How It Works
pdfs2txt converts PDF documents into clean, structured Markdown in a few straightforward steps. Here's what happens behind the scenes when you upload a file.
You select a PDF file (up to 50 MB) and choose your preferred processing options. The file is uploaded securely over HTTPS to our server.
The converter parses the PDF structure and pulls out text, headings, lists, and tables. Text-based PDFs — the kind you can highlight and copy from — are handled directly, preserving the original document structure as Markdown.
If the PDF contains scanned pages or embedded images with text, the basic text extractor alone won't capture that content. This is where OCR and AI options come in (see below).
The resulting Markdown file is ready for a one-time download. After you download it (or after a short period), both the original PDF and the converted file are deleted from our server.
Understanding the Image Processor Options
Not every PDF is the same. Some are born digital — created from a word processor or typesetting tool — and contain real, selectable text. Others are scanned documents where each page is essentially a photograph. Many PDFs are a mix of both. The Image Processor dropdown on the home page lets you choose how to handle pages and images that don't contain selectable text.
None (default)
When set to None, the converter extracts only the text that is already embedded in the PDF. This is the fastest option and works well for fully digital documents — reports exported from Google Docs, LaTeX papers, spreadsheets saved as PDF, and similar files. Images in the document are skipped.
Best for: text-heavy, digitally created PDFs with no important information locked inside images.
OCR
Selecting OCR (Optical Character Recognition) enables an on-server recognition engine that reads text from scanned pages and images. When a page in the PDF contains little or no selectable text, the converter detects this automatically and runs OCR to recover the content.
OCR works well for straightforward scanned documents — typed letters, printed forms, book pages, and similar material. It processes everything on our server, so you don't need an API key or external account.
Best for: scanned documents, photographed pages, or PDFs where you see text on the page but can't select or copy it.
AI-Powered Processing (Coming Soon)
We are working on adding AI vision models — such as OpenAI GPT-4o and Google Gemini — as optional image processors. Once available, these models will go beyond basic character recognition: they will be able to interpret visual content, describe diagrams, read handwritten notes, and extract structured data from tables that trip up traditional OCR.
This feature is not yet available. When it launches, it will require an API key from the respective provider, and each page that lacks selectable text — plus any embedded images — will be sent to the chosen AI model for analysis. We will update this page with full details once the feature is ready.
Will be best for: PDFs with diagrams, charts, mixed layouts, handwriting, or images that contain important information you need captured in the output.
Page Chunks
The Page chunks checkbox splits the Markdown output by
page, inserting a horizontal rule (---) between each page.
This is useful when you want to keep track of where each page begins and
ends in the converted text — for example, when referencing specific page
numbers from the original document.
Which Option Should I Choose?
- Digital PDF, no images matter — use None. Fast and straightforward.
- Scanned document, typed text — use OCR. No API key needed.
- Complex visuals, diagrams, or handwriting — AI-powered options are coming soon. For now, OCR is the best available choice for recovering text from images.
If you're unsure, start with None. If the output is missing text that you can see in the PDF, try again with OCR.