ocr-and-documents

To extract text from a scanned PDF using OCR, the workflow applies marker-pdf to perform optical character recognition on scanned documents and images. It processes local files or remote URLs and outputs the extracted text as Markdown or structured JSON.

The best way to extract text from a PDF URL is by prioritizing the web_extract tool for remote links. This pulls online document content directly before falling back to local extraction methods like pymupdf for text-based PDFs.

Yes, pymupdf works efficiently for extracting text from text-based PDFs. When a document has embedded text rather than scanned images, pymupdf parses the content directly and can output the results as plain text or Markdown.

Yes, you can get Markdown output when extracting document text. The extraction workflow generates plain text, Markdown, and structured results like JSON, enabling seamless integration into downstream data and analytics pipelines.

Besides PDF formats, the document text extraction workflow supports scanned images via OCR and DOCX files via python-docx. Remote URLs are also supported through web_extract before applying local fallback extractors.

You should use marker-pdf instead of pymupdf when dealing with scanned documents or image-heavy files that require OCR. pymupdf is reserved for text-based PDFs where text is already embedded and can be parsed directly without OCR.