ArtificialPersona

Published:

Team project at Queen’s University (with Ovil Miranda, Carter Moore, Zane Prance, Kanika Poonia, Faiz Farook, and Safowan Mostaque) demonstrating real-time anonymisation across both audio and video channels. The pipeline chains:

  1. Speech-to-text transcribes incoming audio.
  2. Text-to-speech synthesises the transcript in a different voice profile.
  3. Deepfake swaps the user’s face for a synthetic persona in the video feed.

The goal: show that high-fidelity end-to-end anonymisation of online communication is feasible with off-the-shelf neural networks running close to real time.

GitHub

Direct Link