AmartyaC/polyglotpodcaster

🎙️ Polyglot Podcaster

A Streamlit-based web application for zero-shot multilingual voice cloning using the Chatterbox TTS model.

🌟 Features

• —Zero-Shot Voice Cloning: Record your voice or upload a short voice sample and generate speech in any supported language
• —Browser-Based Recording: Record audio directly in your browser - no need to upload files!
• —File Upload Support: Alternative option to upload pre-recorded .wav files
• —23 Languages Supported: English, Spanish, French, German, Italian, Portuguese, Polish, Turkish, Russian, Dutch, Arabic, Swedish, Danish, Greek, Finnish, Hebrew, Hindi, Japanese, Korean, Malay, Norwegian, Swahili, and Chinese
• —Simple Web Interface: Easy-to-use Streamlit interface with tabbed input options
• —GPU Acceleration: Automatically uses GPU if available for faster generation

🚀 Installation

Prerequisites

⚠️ Important: PyTorch currently requires Python 3.8-3.12. Python 3.13 is not yet supported.

If you have Python 3.13, you'll need to install Python 3.11 or 3.12:

# Using Homebrew on macOS
brew install python@3.11

Installation Steps

1. 1.Navigate to the project directory:

   cd "/Users/rajrup/Desktop/polyglot podcaster"

1. 1.Create a virtual environment with Python 3.11 or 3.12:

   # If you have Python 3.11
   python3.11 -m venv venv
   
   # OR if you have Python 3.12
   python3.12 -m venv venv
   
   # OR use the default python3 if it's 3.8-3.12
   python3 -m venv venv

1. 1.Activate the virtual environment:

   source venv/bin/activate

1. 1.Install PyTorch first:

   # For CPU-only (smaller download)
   pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu
   
   # OR for GPU support (if you have CUDA)
   pip install torch torchaudio

1. 1.Install remaining dependencies:

   pip install streamlit chatterbox-tts

📖 Usage

1. 1.Run the Streamlit app:

   streamlit run app.py

1. 1.Use the application:

Option A: Record Audio (Recommended)

• —Click the "🎤 Record Audio" tab
• —Click "🎙️ Click to Record" button to start recording
• —Speak for 10-15 seconds
• —Click "⏹️ Click to Stop" to finish recording
• —Your recorded audio will be displayed

Option B: Upload File

• —Click the "📤 Upload File" tab
• —Upload a 10-15 second .wav file of the target voice

Then:

• —Enter the text you want to generate
• —Select the language of your text
• —Click "🎵 Generate Speech"
• —Listen to the generated audio!

📋 Requirements

• —Python 3.10+
• —streamlit
• —streamlit-audiorecorder
• —chatterbox-tts
• —torch
• —torchaudio

🎯 How It Works

1. 1.The app loads the Chatterbox Multilingual TTS model (cached for performance)
2. 2.You provide a voice sample as a reference
3. 3.The model generates new speech in your chosen language using the voice characteristics from the sample
4. 4.The generated audio is saved and played back in the browser

⚠️ Notes

• —The first run will download the Chatterbox model (may take some time)
• —GPU is recommended for faster generation
• —Input audio should be in .wav format
• —Best results with clear, 10-15 second voice samples

📄 License

This project uses the Chatterbox TTS library. Please refer to the Chatterbox TTS license for usage terms.