Markitdown

Source: https://github.com/aipoch/medical-research-skills

When to Use

• Converting research papers or reports (PDF/DOCX/EPUB/HTML) into Markdown for LLM summarization, Q&A, or RAG indexing.
• Extracting tables and structured content from spreadsheets (XLSX/CSV) into Markdown for analysis or documentation.
• Turning slide decks (PPTX) into Markdown notes, including speaker notes and (optionally) AI-generated image descriptions.
• Processing images or scanned documents with OCR to obtain searchable, editable Markdown text.
• Transcribing audio (WAV/MP3) or pulling YouTube transcripts into Markdown for meeting notes, content analysis, or knowledge bases.

Key Features

• Converts many formats to structured Markdown (PDF, DOCX, PPTX, XLSX, images, audio, HTML, CSV, JSON, XML, ZIP, EPUB, YouTube URLs, etc.).
• Produces token-efficient output suitable for LLM pipelines (summarization, chunking, embedding).
• OCR support for images/scans (when OCR dependencies are installed).
• Audio transcription support (when transcription dependencies are installed).
• Optional AI-enhanced image/slide descriptions via an OpenAI-compatible client (e.g., OpenRouter).
• Plugin system to extend format support and custom behaviors.
• Stream-based conversion API for large files.

Dependencies

• Python: >=3.9 (recommended)
• Package:
  • markitdown[all] (installs all optional format handlers)

Optional system dependencies (feature-dependent):

• Tesseract OCR: tesseract-ocr (for image/scanned-text OCR)

Optional external services (feature-dependent):

• Azure Document Intelligence endpoint (for enhanced PDF extraction)
• OpenAI-compatible LLM endpoint (e.g., OpenRouter) for AI image descriptions

Example Usage

Install

pip install 'markitdown[all]'

CLI: Convert a PDF to Markdown

markitdown document.pdf -o output.md

Python: Convert multiple formats (PDF/XLSX/PPTX/DOCX) and save outputs

from pathlib import Path
from markitdown import MarkItDown

md = MarkItDown()

files = [
    "document.pdf",
    "spreadsheet.xlsx",
    "presentation.pptx",
    "notes.docx",
]

for path in files:
    result = md.convert(path)
    out = Path(path).with_suffix(".md")
    out.write_text(result.text_content, encoding="utf-8")
    print(f"Converted {path} -> {out}")

Python: Stream conversion (useful for large files)

from markitdown import MarkItDown

md = MarkItDown()

with open("large_file.pdf", "rb") as f:
    result = md.convert_stream(f, file_extension=".pdf")

with open("large_file.md", "w", encoding="utf-8") as out:
    out.write(result.text_content)

Python: AI-enhanced image/slide descriptions (OpenAI-compatible, e.g., OpenRouter)

from markitdown import MarkItDown
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_OPENROUTER_API_KEY",
    base_url="https://openrouter.ai/api/v1",
)

md = MarkItDown(
    llm_client=client,
    llm_model="anthropic/claude-opus-4.5",
    llm_prompt="Describe this image in detail for scientific documentation.",
)

result = md.convert("presentation.pptx")
print(result.text_content)

Implementation Details

• Conversion entry points

  • MarkItDown().convert(path) converts a file by path/URL and returns an object whose primary payload is result.text_content (Markdown).
  • MarkItDown().convert_stream(stream, file_extension=".pdf") converts from a binary stream; use this for large files or when data is not on disk.

• Format handling

  • Format support is provided by optional extras (e.g., pdf, docx, pptx, xlsx, audio-transcription, youtube-transcription) or all.
  • ZIP inputs are typically processed by iterating through contained files and converting each supported entry.

• OCR

  • For images/scanned documents, OCR is enabled when OCR tooling is available (commonly Tesseract). Ensure the OS-level OCR binary is installed and accessible in PATH.

• AI image descriptions

  • When llm_client, llm_model, and llm_prompt are provided, MarkItDown can request model-generated descriptions for images (including slide images), then inject those descriptions into the Markdown output.
  • Any OpenAI-compatible client can be used (e.g., OpenRouter) by setting base_url and api_key.

• Enhanced PDF extraction (Azure Document Intelligence)

  • When configured with a Document Intelligence endpoint, PDF extraction can be improved for complex layouts (tables, multi-column text, scanned PDFs), producing more faithful Markdown structure.

• Plugins

  • Plugins can be listed and enabled from the CLI (e.g., --list-plugins, --use-plugins) to extend conversion behavior or add new format handlers.