Scanned PDF OCR to Text & Markdown avatar

Scanned PDF OCR to Text & Markdown

Pricing

from $5.60 / 1,000 ocr pages

Go to Apify Store
Scanned PDF OCR to Text & Markdown

Scanned PDF OCR to Text & Markdown

OCR scanned and image-only PDFs into plain text and Markdown with per-page confidence. Pages that already have a text layer are read directly. Blank, unreadable and failed pages are free.

Pricing

from $5.60 / 1,000 ocr pages

Rating

0.0

(0)

Developer

SteadyScrape

SteadyScrape

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

4 days ago

Last modified

Features

Limited permissions

Share

Turn scanned and image-only PDFs into clean plain text and Markdown. Each page is rendered and read with an OCR engine that runs inside the Actor, so your documents are not sent to a third-party AI service. Pages that already contain a real text layer are read directly (faster, and charged at a lower rate).

Who it's for: Teams that need text from scans, filings, invoices and archives, or a mixed batch that includes both digital and scanned PDFs.

Price: $0.008 per readable OCR page + $0.001 per text-layer page + $0.002 once per run, charged only if at least one page is readable. Blank, low-confidence, failed and timed-out pages are free. A 10-page scan in one run = $0.082.

Pairs with PDF to Text / Markdown: that Actor handles digital PDFs and flags scanned ones as image_only; send those here.

Made by SteadyScrape (steady-scrape).

What you get

One dataset row per PDF:

FieldMeaning
statusok, partial, low_confidence, blank, failed, skipped, encrypted, invalid, empty, too_large, fetch_error
text / markdownText of the readable pages, in reading order. Markdown separates pages with ---.
pageCount, pagesProcessed, pagesChargedPage counts
pagesOcr, pagesTextLayerHow each page was read
meanConfidenceMean OCR word confidence (0–100). A screening signal, not an accuracy guarantee.
pagesPer page: page, method (ocr / text_layer), status, confidence, words, chars, charged, note (and text if includePageText is on)

A run summary is saved as OUTPUT in the key-value store.

What you can do with it

  1. OCR a batch of scans. Point the Actor at direct PDF links (or upload files) and get one row of text or Markdown per file.
  2. Mixed digital + scanned PDFs. In auto mode, pages with a real text layer are read directly at the lower rate; only image-only pages go through OCR.
  3. Triage unreadable pages. Low-confidence and blank pages are labeled and free. Their text is still included by default so you can review them.
  4. Document pipelines and search. Feed the Markdown or plain text into a search index, archive or RAG pipeline, keeping source for citations.
  5. Hand-off from the digital extractor. Use PDF to Text / Markdown first, then send image_only files here.

Pricing

  • ocr-page $0.008 — one OCR page with mean confidence at or above minConfidence (default 60) and enough recognized text. Volume discounts on the Store: Bronze $0.0072, Silver $0.0064, Gold and above $0.0056.
  • text-page $0.001 — one page read from an existing text layer (auto mode).
  • actor-start $0.002 — once per run, only when at least one page is charged.

Free: blank pages, low-confidence pages, pages that fail or time out, and files that cannot be downloaded, opened or decrypted. Control spend with maxPages, maxPagesPerFile, maxFiles and the run's maximum charge. The Actor stops before starting a page it cannot charge.

Worked examples: 10 scanned pages in one run = 10 × $0.008 + $0.002 = $0.082. 3 text-layer pages + 7 OCR pages in one run = 3 × $0.001 + 7 × $0.008 + $0.002 = $0.061.

The default maximum charge per run is $1.00, about 10–20 typical runs. Raise it in the run options if you need a larger job.

Input

{
"pdfUrls": ["https://www.archives.gov/files/research/jfk/releases/docid-32112987.pdf"],
"languages": ["eng"],
"mode": "auto",
"format": "both",
"maxPages": 3
}
  • languages: up to 3 of English, Spanish, French, German, Portuguese, Italian.
  • mode: auto (text layer when present, OCR otherwise) or ocr_all (ignore text layers, e.g. when a PDF's text layer is garbled).
  • dpi: 150–400 (default 300). Very large pages are scaled down automatically.

Under LIMITED_PERMISSIONS the Actor downloads the URLs itself — it does not crawl websites for PDFs.

Accuracy

In our tests on 17 synthetic printed pages in English, Spanish and French, 15 came out with under 1% character errors, including blurred, low-resolution, noisy, skewed and two-column pages. A noisy, blurred, JPEG-compressed page had about 6% character errors. Heavily degraded pages, such as a noisy, skewed invoice photocopy, can come out with many errors; in our tests that page scored below the confidence threshold and was free. Handwriting, tables and layout reconstruction are not supported: tables come out as lines of text.

How it compares

Some image-OCR tools on the Store are cheaper per page for simple pictures. This Actor is built for PDFs: mixed digital and scanned files in one run, unreadable and blank pages free, a per-page confidence score, and documents that stay inside the container instead of being sent to a third-party OCR API. For already-digital PDFs, PDF to Text / Markdown is faster and cheaper.

FAQ

Does it send my PDFs to a third-party OCR API? No. OCR runs inside the Actor container.

Can it read handwriting or reconstruct tables? No. It is for printed text. Tables come out as lines of text.

What about digital PDFs? In auto mode, pages with a real text layer are read directly at $0.001 per page. For a batch that is all digital, use PDF to Text / Markdown instead.

What happens to low-confidence pages? They are free. By default their text is still included so you can review it. You can raise or lower minConfidence (30–90).

Password-protected files? They return encrypted and are free.

Can it find PDFs on a website for me? No. Give direct PDF links, or upload files in the Console.

Integrations

  • Apify API and SDKs: call it from Python, JavaScript or any HTTP client and read results from the run's dataset.
  • Schedules and webhooks: run it on a schedule in Apify and trigger your own endpoint when a run finishes.
  • Make, Zapier, n8n: use Apify's own integrations for these tools to start runs and pass dataset rows on.
  • AI agents (MCP): agents can call it through the Apify MCP server. Runs support a maximum charge so an agent can cap spend.

Limits

  • Max 20 files, 200 pages per run, 100 pages per PDF, 50 MB per file, 2 minutes per page.
  • Printed text only. No handwriting. No table structure.
  • Password-protected PDFs are skipped (free).

More from SteadyScrape

See every SteadyScrape (steady-scrape) Actor on our Apify profile.