Motion-designer from Germany united Putin and Trump using neural networks

It turned out “Mr. Trumputin.”

Motion designer and 3D artist Alexander Kurt (Alexander Court) combined Putin and Trump in one person with the help of neural networks. For this, he used the DeepFaceLab program and carried out post-processing using Adobe After Effects.

The result was realistic: there are no obvious artifacts in the video, as is the case with many other dipfaces. In order to train the algorithm, Kurt spent 240 thousand epochs, which means that the neural network learned 240,000 times on one data set.

As a basis for the video, the designer used a 26-second excerpt from Trump’s speech, where he talks about the rigidity of the European Union in relation to the United States. Kurt explained that he made the video as an experiment, not for a political statement.

It is not the first time when a motion designer makes high-quality videos with overlaying faces with neural networks. To do this, he launched a separate project Tiefenschwindel .

Among other examples of his work is the imposition of Mark Zuckerberg’s face on Jesse Eisenberg in the film “Social Network” and “Rejuvenation” by Leonardo DiCaprio.

Back to top button