Regarding 1) and following 2) you only need to make realistic-enough video that it passes as a person and is similar enough to the target person. For example, y

Regarding 1) and following 2) you only need to make realistic-enough video that it passes as a person and is similar enough to the target person. For example, you can have a pretrained model (e.g., using zero targeted data) and search for some configuration that is closest to the target person. You only need to match a few variables (e.g., ethnicity, gender, hair color, age) before the fake is plausible.