🎤 Bhavesh AI Voice Cloner

🚀 Advanced AI Voice Cloning & Text-to-Speech

Transform any text into natural speech using your own voice or any reference audio

---

Bhavesh AI Voice Cloner is a cutting-edge, open-source Text-to-Speech (TTS) and voice cloning system powered by state-of-the-art AI technology. With just a few seconds of reference audio, you can clone any voice and generate natural-sounding speech with emotion control and perfect pronunciation.

✨ Key Features

• 🎯 Zero-shot Voice Cloning: Clone any voice with just seconds of reference audio
• 🎭 Emotion Control: Adjust emotional intensity and expression levels
• ⚡ Lightning Fast: Optimized for real-time generation
• 🧠 0.5B Llama Backbone: Built on state-of-the-art transformer architecture
• 🎵 High Quality Audio: Crystal clear, natural-sounding speech output
• 🔒 Ethical AI: Built-in watermarking for responsible AI usage
• 🌐 Multiple Languages: Supports multiple languages and accents
• 🚀 Easy to Use: Simple API and beautiful Streamlit interface

🎯 Use Cases

• 🎬 Content Creation: Videos, podcasts, audiobooks
• 🎮 Gaming: Character voices and narration
• 📱 Applications: Voice assistants and chatbots
• 🎭 Entertainment: Memes, voice effects, creative projects
• 📚 Education: Language learning and accessibility tools

🚀 Quick Start

📦 Installation

Method 1: Automated Setup (Recommended)

# Clone the repository
git clone https://github.com/bhau23/voice-cloner.git
cd voice-cloner

# Run the setup script
python launch.py setup

# Launch the Streamlit app
python launch.py streamlit

Method 2: Manual Installation

# Clone the repository
git clone https://github.com/bhau23/voice-cloner.git
cd voice-cloner

# Install dependencies
pip install -r requirements.txt

# Run the Streamlit app
streamlit run streamlit_app.py

Method 3: Docker (One-Click)

git clone https://github.com/bhau23/voice-cloner.git
cd voice-cloner
python launch.py docker

Method 4: Install as Package

pip install bhavesh-ai-voice-cloner

💻 Usage

Python API

import torchaudio as ta
from bhavesh_ai_voice_cloner.tts import BhaveshTTS

# Initialize the model
model = BhaveshTTS.from_pretrained(device="cuda")

# Generate speech with default voice
text = "Hello! This is Bhavesh AI speaking. I can clone any voice and make it sound natural!"
wav = model.generate(text)
ta.save("output.wav", wav, model.sr)

# Clone a specific voice
REFERENCE_AUDIO = "path/to/your/reference.wav"
wav = model.generate(text, audio_prompt_path=REFERENCE_AUDIO)
ta.save("cloned_voice.wav", wav, model.sr)

Streamlit Web Interface

For a user-friendly interface, run the Streamlit app:

# Quick launch
python launch.py streamlit

# Or traditional method
streamlit run streamlit_app.py

Command Line Interface

Generate speech directly from command line:

python launch.py cli "Your text here"

Interactive Demo

Run the comprehensive demo:

python launch.py demo

🛠️ Project Management

Setup Development Environment

python launch.py setup

Verify Project Status

python verify_project.py

Deploy to GitHub

python github_setup.py

🎛️ Advanced Parameters

• exaggeration (0.25-2.0): Control emotional intensity (default: 0.5)
• cfg_weight (0.0-1.0): Control speech pacing and clarity (default: 0.5)
• temperature (0.05-5.0): Control randomness in generation (default: 0.8)
• repetition_penalty (1.0-2.0): Prevent repetitive speech patterns (default: 1.2)

🌍 Supported Languages

Currently supporting English with plans to expand to more languages.

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

🙏 Acknowledgements

This project builds upon excellent open-source work:

• Cosyvoice
• Real-Time-Voice-Cloning
• HiFT-GAN
• Llama 3
• S3Tokenizer

🔐 Responsible AI & Watermarking

Every audio file generated by Bhavesh AI includes built-in watermarking technology for responsible AI usage. This helps identify AI-generated content and prevents misuse.

Watermark Detection

import perth
import librosa

# Load the generated audio
audio, sr = librosa.load("generated_audio.wav", sr=None)

# Initialize watermarker
watermarker = perth.PerthImplicitWatermarker()

# Check for watermark
watermark = watermarker.get_watermark(audio, sample_rate=sr)
print(f"AI Generated: {watermark > 0.5}")

📞 Support & Community

• 🐛 Bug Reports: GitHub Issues
• 💬 Discussions: GitHub Discussions
• 📧 Contact: bhavesh23official@gmail.com

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🏆 Citation

If you use this project in your research, please cite:

@misc{bhaveshaivoicecloner2025,
  author       = {{Bhavesh AI}},
  title        = {{Bhavesh AI Voice Cloner}},
  year         = {2025},
  howpublished = {\url{https://github.com/bhau23/voice-cloner}},
  note         = {GitHub repository}
}

⚠️ Disclaimer

This tool is intended for legitimate and ethical uses only. Users are responsible for ensuring they have proper consent before cloning someone's voice. Please use this technology responsibly and in compliance with applicable laws.

---

Made with ❤️ by Bhavesh AI

GitHub • Try Online • Contact