Hallo There! 🎥

An advanced AI-powered video generation tool that creates realistic talking avatars from images and audio.

🌟 Features

• 🎭 Avatar Generation: Create realistic talking avatars from static images
• 🗣️ Voice Processing: Advanced audio diarization using pyannote.audio
• 🎬 Video Synthesis: High-quality video generation with customizable settings
• 🔄 Multi-pose Support: Generate videos with multiple facial poses
• 🎨 Background Customization: Flexible background handling options

📋 Prerequisites

• Python 3.8+
• CUDA-compatible GPU (recommended)
• FFmpeg
• Hugging Face account and access token

🚀 Installation

1. Set up Python environment:

conda create --name hallo-there
conda activate hallo-there

2. Clone the repository:

git clone https://github.com/hiktan44/hallo-there.git
cd hallo-there

3. Install dependencies:

pip install -r requirements.txt
pip install .

4. Install FFmpeg:

• Linux: sudo apt-get install ffmpeg
• Windows: Download from official FFmpeg website and add to system PATH

⚙️ Configuration

1. Create Hugging Face access token:

   • Visit Hugging Face Token Settings
   • Generate new token with required permissions

2. Set up diarization:

python diarization.py -access_token <YOUR_HUGGING_FACE_TOKEN>

📁 Project Structure

hallo-there/
├── source_images/      # Input images (512x512)
├── audio/             # Input audio files
├── diarization/       # Diarization output
├── output/           # Generated video clips
└── docs/             # Documentation

🎮 Usage

1. Prepare source images:

   • 512x512 pixel squares
   • Face should occupy 50-70% of image
   • Place in source_images/ directory

2. Prepare audio:

   • Convert to WAV format
   • Place in audio/input_audio.wav

3. Generate video:

python generate_videos.py
python combine_videos.py

🔧 Advanced Options

• -mode full: Enable subtle head movements during silence
• -background custom: Use custom background image
• -quality high: Generate higher quality output

📚 Documentation

Detailed documentation available in docs/ directory:

• Installation Guide
• Configuration Options
• Troubleshooting

🤝 Contributing

Contributions welcome! Please read CONTRIBUTING.md for guidelines.

📄 License

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

🙏 Acknowledgments

• pyannote.audio for audio diarization
• Hugging Face for AI models and infrastructure