To create a DeepFake video you need a lot of pictures of the victim, which will be analyzed through GAN algorithms and superimposed on video images. There is also the possibility of placing a text that will be played by the person who appears in the video images.
The greatest danger of DeepFake technologies is that the tools that allow it to be created have become accessible to all. To work with these applications you need a modern computer, but this is not an impediment nowadays.
The main tool for creating a DeepFake is the FakeApp application. The original website of the application http://www.fakeapp.org/ has been blocked. However, there are still a number of mirror sites, and on Reddit we find a whole category with links to the application with all the necessary instructions.
Here are some examples of using FakeApp with former US President Barack Obama:
And a fake with Mark Zuckerberg:
The journalist’s experiment was also interesting in that both the author, a Reddit user and the developer of the FakeApp application, who commented on the work, remained anonymous. They did not give their real names and communicated through nicknames and impersonal email addresses.
New York Times journalist Kevin Roose tested the FakeApp app and talked about his experience. To create the video, he used a remote server rented through the Google Cloud platform, which provided sufficient computing power. Even under these conditions, the server needed over eight hours to generate the models. The server lease cost $ 85.96. For this operation an ordinary laptop would need a few days or even weeks. In the experience of creating a DeepFake, over 400 photos of the journalist and over one thousand photos of actor Ryan Gosling were used, because they look very similar. The video images resulting from the experiment were quite unclear.
Source: www.stopfake.org