LegacyLipi

Legacy Font PDF Translator - Translate PDF documents with legacy Indian font encodings to English.

Problem

Millions of government documents, legal papers, and archival materials in Indian regional languages (Marathi, Hindi, Tamil, etc.) were created using legacy font encoding systems (Shree-Lipi, Kruti Dev, APS, Chanakya, etc.). These fonts map Devanagari/regional script glyphs to ASCII/Latin code points, making them unreadable by standard translation tools.

Example:

• What the PDF displays: महाराष्ट्र राजभाषा अधिनियम
• What text extraction produces: ´ÖÆüÖ¸üÖ™Òü ¸üÖ•Ö³ÖÖÂÖÖ †×¬Ö×®ÖμÖ´Ö
• What Google Translate sees: Gibberish

Solution

LegacyLipi:

1. Detects the font encoding scheme used in a PDF (legacy or Unicode)
2. Converts legacy-encoded text to proper Unicode
3. Alternatively, uses OCR (Tesseract or EasyOCR) to extract text from scanned PDFs
4. Translates the Unicode text to the target language
5. Outputs translated text in various formats (text, markdown, PDF) with optional bilingual side-by-side output

Installation

From PyPI (Recommended)

pip install legacylipi

Or with uv (one command, no install):

uvx legacylipi api

Or install as a tool:

uv tool install legacylipi

From Source

git clone https://github.com/biswasbiplob/legacylipi.git
cd legacylipi
uv sync

Docker

# Build the image
docker build -t legacylipi .

# Run with Docker
docker run -p 8000:8000 legacylipi

# Or use Docker Compose
docker compose up

The web UI will be available at http://localhost:8000.

To process local files, mount volumes:

docker run -p 8000:8000 -v ./input:/app/input -v ./output:/app/output legacylipi

System Dependencies

Tesseract (for OCR - recommended):

# Ubuntu/Debian
sudo apt-get install tesseract-ocr tesseract-ocr-mar tesseract-ocr-hin

# macOS
brew install tesseract tesseract-lang

Translate-Shell (recommended translation backend):

# Ubuntu/Debian
sudo apt-get install translate-shell

# macOS
brew install translate-shell

Quick Start

# Basic translation
legacylipi translate input.pdf -o output.txt

# Output as PDF (preserves layout)
legacylipi translate input.pdf -o output.pdf --format pdf

# Bilingual side-by-side output
legacylipi translate input.pdf -o output.pdf --bilingual

# OCR for scanned documents
legacylipi translate input.pdf --use-ocr -o output.txt

# Use local LLM (requires Ollama)
legacylipi translate input.pdf --translator ollama --model llama3.2

# Detect encoding only
legacylipi detect input.pdf

See docs/cli-reference.md for complete CLI documentation.

Web UI

LegacyLipi includes a modern React-based web interface backed by a FastAPI REST API.

Production (single command)

# Serves the built React frontend + API on one port
legacylipi api
# or
uvx legacylipi api

Open http://localhost:8000 in your browser.

Development (hot-reload)

# Start both FastAPI backend and Vite dev server
./scripts/dev.sh

This runs:

• Backend at http://localhost:8000 (FastAPI with auto-reload)
• Frontend at http://localhost:5173 (Vite dev server with HMR, proxies /api to backend)

Workflow Modes:

• Scanned Copy - Create image-based PDF copy (adjust DPI, color, quality)
• Convert to Unicode - OCR + Unicode conversion without translation
• Full Translation - Complete pipeline with OCR, conversion, and translation

Features:

• Drag-and-drop PDF upload
• Workflow-based UI with mode selection
• Multiple translation backends (Translate-Shell, Google, Ollama, OpenAI, etc.)
• OCR support with EasyOCR and Tesseract engine selection
• Structure-preserving or flowing text modes
• Bilingual side-by-side output
• Source language auto-detection from encoding
• Real-time SSE progress streaming
• Direct download of translated files
• Responsive dark-theme design

