Now I'm working on Ciara rtm model that will be able to work in real time without any training, currently I have 400k iterations, but realistically it will probably take another month of training to get decent quality
https://github.com/iperov/DeepFaceLive