PDF to Text API — Extract Text, Layout, and Tables

Extract clean text, layout and table data from any PDF, given either as a public URL or base64. Built on pdfplumber and runs entirely inside the Apify worker — no external call to a third-party OCR service. Built for knowledge engineers, legal/financial analysts, and AI agents that ingest PDFs deterministically.

Free trial

A free Apify account gives you about 5 USD of monthly credits, more than enough to extract several documents and validate the output schema before paying.

Use cases

1. RAG ingestion — drop a batch of PDF reports into the Actor and feed the returned text and tables straight into a vector store.
2. Contract clause mining — extract text per page, then run regex over clauses like Termination, Liability, Governing Law.
3. Invoice line-item extraction — tables array contains one entry per detected table; consume as CSV/JSON for downstream accounting pipelines.

Input

{
  "pdf_url": "https://arxiv.org/pdf/1706.03762.pdf"
}

Field	Type	Default	Description
pdf_url	string	""	Public URL of the PDF (http/https). Mutually exclusive with pdf_base64.
pdf_base64	string	""	Base64-encoded PDF body. Use for private documents.
include_tables	boolean	true	Also extract tabular regions via pdfplumber's table finder.

Output (real run)

[
  {
    "page_number": 1,
    "text": "Provided proper attribution is provided, Google hereby grants permission to reproduce the tables and figures in this paper solely for use in journalistic or scholarly works.\nAttention Is All You Need\nAshish Vaswani* Noam Shazeer* Niki Parmar* Jakob Uszkoreit*\nGoogle Brain  Google Brain  Google Research  Google Research",
    "tables_count": 0
  },
  {
    "page_number": 2,
    "text": "1 Introduction\nRecurrent neural networks, long short-term memory and gated recurrent neural networks in particular, have been firmly established as state-of-the-art approaches in sequence modeling and transduction problems.",
    "tables_count": 0
  }
]

Field	Type	Description
page_number	integer	1-based page index.
text	string	Extracted plain text for the page (whitespace normalised).
tables_count	integer	Number of tabular regions detected on the page.

Pricing (pay-per-event)

• No subscription. Free monthly credits apply first.
• 1 result = $0.001 USD (about $1 per 1 000 results).
• Charged per item returned in the dataset; failed runs are not charged.

Source, licence, limits

• Source: any PDF (URL or base64) you supply.
• Engine: pdfplumber, MIT licence.
• Limits: pure-KNN Python implementation; runs locally inside the Actor memory. No OCR — image-only PDFs (e.g. scans) will produce empty text. Encrypted/password-protected PDFs are rejected.

FAQ

Why are some pages empty? Image-only PDFs (scans) need OCR, which is intentionally out of scope for this Actor. Pre-process with an OCR engine or use a different Actor.

Does it handle multi-column layouts? Yes — pdfplumber preserves column order by reading text clusters left-to-right within x_tolerance.

Can it preserve table structure? Tables are summarised as tables_count. The raw table cells are also written when include_tables=true; downstream code reconstructs rows by grouping cells with overlapping y-values.

Is the upload size limited? Yes — Apify's per-run dataset cap is 5 MB per item. Very large PDFs should be split into per-page runs.

What about forms and checkboxes? pdfplumber does not parse AcroForm fields. If you need form data, use a PDF form-aware library such as pypdf.

How do I extract text from a private URL? Pass the file as base64 in pdf_base64. Anything that requires authentication is unsupported by design.

Other Actors from DevOne-Studio

• ted-tenders-api — TED (Tenders Electronic Daily) EU public procurement search.
• pubmed-articles-api — NCBI PubMed biomedical literature with MeSH/DOI/year filters.
• federal-register-api — US federal agency rules, proposed rules, notices, presidential docs.
• sec-company-tickers — Resolve SEC CIK from ticker or company name (10,400+ issuers).
• sec-edgar-fulltext-search-api — Full-text search across 35M+ SEC EDGAR filings and exhibits.
• sec-edgar-xbrl-frames-api — XBRL frames (Revenues, Assets, NetIncomeLoss) for ~6000 US issuers.
• cfpb-consumer-complaints-api — 18M+ US CFPB consumer complaints since 2011.
• nhtsa-safety-ratings-api — NHTSA NCAP 5-star vehicle safety ratings.
• openfda-udi-api — FDA Unique Device Identifier database by device/manufacturer.
• eia-state-electricity-api — US state-level electricity generation and consumption, 1990-2025.

Changelog

• 2026-10-12 — SEO metadata (seoTitle, seoDescription); README rewritten with real output sample, FAQ, cross-links.
• 2026-10-07 — Pricing switched to Apify pay-per-event ($0.001 per dataset item).
• 2026-09-30 — Initial public release on the Apify Store.