🧠 Synthetic-Voice-Detection-Vocoder-Artifacts

---

📁 LibriSeVoc Dataset

1. We are the first to identify neural vocoders as a source of features to expose synthetic human voices.
   Here are the differences shown by the six vocoders compared to the original audio:

   

2. We provide LibriSeVoC as a dataset of self-vocoding samples created with six state-of-the-art vocoders to highlight and exploit the vocoder artifacts.
   The composition of the dataset is shown in the following table:

   

   The source of our dataset ground truth comes from LibriTTS. Therefore, we follow the naming logic of LibriTTS.
   For example:
   27_123349_000006_000000.wav →

   • 27 is the reader's ID
   • 123349 is the ID of the chapter

---

🎯 Deepfake Detection

We propose a new approach to detecting synthetic human voices by:

• Exposing signal artifacts left by neural vocoders
• Modifying and improving the RawNet2 baseline by adding multi-loss

✅ This lowers the error rate from 6.10% to 4.54% on the ASVspoof Dataset.

Here is the framework of the proposed synthesized voice detection method:

---

📄 Paper & Dataset

• 📘 Paper:
  AI-Synthesized Voice Detection Using Neural Vocoder Artifacts – CVPRW 2023

• 📦 Dataset:
  Download LibriSeVoc

---

🛠️ Usage

🏋️‍♀️ To train the model, run:

python main.py --data_path /your/path/to/LibriSeVoc/ --model_save_path /your/path/to/models/

🧪 To test with your sample, run:

python eval.py --input_path /your/path/to/sample.wav --model_path /your/path/to/your_model.pth

📥 Pretrained Model Weights

Download the trained model weights from the link below:

https://drive.google.com/file/d/15qOi26czvZddIbKP_SOR8SLQFZK8cf8E/view?usp=sharing

🌐 In-the-Wild Testing

You can test audio samples live on our lab's Deepfake O Meter platform:

https://zinc.cse.buffalo.edu/ubmdfl/deep-o-meter/landing_page

📄 License

This repository is licensed under the MIT License.
You are free to use, modify, and distribute the code with proper attribution.

• 🔗 MIT License