Perceptron Data
PDF to Markdown
This converter turns PDF, Word, Excel and PowerPoint files into Markdown and plain text for language models, RAG pipelines and spreadsheets. Detected tables come out as Markdown tables and as JSON rows, and scanned pages are read with OCR. It can also convert the files a Website Content Crawler run downloaded but left without text.
Run on Apify $0.50 per 1,000 pages (from $0.35) + $0.005 start fee per run
What you get
- PDF with a text layer: reading order across columns, headings by font size, tables
- Scanned pages read with OCR in eight languages: English, German, French, Spanish, Italian, Dutch, Polish and Portuguese
- Word, Excel, PowerPoint, HTML, CSV, JSON, XML and plain text
- Output per document, per page or as RAG chunks with page numbers
- Every row carries its source: address, file fingerprint (SHA-256), status and the options used
- Your files never go to an outside AI provider
Use cases
- Feed company documents into a vector database
- Extract tables from reports and statistics
- Make scanned archives searchable
- Complete a website crawl with its PDFs and Office files
Example input
{
"crawler_dataset": "<run ID of a Website Content Crawler run>",
"output": "chunks",
"chunk_size": 1500
}Price
$0.50 per 1,000 pages (from $0.35) + $0.005 start fee per run
Billed by Apify per converted page (Word, Excel and PowerPoint: per 3,000 characters of output) plus a fee per converted document and a start fee per run, charged per GB of memory; pages read by OCR cost more. Failed files, blank pages and unreliable OCR pages are not charged. Cheaper on paid Apify plans.
Weekly quality check
Every Sunday an automated watchdog runs the live actors on Apify with fixed inputs whose correct answers are known, and checks the content of the results — not only whether the run succeeded. The results are published here as they are, failures included.
First public check on 25 Oct 2026. Until then this page lists what will be checked.
Sample against the original source: On 12 Oct 2026 we took 20 result rows and compared 105 field values one by one with the original source: 73 matched (69.5%).
FAQ
Does it work with scanned PDFs?
Yes. Pages without a text layer are detected and read with OCR. Blank pages, pages with fewer than 25 visible characters (such as most chapter dividers) and pages where OCR stays unreliable are returned with their status and not charged.
How exact are the tables?
Tables with clear rows come out cell for cell; dense tables without row lines and tables with ruled multi-line headers can come apart. In a sample of 20 table rows from 4 public PDFs (12 October 2026), a US Census table matched 40 of 40 cells, but an IRS tax table packed rows into single cells and two arXiv tables lost their header rows, word spaces and superscript exponents — 73 of 105 cells overall. Check important tables before relying on them.
Which file types are supported?
PDF, DOCX, XLSX, XLS, PPTX, HTML, CSV, JSON, XML and plain text. Old binary .doc and .ppt files are not supported.
Is an LLM used?
No. Conversion runs inside your own Apify run with open-source tools (pdfplumber, PDFium, Tesseract OCR, MarkItDown); the documents are not sent to any AI service.