AuraAuth is an AI-powered tool that detects whether audio content is AI-generated or human-spoken. Built with deep learning, it's designed to promote media authenticity in an era of synthetic content.
🎥 Video deepfake detection is currently under development and will be integrated into future releases.
- 🎤 Audio Deepfake Detection
- Classifies audio as AI-generated or human.
- Provides a confidence/probability score.
- 🧠 Deep Learning-Based Classification
- Trained on a variety of real and synthetic speech datasets.
- ⚙️ Modular Python Backend
- Built using FastAPI/Flask (customizable).
- 💻 Modern Frontend Interface
- Built with React and Vite for seamless user interaction.
- 🔒 Future Expansion
- 📹 Video authenticity detection coming soon.
AuraAuth/
├── backend/ # Python backend (FastAPI or Flask)
│ ├── main.py
│ ├── Model/
│ │ └── model.py
│ └── visualize.py
├── frontend/ # React + Vite frontend
│ ├── src/
│ │ └── App.jsx
│ ├── index.html
│ └── package.json
├── app.py # (optional) wrapper entry point
└── README.md
cd backend
python -m venv venv
venv\Scripts\activate # On Windows
pip install -r requirements.txt
python main.pycd frontend
npm install
npm run dev- Input: Audio file (.wav, .mp3)
- Output: Classification (
Human/AI-Generated) + confidence - Model located in:
backend/Model/model.py
- ✅ Audio deepfake detection (stable)
- 🚧 Video deepfake detection (under development)
- ⏳ Model visualization (spectrogram overlays)
- ⏳ Docker support
- ⏳ REST API docs with Swagger / Redoc
We welcome contributions, ideas, and feedback!
git clone https://github.com/OP-Prajwal/AuraAuth.gitPlease open an issue or PR for major changes.
This project is licensed under the MIT License.
- PyDub, Librosa – audio processing
- Scikit-learn, TensorFlow – model training
- OpenAI & Google for public research datasets
⚠️ AuraAuth is an experimental project and should not be used in critical applications without further validation.