Deepfakes' entire story, from figures to creation...

S Venkateshwari
Deepfakes' entire story, from figures to creation...

In order to produce a deepfake, the concerned person's photo must first be chosen. That is to say, you must first choose the subject and scenario you wish to deepfake. A subsequent phase then begins.

Before using the AI tool to produce deepfakes, an AI system known as Encoder determines which celebrity's data-face will match it the closest. Experts state that an encoder's task is to compare two data-faces for similarities. The decoder AI algorithm is then applied. The individual whose deepfake is to be created is discovered using the decoder. After that, a deepfake image or video is produced using a data-face swap.

To further understand this, consider the following example: In order to create a deepfake of a person named A, the individual's entire profile must first be extracted. The encoder then determines where the picture should fit. Once the encoder has made this decision, the decoder uses a facial swipe to produce a deepfake image or video. Both person A and person B could lose in this situation.

The Guardian claims that the first known instance of a deepfake occurred in 2017. Reddit users at the time were posting celebrity footage. Among these superstars, Scarlett Johansson, Taylor Swift, and Gal Gadot were well-known names. Following this, the deepfake market grew steadily. A request to produce 15,000 deepfakes was received by AI startup DeepTrace in september 2019. Of these, cases involving obscenity accounted for 96%.

Find Out More:

Related Articles: