PDF to Markdown with OCR (RAG-ready) — by Brainiall

Convert any PDF document (native or scanned) into clean, structured Markdown ready for RAG pipelines, LLMs, AI agents, vector databases and knowledge bases — in seconds.

Provide one or more PDF URLs and get back clean, chunk-friendly Markdown preserving headings, paragraphs, lists, tables, and document layout without clutter.

What it does

• Native PDF Parsing — fast text, table and layout extraction from standard PDFs.
• Automatic & Forced OCR — embedded visual OCR engine extracts text seamlessly from scanned documents, forms, and image-based PDFs.
• Page Range Selection — specify exact page ranges (e.g. 1-5, 3-10) to process only what you need and optimize token usage.
• RAG-ready Output — outputs clean heading structures and word counts per document, perfectly formatted for semantic chunking and embedding.
• Batch Processing — process multiple PDF URLs concurrently with built-in retries and error isolation.

Powered by the Brainiall PDF engine (api.brainiall.com) — high-performance document intelligence for production AI workloads.

Who it's for

• RAG & Vector DB Engineers — ingest research papers, whitepapers, contracts, and technical manuals directly into Pinecone, Qdrant, Chroma, or Weaviate.
• AI Agent Developers — give your agents token-efficient, clean Markdown from complex PDF documents instead of messy raw text extractions.
• Data & Compliance Teams — digitize and structure legacy scanned archives into structured text.
• LLM Fine-Tuners — build high-quality Markdown datasets from domain-specific PDF libraries.

Input

{
    "startUrls": [
        { "url": "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf" },
        { "url": "https://example.com/scanned-contract.pdf" }
    ],
    "pageRange": "1-5",
    "useOcr": false
}

Field	Description
startUrls	List of HTTP/HTTPS URLs pointing to PDF documents.
pageRange	Optional range of pages to convert (e.g., "1-5", "1,3,5"). Leave empty for all pages.
useOcr	true to force visual OCR processing. false uses native extraction with auto-OCR fallback for scanned pages.

Output

Each document produces a structured dataset item:

{
    "url": "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf",
    "status": "SUCCESS",
    "markdown": "# Dummy PDF Header\n\nThis is a sample PDF document converted into clean Markdown...",
    "pages": 1,
    "word_count": 42,
    "ocr_used": false,
    "page_range": "1-5",
    "error": null
}

If a document URL is unreachable or invalid, an item with status: "ERROR" and details is pushed to the dataset without crashing the Actor run.

Pricing

You pay per page converted. Pages in documents that fail or cannot be processed are not charged. No monthly lock-ins, no minimums — billing scales exactly with your document volume.

Tips

• For multi-hundred page documents, use pageRange to extract only relevant sections.
• Leave useOcr as false by default — automatic fallback will trigger OCR if minimal text is detected in scanned PDFs.

---

Built and maintained by Brainiall — production AI APIs for speech, documents, vision and the web.