PDF Toolkit — extract pages, split, merge, tables, redact, compress

One Apify Actor for the everyday PDF jobs. No Ghostscript, no external service: everything runs inside the actor container, results come back as structured dataset items plus downloadable files in the key-value store.

Operations

operation	What it does	Output
extract_pages	Keep only the pages you select (1-3,7,10-, negative = from the end)	new PDF + page list
split	Split into single pages or every N pages	one PDF per part
merge	Merge several PDFs in the given order	one PDF
rotate	Rotate selected pages by 90/180/270°	new PDF
text	Text per page (fast, no layout reconstruction)	text + char count per page
metadata	Page count, document info, encryption flag	JSON
extract_tables	Tables via pdfplumber, returned as row/column JSON	tables.json
to_images	Render pages to PNG/JPEG at a chosen DPI	image files
watermark	Diagonal text watermark on selected pages	new PDF
redact	Remove terms from the text layer and black out the area	new PDF + hit list
compress	Strip duplicates, deflate streams, re-encode images as JPEG	smaller PDF + savings
fill_form	Fill AcroForm fields by name	new PDF + filled/missing field list
markdown	Convert pages to LLM-ready Markdown: headings, lists, tables, page anchors	document.md
ocr	Recognise text and word boxes on scanned pages — Tesseract runs inside the container, no external OCR service	ocr.json with text, word boxes and confidence per page

Input example

{
  "operation": "extract_pages",
  "pdfUrls": ["https://example.com/invoice.pdf"],
  "pages": "1-2,5"
}

{
  "operation": "redact",
  "pdfUrls": ["https://example.com/contract.pdf"],
  "redactTerms": ["Max Mustermann", "AT61 1904 3002 3457 3201"],
  "pages": ""
}

{
  "operation": "extract_tables",
  "pdfUrls": ["https://example.com/report.pdf"],
  "pages": "2-6"
}

{
  "operation": "markdown",
  "pdfUrls": ["https://example.com/report.pdf"],
  "pages": ""
}

{
  "operation": "ocr",
  "pdfUrls": ["https://example.com/scan.pdf"],
  "pages": "1-3",
  "ocrLanguage": "eng+deu",
  "ocrDpi": 300
}

Output

Every run pushes one item per operation to the dataset:

{
  "operation": "extract_pages",
  "pages": [1, 2, 5],
  "pageCount": 3,
  "file": "out/extracted.pdf"
}

Generated files are also stored in the default key-value store (OUTPUT records plus the binary files), so a follow-up actor or your own script can pick them up directly.

Pricing (pay per event)

Event	When	Price
apify-actor-start	once per run and per GB of memory (2 GB default)	$0.005
page-processed	per page in the result	$0.002
image-rendered	per rendered page image	$0.001
table-extracted	per table found	$0.02
page-redacted	per page that contained a match	$0.01
form-filled	per filled form field	$0.002
page-ocred	per page run through OCR (replaces page-processed for this operation)	$0.01

A 10-page extract therefore costs about $0.03; a 3-table report extraction about $0.08. OCR is the expensive operation: recognition needs roughly a second of CPU per page, so an OCR page is charged at $0.01 and page-processed is not charged on top of it. The start event is charged once per GB of memory, so the 2 GB default costs $0.01. The code charges the remaining events.

Limits and honest notes

• text reads the PDF text layer. Scanned pages have none — use ocr for those instead.
• markdown is a heuristic, not a layout engine: headings come from font size, lists from bullet characters, tables are appended per page as pipe tables, and the reading order follows the position on the page. Multi-column layouts will come out interleaved — check the output before feeding it into a pipeline.
• ocr uses Tesseract with English and German models. It is accurate on clean scans and printed text, weaker on handwriting, unusual fonts and rotated pages. The confidence per word is in the output so you can see where it struggled instead of trusting a single number.
• OCR is CPU work: expect roughly 1–2 seconds per page at 200 DPI on a 2 GB run. For a few hundred pages, raise the memory setting — the apify-actor-start price scales with it.
• Which DPI? ocrDpi defaults to 200, and that is measured, not guessed: on a clean two-page scan, 150 / 200 / 300 DPI produced the same text in 17.6 s / 23.6 s / 31.9 s. Higher resolution does not improve clean input, it only costs runtime. Use 300 when the source is noisy, faxed or a poor photocopy — there the extra pixels do help.
• redact removes the matched text from the content stream and paints a filled box. It is not a compliance-grade redaction suite: hidden metadata, annotations or attached files are not scrubbed.
• compress re-encodes embedded images; scanned PDFs shrink a lot, text PDFs only slightly.
• Encrypted PDFs are rejected with a clear error (no password support yet).
• Max input size is limited by the Apify memory setting of the run.

Reporting a problem

Please use the Issues tab on the Actor page instead of email. The report then sits with the right Actor, together with the input and the run it came from — usually exactly what is missing to reproduce it. Issues are read and answered there.

What happens to your data

Input files and results stay in your run's storage on Apify. Apify keeps the ten most recent runs of an Actor indefinitely; older runs and their default storages are deleted under the platform's retention rules. I do not export or archive your files anywhere else, and the Actor never sends them to a third-party service.

Local development

python -m venv .venv && .venv/bin/pip install -r requirements.txt
.venv/bin/python -m pytest tests -q
echo '{"operation":"extract_pages","pdfUrls":["/path/file.pdf"],"pages":"1-2"}' > /tmp/in.json
.venv/bin/python -m src --input /tmp/in.json --outdir out --push

The CLI path is the same code as the actor run, so what passes locally behaves identically on Apify.