BakungaBronson/Luganda-Speech-To-Speech-Lite

A lite weight speech to speech Luganda assistant # Luganda Speech-to-Speech Lite A lightweight Luganda speech-to-speech system with support for Apple Silicon GPU acceleration and OpenAI-compatible APIs. ## Features - Speech-to-Text and Text-to-Speech for Luganda language - GPU acceleration support for Apple Silicon (M1/M2) via PyTorch MPS - OpenAI-compatible API endpoints - Modern React frontend with Shadcn UI - Docker support with multi-architecture builds ## Models This project uses the following open-source models: ### Speech-to-Text - Model: allandclive/whisper-tiny-luganda-v2 - Type: Fine-tuned Whisper model optimized for Luganda speech recognition ### Text-to-Speech - Text to Mel-Spectrogram: Sunbird/sunbird-lug-tts - Type: Tacotron2 model trained for Luganda speech synthesis - Vocoder: speechbrain/tts-hifigan-ljspeech - Type: HiFiGAN vocoder for high-quality waveform generation ## Prerequisites - Python 3.9+ (3.10+ recommended for Apple Silicon) - Node.js 18+ - Docker and Docker Compose (optional) - FFmpeg ## Installation ### Using Docker (Recommended) 1. Clone the repository: ```bash git clone cd luganda-speech-to-speech-lite ``` 2. Start the services using Docker Compose: ```bash # For Apple Silicon users (enables GPU acceleration) USE_MPS=true docker-compose up --build # For other platforms docker-compose up --build ``` The application will be available at: - Frontend: localhost - Backend API: localhost - API Documentation: localhost ### Manual Installation with Python Virtual Environment 1. Create and activate a Python virtual environment: ```bash # Create a new virtual environment python -m venv venv # Activate the virtual environment # On Windows: .\venv\Scripts\activate # On macOS/Linux: source venv/bin/activate ``` 2. Set up the backend: ```bash # Install Python dependencies pip install -r requirements.txt # Optional: Enable GPU acceleration on Apple Silicon export USE_MPS=true # Start the FastAPI server cd backend uvicorn main:app --host …