🎶 Fake vs Real Music Detector

A web-based system to classify audio as real (human-composed) or fake (AI-generated) using pretrained MERT embeddings and a CNN model.

🚀 Features

• 🎵 Upload .wav audio files and detect whether they are AI-generated or human-composed.
• 🔍 Uses m-a-p/MERT-v1-330M to extract audio embeddings.
• 🧠 CNN model trained on the FakeMusicCaps dataset.
• 🌐 Interactive Flask web interface.

---

🛠️ Installation

1. Clone the repository

git clone https://github.com/Nutan22341/Music-DeepFake.git
cd Music-DeepFake

2. Set up a virtual environment

python -m venv venv

Activate the environment:

• On Windows:

  venv\Scripts\activate

• On macOS/Linux:

  source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

If you're on Windows and need local .whl files for PyTorch, install them like:

pip install torch-2.1.0+cpu-cp39-cp39-win_amd64.whl
pip install torchaudio-2.1.0+cpu-cp39-cp39-win_amd64.whl
pip install torchvision-0.16.0+cpu-cp39-cp39-win_amd64.whl

4. Run the Web App

python app.py