This is a 288x288 wav2lip model version. Some Features I will implement here
- input size 288x288
- PRelu
- LeakyRelu
- Gradient penalty
- Wasserstein Loss
- Kubeflow for training process
- Orthogonal regularization with gradient penalty
I trained my own model on AVSPEECH dataset and then transfer learning with my private dataset.