MarkPrep
scanned-pdf → markdown

Scanned PDF to Markdown

OCR scanned PDFs into Markdown in your browser, with no page limit and no upload.

Drop files, paste, or click to browse
PDF, Word, Excel, PowerPoint, HTML, CSV, JSON, EPUB, images and more. Nothing is uploaded.
Live count of requests to other websites during conversion: 0To see for yourself, press F12, open the Network tab and convert a file.

MarkPrep converts a scanned PDF to Markdown with OCR (optical character recognition) for free, in your browser, so the scan is never uploaded. Many 'scanned' PDFs already carry an invisible text layer, so MarkPrep checks each page first and uses that text when it finds it. When a page is only an image, the Tesseract OCR engine reads it in English. Each such page is first rendered at a larger size and turned black and white for OCR. OCR is on by default, and there is no page limit.

How to convert Scanned PDF to Markdown

  1. 1Drop your scanned PDF onto the box above, or click the box to pick a file.
  2. 2MarkPrep converts it to Markdown in your browser and shows an estimated token count for the AI model you pick.
  3. 3Check the result on the Preview tab. The Chunks tab holds the chunk preview, which shows where the text splits into chunks of the size you set. The Verify tab flags tables with uneven rows, pages where OCR was unsure and text that didn't convert cleanly.
  4. 4Copy the Markdown, or download it as a .md file. You can also export the chunks as JSONL, a file with one chunk per line. A RAG pipeline loads that file so an AI model can look up passages in your documents before it answers.

Format notes & gotchas

  • To skip OCR, switch off 'OCR scanned pages (local)' under Output. Image-only pages then come out empty, and the Verify tab lists each one.
  • Poor scans and small print give less accurate text than a digital PDF, and handwriting isn't read. The Verify tab flags pages where OCR was unsure.
  • OCR on a long scan takes a lot of processing, so a computer is faster than a phone.

Related converters

Frequently asked questions

How is this different from hosted OCR?+

Hosted OCR is a service that sends each page to its own server to be read. MarkPrep's OCR runs in your browser instead, so the scan stays on your device.

Is my file uploaded anywhere?+

No. Conversion runs in your browser tab, and there is no upload step. Two optional features do send something, and each says so where you use it. URL to Markdown asks MarkPrep's server to fetch the public web page you name. AI enhance is a button that asks an AI model to tidy the structure of the Markdown. It sends the converted text to Anthropic with your own Anthropic API key, a personal access key you create on Anthropic's site and add in Settings.

How much does it cost?+

Conversion and the AI-Ready Pack are free and unlimited, because they run on your device. The AI-Ready Pack is the token count, the chunk preview and the JSONL export. MarkPrep Pro costs £19 once and adds tools that run outside the browser. Its command-line converter turns files into Markdown from a terminal, and it can watch a folder and convert each file you put in it. Its local MCP server lets AI apps such as Claude Code convert files on your computer. MCP is the standard those apps use to call outside tools.

Are the token counts accurate?+

Every count is an estimate, marked "approx" next to the number. OpenAI hasn't published the encoding its GPT-6 models use, so MarkPrep counts them with o200k_base, the encoding of OpenAI's earlier models. Anthropic and Google offer no tokenizer that runs in a browser. For Claude, MarkPrep counts about 2.5 characters as one token, and for Gemini about 4, the averages the two companies give. Real counts vary with the text, so leave some room below the model's limit.

MarkPrep Pro, £19 once

Pro lets you convert files to Markdown from a terminal, or from AI apps such as Claude Code, on your own computer.

See pricing