git clone --recursive https://github.com/VintellX/VeriVoice.git
cd VeriVoice

Overview

VeriVoice is a DeepFake Audio Detection model, this is a PyTorch-based framework for detecting cloned or syntthesized speech. It uses a Graph-Enhanced Audio Transformer to capture both local spectral artifacts and global temporal patterns that distinguish real voices from their deepfakes or clones.

Quick Start

1. Install dependencies:

   conda create -n VeriVoice python=3.10
   conda activate VeriVoice
   pip install -r requirements.txt

2. Prepare data (check out usage.md for details).
3. Train the mdoel:

   python -m ipykernel install --user --name=VeriVoice --display-name "VeriVoice-Env"
   jupyter notebook

4. Test an audio file:

   python3 -m predicto path/to/audio.wav

License

This project is licensed under the Apache License 2.0. See LICENSE for details.