download. First run downloads model weights and may spawn a local inference server; expect a large download.
license. Model weights are free for research, personal use and small startups; check the license for commercial use.
Who it is forA researcher or analyst with a paper or report they want as editable text.
The problemCopying text out of PDFs breaks tables, math and reading order.
The outcomeA markdown file of the document in the output folder you choose.
What you need first
- Python 3.10+ and PyTorch
- pip
- A GPU is best, CPU and Apple Silicon also work in fast mode
Steps
- Install with
pip install marker-pdf - Run
marker_single /path/to/file.pdfto convert one file - Add
--output_formatand--output_dirto choose format and folder - Open the markdown and check tables and equations
Prompt for your agent
Using datalab-to/marker, install it with `pip install marker-pdf`, run `marker_single /path/to/file.pdf --output_format markdown --output_dir ./out` on <my PDF>, and summarise any pages where the output looks wrong.
Caveat. Balanced mode needs a GPU for best speed; CPU runs default to fast mode.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for datalab-to/marker and "Convert a PDF to clean markdown"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer