Zerox

Convert PDFs, DOCX, PPTX, and images to Markdown using zerox with GPT-4o vision, including OCR for scanned documents.

Zerox Document Converter

Convert various document formats to Markdown using the zerox library and GPT-4o vision.

Supported Formats

• PDF (scanned and text-based)
• Microsoft Word (DOCX)
• Microsoft PowerPoint (PPTX)
• Images (PNG, JPG, etc.)
• And more via OCR

Convert Document (Foreground)

For small files (< 30 seconds):

node {baseDir}/scripts/convert.mjs <filePath> [outputPath]

Examples

# Convert PDF - saves to {baseDir}/output/document.md by default
node {baseDir}/scripts/convert.mjs "/path/to/document.pdf"

# Convert PDF with custom output path
node {baseDir}/scripts/convert.mjs "/path/to/document.pdf" "/path/to/output.md"

# Convert Word document - saves to {baseDir}/output/document.md
node {baseDir}/scripts/convert.mjs "/path/to/document.docx"

Convert Document (Background)

For large files or scanned PDFs that take minutes:

node {baseDir}/scripts/convert-bg.mjs <filePath> [outputPath]

Features

• Runs conversion in background (no timeout issues)
• Logs progress to {baseDir}/output/convert-bg.log
• Sends macOS notification when complete
• Detached from terminal (safe to close)

Examples

# Convert large scanned PDF in background
node {baseDir}/scripts/convert-bg.mjs "/path/to/scanned-document.pdf"

# Monitor progress
tail -f {baseDir}/output/convert-bg.log

Requirements

• APIYI_API_KEY: Your OpenAI-compatible API key (environment variable)

Notes

• The conversion uses GPT-4o vision to extract text, so it works even with scanned documents
• Large documents may take some time to process
• Output is plain Markdown text