I’ve been working on a deepfake audio and voice detection tool called VoiceAuth. It uses deep learning models to detect if audio is manipulated or if it’s generated by a deepfake algorithm. I’ve built this tool with a simple, user-friendly interface in Streamlit, making it easy to test out. It supports a variety of audio and video formats.

Key Features:

    Multiple Models: Includes detection using Random Forest, Melody, and 960h deep learning models.
    Audio Extraction: Supports extracting audio from videos (MP4, MKV, AVI, etc.).
    Visualization: Displays audio features like MFCCs and Mel Spectrograms for analysis.
    Real-Time Predictions: Instantly analyzes the uploaded audio/video and gives you a confidence score for deepfake detection.

    Upload an audio or video file.
    Select the model(s) you'd like to run (Random Forest, Melody, 960h, or All).
    Click on Run Prediction.
    View the results in real time, including confidence scores, audio visualizations, and file metadata.

With the rise of deepfake technology, detecting manipulated audio is becoming critical for verifying media authenticity. VoiceAuth provides a simple way to analyze whether an audio clip or video might contain a deepfake.

You can try it out live by uploading your own files directly on the app. Demo:

https://voicedetector.streamlit.app/

Let me know what you think! Any feedback or suggestions would be greatly appreciated.