How hard is it to make a deepfake?

Creating a deepfake can vary in difficulty depending on the resources available and the level of sophistication desired in the final product.

The basic process of creating a deepfake involves the use of machine learning algorithms to swap faces in a video. This is typically done using a method known as autoencoding.

In the initial stages, you would need a large dataset of images of the person you wish to 'deepfake'. This is to train the AI. The larger and more varied the dataset, the more convincing the final result will be.

Once a sufficient dataset is collected, the autoencoder learns how to recreate the person's face from these images. At the same time, it learns the face of the person in the original video.

The autoencoder then swaps the faces, creating the deepfake.

While this process may sound straightforward, it involves a significant understanding of machine learning and computer vision techniques. Additionally, it requires access to high-performance computing resources, as the training process can take a significant amount of time and computational power.

However, with the rise of deepfake applications and software, the technical barrier is lowering. Some applications now provide user-friendly interfaces that guide users through the process with minimal technical knowledge required. But, creating a highly realistic and convincing deepfake still remains a complex task.

It's also important to note that while creating deepfakes isn't illegal, distributing them may be subject to legal restrictions in some states. Misuse of deepfakes can lead to serious legal and ethical issues.