mmpdfkit

Convert Burmese PDFs to clean, usable Markdown and text for AI applications, data analysis, and vectorization.

What is mmpdfkit?

mmpdfkit solves a critical problem for anyone working with Burmese/Myanmar text: extracting usable content from PDFs with mixed encodings, legacy fonts, and scanned documents.

Burmese PDFs often contain text in multiple non-Unicode encodings (Win Myanmar, Zawgyi) or are entirely scanned. This makes them unsuitable for AI model input, vectorization, or modern text processing pipelines. mmpdfkit automatically:

• Detects and converts legacy Myanmar encodings (Win Myanmar, Zawgyi) to proper Unicode
• Extracts text with layout preservation via Markdown formatting
• OCR scans for documents that are image-based
• Preserves structure (headings, paragraphs, spacing) during conversion

Use Cases

• AI/LLM contexts — Clean Burmese text for prompt context or fine-tuning
• Vectorization — Prepare PDFs for embedding and vector databases
• Text analysis — Linguistic research on Burmese corpora
• Content migration — Convert legacy Burmese digital archives to modern formats

Install

As a CLI Tool (Recommended)

With full OCR support for scanned PDFs:

uv tool install "mmpdfkit[ocr]"

Without OCR (faster, for digital PDFs only):

uv tool install mmpdfkit

Then use the mmpdfkit command:

mmpdfkit example.pdf              # Convert to Markdown
mmpdfkit inspect example.pdf      # Extract metadata as JSON

As a Python Library

Standard install:

pip install mmpdfkit

With OCR support:

pip install "mmpdfkit[ocr]"

For Development

git clone https://github.com/kaungsithu/mmpdfkit.git
cd mmpdfkit
pip install -e ".[dev,ocr]"
pre-commit install

Usage

CLI Examples

After installing with uv tool install "mmpdfkit[ocr]":

# Convert PDF to Markdown (output next to input)
mmpdfkit example.pdf                        # → example.md

# Convert all PDFs in a directory
mmpdfkit samples/                           # → all .md files in same dir

# Save to custom output directory
mmpdfkit example.pdf --output-dir ./out/

# Disable OCR for faster processing (digital PDFs)
mmpdfkit example.pdf --no-ocr

Inspect PDF Metadata

# Extract font/text metadata as JSON
mmpdfkit inspect example.pdf    # → example_inspection.json

# Inspect all PDFs in directory
mmpdfkit inspect samples/

One-Shot Usage (No Install Required)

Use uvx to run mmpdfkit without installing it globally:

# Without OCR (fastest)
uvx mmpdfkit example.pdf

# With OCR for scanned PDFs
uvx "mmpdfkit[ocr]" example.pdf

Tip: For repeated use, uv tool install is faster than uvx since it caches the installation.

Library Usage

from mmpdfkit import pdf_to_markdown, inspect_pdf

# Convert PDF to markdown string
md = pdf_to_markdown("example.pdf")

# Inspect PDF metadata
inspection = inspect_pdf("example.pdf")

OCR Support

When installed with [ocr] extra, scanned PDFs are automatically processed with optical character recognition.

Disable OCR by default (optional configuration at ~/.mmpdfkit/config.yaml):

enable_ocr: false

Or use --no-ocr flag for individual conversions:

mmpdfkit scanned.pdf --no-ocr

Running from Source (Development)

After cloning and installing with pip install -e ".[dev,ocr]":

# CLI (same as installed version)
mmpdfkit example.pdf
mmpdfkit inspect example.pdf

Testing

Run the test suite:

pytest tests/ -v

Test fixture: test-pdfs/test.pdf is a minimal 3-page fixture combining sample pages from various Myanmar PDFs (digital typeset + scanned pages) for testing both text extraction and OCR pipelines.