You will need to install cmake first (required for dlib, which is used for face alignment). You will also need to download shape_predictor_68_face_landmarks.dat for dlib, put it in the main folder
conda create --name torchfakes
activate torchfakes
git clone https://github.com/IanSullivan/DeepFakeTorch.git
cd DeepFakeTorch
pip install -r requirements.txt
youtube video https://www.youtube.com/watch?v=XqluthtTenI
paper written by deepfacelab https://arxiv.org/abs/2005.05535
python face_detect.py -video_src data_src.mp4 -out_name a
python face_detect.py -video_src data_dst.mp4 -out_name b
python train.py -face_a_dir a -face_b_dir b -n_steps 100000
python video_writer.py -original_video [video name] -model_location saved_models/model.pt -out_name myswappedvideo -decoder [either 'a' or 'b']
- Smooth images on swapped video
- Add colour correction to swapped faces