Marker Document Converter

Convert PDF, EPUB, PPTX, DOCX, XLSX, and HTML files to clean Markdown format using the marker-pdf tool with multimodal LLM enhancement.

Prerequisites

# Install marker-pdf with full document support
uv tool install marker-pdf[full]

Requires Python 3.10+ and PyTorch.

Usage

When the user provides a document file path, run:

marker_single "<file_path>" \
  --output_format=markdown \
  --output_dir="<output_directory>" \
  --use_llm \
  --llm_service=marker.services.claude.ClaudeService \
  --claude_model_name=claude-haiku-4-5 \
  --claude_api_key=$ANTHROPIC_API_KEY \
  --disable_image_extraction

Note: --disable_image_extraction generates plain text output. Remove this flag if images need to be preserved.

Parameters

• <file_path>: Path to the input document (PDF, EPUB, PPTX, DOCX, XLSX, HTML)
• --output_format: Output format, options: markdown (default), html, json, chunks
• --output_dir: Directory to save output files (defaults to same directory as input)
• --use_llm: Enable LLM-based accuracy improvements for tables, formatting, and form extraction
• --page_range: Process specific pages (e.g., "0,5-10,20")

Examples

Convert PDF to Markdown

marker_single "./docs/report.pdf" \
  --output_format=markdown \
  --output_dir="./docs/" \
  --use_llm \
  --llm_service=marker.services.claude.ClaudeService \
  --claude_model_name=claude-haiku-4-5 \
  --claude_api_key=$ANTHROPIC_API_KEY \
  --disable_image_extraction

Convert PPTX to HTML

marker_single "./slides/presentation.pptx" \
  --output_format=html \
  --output_dir="./slides/" \
  --use_llm \
  --llm_service=marker.services.claude.ClaudeService \
  --claude_model_name=claude-haiku-4-5 \
  --claude_api_key=$ANTHROPIC_API_KEY \
  --disable_image_extraction

Convert specific pages

marker_single "./docs/book.pdf" \
  --output_format=markdown \
  --output_dir="./docs/" \
  --page_range="1-10" \
  --use_llm \
  --llm_service=marker.services.claude.ClaudeService \
  --claude_model_name=claude-haiku-4-5 \
  --claude_api_key=$ANTHROPIC_API_KEY \
  --disable_image_extraction

Advanced Options

• --force_ocr: Force OCR on entire document, converts inline math to LaTeX
• --redo_inline_math: Highest quality inline math conversion (use with --use_llm)
• --disable_image_extraction: Skip image extraction
• --paginate_output: Add page separators to output
• --debug: Enable diagnostic logging

Output

The tool generates:

• Formatted Markdown with tables, LaTeX equations ($$-fenced), code blocks
• Extracted images saved alongside the output
• Metadata with table of contents and page statistics

Instructions

1. Confirm the input file path exists
2. Determine output directory (default: same as input file)
3. Check if $ANTHROPIC_API_KEY environment variable is set
4. Run the marker_single command with appropriate options
5. Report the output file location and any extraction notes