Authors: Hoang Tang, Derek Duenas, Ishita Gupta
This repository contains demo code for generating deepfake datasets for the Generative AI course (10-623). The project focuses on transforming real videos with corresponding audio into deepfake videos by replacing the original audio with synthetic ones, creating a realistic speech-to-face synthesis.
- 📄 ArXiv | 🌐 Project Page | 💻 GitHub Repo
The project utilizes multiple datasets to train and test the deepfake generation. Below is a list of datasets used in the project:
- LibriSpeech (real audio)
- LibriSeVoc (fake audio)
- FakeAVCeleb (real videos)
- GRID (audio-visual speech corpus)
- LipSyncTIMIT