Chatterbox Voice Cloning - Local Setup

This package contains everything you need to run Chatterbox Voice Cloning locally with a simple UI. After evaluating several open-source voice cloning solutions, Chatterbox TTS was selected as the best option due to its MIT license, state-of-the-art performance, simple Python API, and built-in web UI capabilities.

Quick Start

Windows Users

1. Double-click start_voice_cloning.bat
2. Wait for the UI to launch in your browser
3. Start cloning voices!

Linux/Mac Users

1. Open a terminal in this folder
2. Run chmod +x start_voice_cloning.sh (first time only)
3. Run ./start_voice_cloning.sh
4. Wait for the UI to launch in your browser

Requirements

• Python 3.11 ONLY (application will not run on other versions)
• CUDA-compatible GPU recommended for faster processing (but CPU works too)

First-Time Setup

If the launchers don't work immediately, you may need to:

1. Install Python from https://www.python.org/downloads/
2. Open a terminal/command prompt in this folder
3. Run: python -m venv venv
4. Activate the environment:
   • Windows: venv\Scripts\activate
   • Linux/Mac: source venv/bin/activate
5. Install dependencies: pip install -r requirements.txt
   • Note: If you encounter dependency resolution issues, try using pip install --no-deps -r requirements.txt followed by pip install torch torchaudio
6. Then try the launcher again

Windows-Specific Notes

• If you encounter dependency resolution issues during installation, the requirements.txt file contains pinned versions that should work together
• For Windows users with Python 3.13+:
  • The updated requirements.txt uses numpy<1.26.0 to avoid build failures on Python 3.13
  • If you still encounter numpy build errors, try installing with pip install --only-binary=numpy -r requirements.txt
  • Alternatively, consider using Python 3.10-3.12 for better compatibility with all dependencies
• If you see UnicodeDecodeError during installation, try setting your console to UTF-8: chcp 65001 before running pip

Using the Voice Cloning UI

1. Enter the text you want to convert to speech
2. Upload a voice reference audio file (5-10 seconds of clear speech works best)
3. Click "Generate Speech" to create your cloned voice audio
4. Adjust the advanced settings as needed:
   • Exaggeration: Controls emotional intensity (higher = more expressive)
   • Temperature: Controls randomness (higher = more variation)
   • CFG Weight: Controls adherence to text (higher = more precise)

Tips for Perfect Voice Cloning

For streamer chat TTS specifically:

• Use a high-quality reference audio with clear speech
• Keep messages relatively short for best results
• For consistent voice quality, use the same reference audio for all generations
• Try lower CFG weight values (~0.3) and higher exaggeration (~0.7) for more expressive speech

Troubleshooting

If you encounter any issues:

1. Make sure Python is installed and in your PATH
2. Check that you have enough disk space
3. For GPU acceleration, ensure you have CUDA installed
4. Try running the commands manually as described in "First-Time Setup"