Blazing Fast PDF to Markdown Converter

Convert PDF to Markdown with layout detection — preserving images, tables, formulas, captions, headers, and footnotes. Built with Rust, NCNN, and MuPDF for maximum performance.

Try the free online converter: pdf2md.deepdiy.net

Features

• Layout-aware Markdown — Uses DocLayoutNet YOLO-based detection to understand document structure. Output preserves headings, paragraphs, tables, lists, formulas, captions, and more in proper reading order.
• Images & Assets — Automatically extracts embedded images and saves them alongside the Markdown output.
• Clean Output — No unnecessary line breaks within paragraphs. Produces readable, well-formatted Markdown.
• Self-hostable — Pre-built binaries for macOS, Linux, and Windows. No Docker or external services required.
• Free Web API — No API key needed. Send a PDF and get back Markdown, image links, and a ZIP download.

Performance Comparison

Faster than other PDF to Markdown tools on equivalent hardware.

Runs efficiently on a 1-core 1GB RAM VPS.

DocLayoutNet detection keeps the original layout intact.

No broken inline text — every paragraph stays together.

No sign-up required. Upload and convert instantly.

Pre-built Binaries

Download pre-compiled binaries for 4 platforms from the dist/ directory:

Platform	Binary
macOS (Apple Silicon)	dist/pdf2md-macos-arm64
Linux (x86_64)	dist/pdf2md-x86_64-unknown-linux-gnu
Linux (ARM64)	dist/pdf2md-aarch64-unknown-linux-gnu
Windows (x86_64)	dist/pdf2md-win10-x64.exe

Step 1 — Move files to your working directory

mv dist/pdf2md-<platform> <workdir>/
mv yolo26n-doclaynet_ncnn_model/ <workdir>/

Step 2 — Run conversion

cd <workdir>
./pdf2md-<platform> <input.pdf>

Arguments

Argument	Description
input.pdf	Input PDF file
output.md	Output Markdown file (optional, defaults to stdout)

Extra options

Option	Description
--asset-dir DIR	Directory to export page assets
--detect-dpi N	DPI for layout detection (default: 72)
--asset-dpi N	DPI for asset export (default: 150)
--page N	Process only the specified page
--model-dir PATH	Path to the model directory (default: ./yolo26n-doclaynet_ncnn_model/)

Build from Source

cargo build --release --bin pdf2md

The compiled binary will be at target/release/pdf2md.

Run from Source

cargo run --release --bin pdf2md -- ./input.pdf ./output.md

Self Hosting Streamlit App

A browser-based UI for uploading PDFs and previewing Markdown output with images.

The app automatically detects your OS and architecture to find the right binary in dist/. You can also specify a custom path:

pip install streamlit
streamlit run streamlit_app.py

Specify a custom binary or model directory:

streamlit run streamlit_app.py -- \
  --pdf2md-bin ./dist/pdf2md-<platform> \
  --model-dir /path/to/yolo26n-doclaynet_ncnn_model

Free PDF to Markdown API

No API key required. Submit a PDF and receive Markdown, extracted images, and a downloadable ZIP.

Endpoint

POST https://pdf2md.deepdiy.net/v1/convert
Content-Type: application/pdf

curl example

curl -X POST "https://pdf2md.deepdiy.net/v1/convert" \
  -H "Content-Type: application/pdf" \
  --data-binary @paper.pdf

Success response

{
  "status": "succeeded",
  "markdown": "# Paper title\n\nConverted Markdown...",
  "images": [
    {
      "path": "assets/page_0001_order_0001_class_6.png",
      "url": "https://..."
    }
  ],
  "zip_url": "https://...",
  "download_url": "https://...",
  "expires_in": 300
}

Error response (HTTP 429)

{
  "error": "busy"
}

The system processes one request at a time across all users. If the server is busy, it returns HTTP 429. Wait 1 second and retry. Each conversion runs for up to 120 seconds — you will likely get a slot within that window.

API Limits

Item	Value
Price	Free
Max PDF size	20 MB
Concurrency	One request at a time (returns 429 if busy)
Max task duration	120 seconds
Conversion timeout	150 seconds
Request timeout	180 seconds
ZIP download expiry	5 minutes

Detection Classes

You can use these class IDs to filter or block specific elements (e.g., Page-header, Footnote) from the output:

0: Caption, 1: Footnote, 2: Formula, 3: List-item, 4: Page-footer, 5: Page-header, 6: Picture, 7: Section-header, 8: Table, 9: Text, 10: Title