Modes of Operation

Generating believable fake content requires taking care of multiple aspects to ensure that the results are as authentic as possible. A typical deepfake setup requires a source, a target, and the generated content.

• The source, denoted with subscript $s$, is the driver identity that controls the required output.

• The target, denoted with subscript $t$, is the identity being faked.

• The generated content, denoted with subscript $g$, is the result following the transformation of the source to the target.

Now that we have some basic terminology, let’s explore the different ways to generate deepfake content.

Disclaimer: This demonstration uses deepfake technology to show how images and videos can be manipulated using AI. It is intended purely for educational purposes to explore the capabilities of the technology, not to deceive or mislead. Ethical considerations are crucial when using this technology, and unauthorized use without consent is not supported.

Replacement

This is the most widely used form of generating fake content. The aim is to replace the specific content of the target ($x_t$) with that from the source ($x_s$). Face replacementDepFA. 2022. “Dfaker/Df.” GitHub. March 24, 2022. https://github.com/dfaker/df. has been an active area of research for quite some time now. The image below shows Donald Trump’s face being replaced with Nicolas Cage’s. The figure displays both source ($x_s$) and target ($x_t$) identities, while the generated content ($x_g$) is shown in the last column: