According to the programmer, his version of the algorithm modifies not only the face, but the entire "person" as a whole: you only need one source photo, target image, five approximate parameters of complexion and "0.1 second of patience" of the user.
After the stage of 3D-fication (3Dfying) of the original photo (creating a three-dimensional version of a two-dimensional image), the output is a completely modified "person", head, hair, skin color, as well as complexion, which repeat those in the original photo.
Unlike the old Deepfake, the new version changes not only the face, but also the head, hair, skin color and even the figure. Moreover, the technology can change any photo, and not pre-built into the algorithm.
According to Chaplygin, he developed this technology alone for two years, and it was "very difficult to create something that no one else has been able to do before, especially if you work alone." At the moment, the program is running on his "home computer", and in order to scale the project, the developer needs "servers, high-speed Internet and a team that supports the entire setup." Chaplygin plans to find funds for the mass introduction of the new technology from interested persons or organizations using the “six handshakes” method.
According to the developer, such "deepfakes" can be used for "identity theft / fake", but, according to Chaplygin, Deepfake 2.0 was created for "good purposes", one of which is to create a virtual fitting room for online clothing stores, where everyone can upload your photo to evaluate a future purchase.
Deepfake 2.0 can be embedded as a plug-in (subroutine) into any software or device, so that the user can instantly see himself in any photo. Also, the technology can be used in photo editors, filmmaking, when creating animation, etc.
The quality of the new algorithm can be assessed on a demonstration video posted on Alexey Chaplygin's YouTube channel.
