If you use Windows and don't want to install Python or type commands, use an app instead of MarkItDown, Docling, Marker or Pandoc. MD Factory does the same job in a window: add files or a folder, and a Markdown file appears next to each one. The open-source tools are free, run on other systems and can be scripted, and this page says where that matters.
A disclosure: MD Factory is our product. Statements about the other tools come from their READMEs, documentation and source code, read on October 27, 2026. We did not run the tools for this page, so it does not compare the quality of their output. For that, see Best PDF to Markdown converters. This page compares what you have to install and do.
Short answer
| Tool | What you install | How you run it | Scanned PDFs |
|---|---|---|---|
| MarkItDown | Python 3.10 to 3.14 and the markitdown package | Command line or Python | Not in the built-in PDF reader; needs a plugin with a vision model, or Azure |
| Docling | Python 3.10 or later and the docling package; models download on first use | Command line or Python | OCR is on by default |
| Marker | Python 3.10 or later, PyTorch, and Docker or llama.cpp for OCR | Command line, Python, or a small browser app | OCR through a local inference server; skipped with --disable_ocr |
| Pandoc | One installer, or a zip, Chocolatey or winget | Command line | Does not read PDF at all |
| MD Factory | One app from the Microsoft Store | A window; add files or a folder | Local OCR, language chosen or detected |
What each tool asks you to set up
MarkItDown. The README says it requires Python 3.10 through 3.14 and recommends a virtual environment so dependencies do not clash. You install it with pip, then run markitdown with a file name. The README's optional extras decide which formats work, for example PDF, Word or Outlook messages; the [all] option installs them together. The maintainers also say that desktop applications are out of scope for the repository and should be separate projects.
Docling. The README says pip install docling works on macOS, Linux and Windows, and asks for Python 3.10 or higher. Its documentation says models download automatically on first use, with a separate command to download them ahead for offline use. Its supported formats page says legacy DOC and RTF files need LibreOffice.
Marker. The README asks for Python 3.10 or later and PyTorch. Formats other than PDF need the marker-pdf[full] install. OCR runs through a local inference server that Marker starts on first use, and you provide what it runs on: Docker plus the NVIDIA Container Toolkit on an NVIDIA GPU, or the llama-server program from llama.cpp on a CPU. Marker also includes marker_gui, a Streamlit app for trying it with basic options, installed with two more pip packages. The code is Apache 2.0. The README says the model weights are free for research, personal use and startups under $5 million in funding or revenue, and larger commercial use needs a paid license from Datalab.
Pandoc. This is the lightest to install. The installation page lists an installer for Windows, a zip with the programs, Chocolatey and winget. It is a command-line tool, as the manual says. It reads the document formats in its input list, not everything: see the next two sections.
Scanned PDFs
A scanned PDF is a picture of a page with no text inside. A converter needs OCR to read it, and the tools differ in how much work that takes.
- MarkItDown. The built-in PDF converter in its source code has no OCR step. The README offers two routes for scans: the
markitdown-ocrplugin, which sends page images to a vision model through an OpenAI-compatible client, or Azure Document Intelligence, whose calls are billed. The README adds that if the plugin has no client, OCR is silently skipped. In both routes you supply a model or a cloud service and its key. - Docling. The CLI reference shows OCR enabled by default, with Tesseract among the engines you can pick. Nothing extra to add for a basic scan.
- Marker. OCR is part of the tool but depends on the inference backend above. With
--disable_ocr, the README says scanned pages are skipped. - Pandoc. The FAQ answers the question "How can I convert PDFs to other formats using pandoc?" with "You can't." PDF appears in the manual only as an output format, and image files are not in the input list either. A scan cannot go through Pandoc at all.
- MD Factory. Scans are read with local OCR on your PC, in a language you choose or left on automatic. A page with low OCR confidence or no recognized text is marked Needs review instead of passing silently.
More on scans: converting scanned PDFs to Markdown.
