MarkItDown — Convert Any File to Markdown for LLMs

Supported Formats

Batch Conversion

from pathlib import Path

md = MarkItDown()
for file in Path("./documents").glob("*.*"):
    result = md.convert(str(file))
    Path(f"./markdown/{file.stem}.md").write_text(result.text_content)

Image OCR (with LLM)

md = MarkItDown(llm_client=openai_client, llm_model="gpt-4o")
result = md.convert("screenshot.png")
# Uses vision model to describe and extract text from images

Audio Transcription

result = md.convert("meeting_recording.mp3")
# Uses Whisper for speech-to-text, outputs as Markdown

RAG Pipeline Integration

from markitdown import MarkItDown
from langchain.text_splitter import RecursiveCharacterTextSplitter

md = MarkItDown()
doc = md.convert("quarterly_report.pdf")
chunks = RecursiveCharacterTextSplitter(chunk_size=512).split_text(doc.text_content)
# Feed chunks into your vector database

MarkItDown vs Docling

Key Stats

• 8,000+ GitHub stars
• By Microsoft
• 10+ input formats
• Image OCR and audio transcription
• Single API for all formats

FAQ

Q: What is MarkItDown? A: A Python library by Microsoft that converts PDF, DOCX, PPTX, XLSX, images, audio, and HTML to clean Markdown for LLM consumption.

Q: Is MarkItDown free? A: Yes, open-source under MIT license.

Q: MarkItDown or Docling? A: MarkItDown for diverse formats (10+ types). Docling for high-accuracy PDF parsing with complex layouts.