AI Driven Deepfake Detection System

Welcome

This system detects AI-generated (deepfake) content in images and audio using transformer-based deep learning models.

Supported inputs

  • Images: JPG, PNG (face-centric images recommended)
  • Audio: WAV, MP3, FLAC, M4A, OGG formats (clear speech preferred)

How to use

  1. Select a detection mode using the tabs above.
  2. Upload an image or audio file.
  3. Click Submit to start analysis.
  4. Review the prediction, confidence score, and risk assessment.

Understanding the results

  • Prediction: Model decision (Real / Fake)
  • Confidence: Certainty percentage of the prediction
  • Risk Assessment:
    • High likelihood → strong indication
    • Possibly → caution advised
    • Uncertain → manual review recommended

Explainability

For images, attention heatmaps highlight the facial regions that influenced the model’s decision, supporting transparency and forensic analysis.

Data privacy & intended use

Uploaded files are processed temporarily and are not stored. This system is intended as a decision-support tool and should not be used as the sole source of verification.