Supported Encodings

Encoding	Font Family	Language	Status
shree-dev	SHREE-DEV-0708, 0714, 0715, 0721	Marathi	Built-in
shree-lipi	Shree-Lipi, SDL-DEV	Marathi	Built-in
dvb-tt	DVBWTTSurekh, DVBTTSurekh	Marathi	Built-in
kruti-dev	KrutiDev010, KrutiDev040	Hindi	Built-in
chanakya	Chanakya	Hindi/Sanskrit	Built-in
aps-dv	APS-DV-TT	Hindi	Built-in
walkman-chanakya	Walkman Chanakya	Hindi	Built-in
shusha	Shusha	Marathi/Hindi	Built-in

Translation Backends

Backend	Description	Setup
trans	translate-shell CLI (recommended)	brew install translate-shell
google	Google Translate (free API)	Works out of the box
mymemory	MyMemory API (free)	Works out of the box
ollama	Local LLM via Ollama	Ollama required
openai	OpenAI GPT models	Set OPENAI_API_KEY
gcp_cloud	Google Cloud Translation	GCP project + credentials

See docs/translation-backends.md for detailed setup guides.

OCR Backends

Both OCR engines are included as core dependencies:

Backend	Description	GPU Support
EasyOCR	Local, free, good for Indian languages (default)	CUDA, MPS (Apple Silicon)
Tesseract	Local, free, most language packs	CPU only
Google Vision	Cloud, paid, best accuracy	N/A

Google Vision requires an additional install: pip install legacylipi[vision]

See docs/cli-reference.md for detailed OCR options and language codes.

CLI Commands

Command	Description
api	Launch the React web UI + FastAPI REST API
translate	Full pipeline: parse, detect, convert, translate, output
convert	Convert legacy encoding to Unicode (no translation)
extract	Extract text from PDF (OCR or font-based)
detect	Analyze PDF and report detected encoding
scan-copy	Create an image-based scanned copy of a PDF
encodings	List supported font encodings
usage	Show API usage statistics

See docs/cli-reference.md for full command reference.

Architecture

┌─────────────────────────────────────────────────────────────────────────┐
│                              LegacyLipi                                 │
├─────────────────────┬───────────────────────────────────────────────────┤
│   React Frontend    │                  FastAPI Backend                   │
│  (Vite + TS + TW)   │                                                   │
│                     │   ┌──────────────────────────────────────────┐    │
│  FileUploader       │   │              REST API                    │    │
│  WorkflowSelector   │   │  /api/v1/config/*     GET config         │    │
│  Settings panels    │◄─▶│  /api/v1/sessions/*   Upload/delete      │    │
│  StatusPanel (SSE)  │   │  /api/v1/sessions/*/  Start pipeline     │    │
│  DownloadButton     │   │  /api/v1/sessions/*/progress  SSE stream │    │
│                     │   │  /api/v1/sessions/*/download  Get result │    │
│                     │   └────────────────────┬─────────────────────┘    │
├─────────────────────┘                        │                          │
│                                              ▼                          │
│  ┌──────────────────────────────────────────────────────────────────┐   │
│  │                      Core Pipeline                               │   │
│  │                                                                  │   │
│  │  PDF Parser / OCR Parser (Tesseract + EasyOCR)                   │   │
│  │       │                                                          │   │
│  │  Encoding Detector → Unicode Converter                           │   │
│  │       │                                                          │   │
│  │  Translation Engine (trans, Google, Ollama, OpenAI, GCP, ...)    │   │
│  │       │                                                          │   │
│  │  Output Generator (.txt, .md, .pdf, bilingual)                   │   │
│  └──────────────────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────────────────┘

Development

See docs/development.md for setup instructions, running tests, project structure, and adding new encodings.

License

MIT

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Run checks (./scripts/check.sh)
4. Commit and push
5. Open a Pull Request