Voice-to-Voice Translation System

A Gradio-based application for end-to-end voice translation. It uses pyannote for speaker diarization, NVIDIA Canary-1b-v2 for translation, and Piper-TTS for voice synthesis.

---

Installation

Prerequisites

• NVIDIA GPU with CUDA Toolkit 12.8 installed.
• Python 3.10 (ensure it's added to your system's PATH).

Setup Steps

1. Clone the repository:

   git clone https://github.com/Juste-Leo2/VoiceToVoice-Translation.git
   cd VoiceToVoice-Translation

2. Create and activate a virtual environment:

   python -m venv .venv
   
   # On Windows
   .venv\Scripts\activate
   
   # On Linux / macOS
   # source .venv/bin/activate

3. Install dependencies using uv:

   python -m pip install --upgrade pip
   pip install uv
   uv pip install -r requirements.txt --no-deps --index-strategy unsafe-best-match

Usage

1. Run the application:

   python app.py

2. Open the local URL provided in the terminal (e.g., http://127.0.0.1:7860) in your browser.