A YouTube channel Ctrl Shift Face demonstrates how far the DeepFake technology has developed in a short time. In the beginning it was mainly pornos that were the goal, in which the faces of the actresses were exchanged for the well-known actresses with more or less convincing results. DeepFake is used to rewrite cinema history and famous films can be cast as desired - Sylvester Stallone is then the Terminator, Eward Norton becomes Brad Pitt at the end of Fight Club or Heath Ledger&s Joker plays the hero in A Knight&s Tale.
Sylvester Stallone as Terminator
How far the technology has developed can also be seen in how well a complex incidence of light, including shadows, is now simulated on the replaced face. In some of the clips Ctrl Shift Face lets the faces morph imperceptibly into each other without the viewer being able to tell exactly when one face ends and when the other begins.
DeepFakes - a conglomerate of Deep Learning and Fake - use Deep Learning technology/neural networks, which are trained using sample material from one face to trade it for another. Theoretically, anyone with enough expertise, relatively powerful hardware (a PC for gamers or 4K video editing, for example, is enough) and enough source video of the face to be exchanged can create such fakes.
If this expertise is poured into tools, then no special expertise is needed - then really anyone can easily produce DeepFakes from anyone. The videos of Ctrl Shift Face show what an amateur can do now.
It is only a matter of time until the faces can be exchanged almost perfectly and the voice can also be simulated appropriately - the voice simulation is also making great progress at the moment via machine learning. From now on, the conclusiveness of videos can only be trusted after a thorough analysis or if their origin is completely certain.