Allows to train Deep Neural Network to reenact face from source to destination in an image or a video. Also allows for face recongition and extraction. This is based on widely available framework of recognizing and learning face/objects using multi-layer neural network. This particular approach introduces new model and allows for dynamic vram growth. Was tested on Nvidia Titan V(12GB) and Intel I9-7980XE. You can adapt this scripts to run in linux. Tested on Ubuntu. Speed of neural network training is faster on Ubuntu (18.10) vs Windows 10 (27% faster). Same machine. More headache to configure (Docker approach recommended). You will need tensorflow and keras, nvidia cudnn, python installed in anaconda(or whatever else you using). This can be used with google colab for those who a thrifty with electricity bills..