Folders and batches
Docling's docling convert accepts a file, a directory or a URL (CLI reference). Marker's marker command takes an input folder and has --skip_existing to resume a run. For both, you type the command and choose where the output goes.
MarkItDown's README shows one file per command. In a June 2026 post announcing a desktop wrapper, its author wrote that converting a folder with MarkItDown "means writing your own script". That is one developer's description of his own experience, but it matches the README's examples.
Pandoc's manual says that when you give several input files, it concatenates them before parsing, with blank lines between them; --file-scope parses them individually. Getting one Markdown file per document is a shell loop.
In MD Factory you add a folder, or drag it into the window, and each document gets its own .md next to it, with pictures in a <name>_images folder. Source files are never changed. Each file ends as Completed, Needs review or Failed. Details: batch converting documents to Markdown.
Email, e-books and archives
All of the open-source tools cover Word, PowerPoint, Excel and HTML. The differences start with less common formats. "Not listed" means the format does not appear in the tool's README, format list or, for MarkItDown, the converters registered in its source.
| Format | MarkItDown | Docling | Marker | Pandoc | MD Factory |
|---|---|---|---|---|---|
| Outlook MSG | Yes, with the Outlook extra | Yes | Not listed | Not listed | Yes |
| EML | Not listed | Yes | Not listed | Not listed | Yes |
| EPUB | Yes | Yes | Yes | Yes | Yes |
| FB2 | Not listed | Not listed | Not listed | Yes | Yes |
| MOBI | Not listed | Not listed | Not listed | Not listed | Yes |
| ZIP | Yes, as one combined document | Yes, as separate documents (documented for the Python API) | Not listed | Not listed | Yes, one folder with a Markdown file per document |
| 7Z and RAR | Not listed | Not listed | Not listed | Not listed | Yes |
Docling is the broadest of the free tools here. Its supported formats page also lists Apple Pages, Numbers and Keynote, OpenDocument, audio and more, which MD Factory does not claim. If your files are EML and MSG, MD Factory output is covered in our EML and MSG guides; for e-books and archives, see MOBI and archives.
Free graphical wrappers
Community projects put a window on top of these libraries. One example is MDFlux, a desktop app built on MarkItDown. Its author describes it as a portable app that you unzip and run without a terminal, with local OCR and folder batches, and says it is MIT-licensed. We have not tried it. If a free project from GitHub suits you, it is worth a look. If you want an app from the Microsoft Store with a free trial, that is where MD Factory fits.
The same job in MD Factory
- Install MD Factory from the Microsoft Store.
- Add files or a folder, or drag them into the window. Conversion starts right away and runs on your PC; nothing is uploaded.
- Open the
.mdnext to each source file.
Converting again is safe. If the earlier result is still MD Factory's own, unedited output, it is replaced. If you edited it, the new result is saved as a numbered copy such as report (2).md.
Where the free tools are better
- Price. MarkItDown, Docling, Marker (within its weights license) and Pandoc are free. MD Factory has a 7-day free trial and then a one-time purchase.
- Scripting. They run from a script, a scheduled task or a server. MD Factory has no command line and cannot be called from code.
- Other systems. Docling's README lists macOS, Linux and Windows, and Pandoc's installation page covers macOS and Linux too. MD Factory runs on Windows only.
- Reach. Docling reads more formats than MD Factory, and its README lists integrations with LangChain, LlamaIndex and other AI frameworks. MD Factory has none.
- Word and EPUB sources. If you have the DOCX or EPUB and are fine with a command line, Pandoc converts it for free.
Which one to choose
- You write Python or run a pipeline: MarkItDown or Docling, then Marker if you have the hardware and the license fits.
- You have a DOCX or EPUB and a terminal does not bother you: Pandoc.
- You want a window, scans, whole folders, or email, e-book and archive files, on Windows: MD Factory.
Whichever you pick, convert five or ten of your own files first and compare the Markdown with the originals, especially tables and scans. The tools change often; check the linked pages before you decide. To keep documents on your PC while you do it, see converting documents without uploading them.