MarkItDown GUI

A desktop frontend for Microsoft's markitdown library, with optional NVIDIA NIM-powered OCR / image understanding.

Convert PDFs, Office documents, images, HTML, and more into clean Markdown — with a drag-and-drop queue and a live, themed preview.

Features

• Drag-and-drop file queue with per-file status (pending / converting / done / error)
• Optional OCR via NVIDIA NIM (Nemotron vision models)
• Live Markdown preview (rendered + raw) with copy / save
• Light / Dark / System themes
• Batch conversion that never halts on a single failed file

Setup

pip install -r requirements.txt

Create your environment file and add your NVIDIA API key (only needed for OCR):

copy .env.example .env      # Windows
# cp .env.example .env       # macOS / Linux

Then edit .env:

NVIDIA_API_KEY=nvapi-your-key-here

Get a key at https://build.nvidia.com (free tier available). You can also paste the key later via the gear (settings) dialog in the app.

Run

python main.py

Notes

• OCR is optional. Turn it off to convert documents that don't need image understanding (no API key required).
• Output .md files are written next to each source file by default, or to a folder you choose in the settings bar.
• Settings (theme, model, OCR, output folder) persist in config.json, created on first run.