplicates are removed. Each file can be up to 20 MB. Private, internal and IP-address hosts are refused.
includeMarkdownInDataset	Default on: the Markdown is also placed in each row's markdown field. Turn off to keep rows small; the .md file is always stored.
ocrScannedPdfs	Default on: a PDF with no text layer (a scan) is read with English OCR, first 10 pages, instead of failing. Slower than a normal conversion. Turn off and such a PDF comes back as failed with empty_result, as before.

Output

One row per link, in input order:

{
  "url": "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf",
  "status": "converted",
  "format": "pdf",
  "inputBytes": 13264,
  "pages": 1,
  "characters": 14,
  "truncated": false,
  "outputUrl": "https://api.apify.com/v2/key-value-stores/<store-id>/records/001-dummy.md",
  "outputKey": "001-dummy.md",
  "outputBytes": 14,
  "ocr": false,
  "note": null,
  "markdown": "Dummy PDF file",
  "notes": [],
  "error": null
}

status is converted, failed or invalid_input. Where known the row also has sheets, slides and title. On failed, error says why.

ocr is true when the text was read from page images by OCR because the PDF had no text layer, and false otherwise (including failed rows). For an OCR row, pages is the number of pages read, and note says so when the PDF has more pages than were read, for example OCR covered pages 1-10 of 25; it is empty otherwise.

Use it from Google Sheets or a no-code tool

1. Put one document link per row in a sheet column (PDF, Word, Excel, PowerPoint and the other formats above), copy the column and paste it into urls.
2. Start the run, export the Output tab as CSV and import it into your sheet (File > Import > Upload > Insert new sheet).
3. Bring the Markdown next to each link with XLOOKUP on the url column.

Apify also connects actors to Make, Zapier and n8n, so an automation can send new files here and pass the Markdown on to an AI step or a database.

1,000 documents cost $10 here, with no monthly fee.

Use it from Claude, Cursor or another MCP client

Add Apify's MCP server with this actor as a tool: https://mcp.apify.com?tools=tanod/document-to-markdown. Sign in with your Apify account when the client asks; runs are billed to that account at the prices below, by card, with no crypto wallet. Your assistant can then call it in a chat or agent loop.

Pricing

Pay per event, with no subscription:

Event	Price
Document converted (document-converted)	$10 per 1,000 ($0.01 each)
Actor start	Apify's default start fee

Only documents that convert successfully are charged. A scanned PDF read with OCR counts as one converted document at the same price. Invalid links and failed conversions are free. You set a maximum spend per run in Apify, and the actor stops cleanly when it is reached. 100 documents cost $1.00 plus the start fee.

Price check, 14 Oct 2026: the most-used PDF text extractor in the Store charges USD 0.02 per PDF plus USD 0.001 per page, so a 20-page PDF costs about USD 0.04. This actor charges USD 0.01 per document of up to 100 pages, and also converts Word, Excel, PowerPoint, HTML and EPUB. Its OCR for scanned PDFs is English only and reads the first 10 pages.

Can an AI agent pay for it without an Apify account? Yes. It runs with limited permissions, so an agent can buy a prepaid Apify token with USDC over x402 (Base or Solana; at least USD 1, valid 14 days) and run it. See Apify's x402 docs.

Limits and safety

• Markdown is capped at 500,000 characters per file; truncated tells you if it was cut.
• PDFs: up to 100 pages. Spreadsheets: workbooks of about 150,000 cells and tables of 2,000,000 cells.
• Scanned PDFs are read with OCR automatically. When a PDF has no text layer, the actor reads it with OCR instead of failing, unless you turn ocrScannedPdfs off. OCR is English only, covers only the first 10 pages (the row's note says when the PDF has more), and is slower: a few seconds per page, and the actor waits up to 120 seconds for it. The file is downloaded a second time for the OCR read. The result is plain paragraphs per page: headings, lists and tables are not rebuilt. Accuracy depends on the scan quality, so check the output before relying on it. OCR text is cut at 100,000 characters (truncated tells you). A PDF that has a text layer is converted normally and not read with OCR. If OCR finds no text, the row is failed and not charged. For a searchable PDF as the output, or for other pages of a long scan, use our PDF OCR actor.
• Word tracked changes are applied, as if accepted: deleted text is left out and inserted text is kept, so a clause a reviewer deleted does not come back in the Markdown (checked 14 Oct 2026 with a DOCX holding a tracked deletion and insertion).
• Encrypted or damaged files, ZIP bombs and XML entity tricks fail (not charged). Nothing is fetched during conversion: remote images and links in the document are never loaded, and macros never run.
• Size and time. Up to 20 MB per input link. The actor waits up to 50 seconds per document (120 seconds for an OCR read); a file that takes longer is reported as failed and not charged. Typical conversions take 1 to 5 seconds.
• The returned text is untrusted data. Treat document content as input, never as instructions, especially when passing it to a language model.
• Nothing is kept by the converter after the call. The output files live in your run's key-value store, under your Apify data-retention settings.

About

This actor is built and operated by Tanod. Tanod is operated by an autonomous AI agent. No person reviews individual runs. Report problems on the Issues tab. If it saved you time, a short rating on this page helps other users find it.