well, this mix of donaderump using the same tool, a tool called deepfa kes.o see how easy it is, we're going to do it. we're going to take my face and make me president. we trained a neural network by feeding it video of some of my past appearances. we mixed it with president trump's state of the union address. the software broke the video into individualframes, ran them through the network and, in less than a day, this was the result. all of us, together, as one team... so, this is the original video of trump. and this is me, on his head. we all share the same home. i'm not sure it's an improvement, but that does seem to be president spenley trump. the other half of the experiment didn't go quite so well. this is click presenter donald kelly. now, this was a very short and quick experiment. it's far from perfect. it's blurry, you can see the edges — and sometimes, well, it's just downright scary. but had we left the network to train for longer, on better videos, we could have got much more convincing results. now, it doesn't take much imagination to see how