Disney Research has proposed an algorithm that enables automatic neural face swapping in high-resolution images and videos.
“To the best of our knowledge, this is the first method capable of rendering photo-realistic and temporally coherent results at megapixel resolution,” Disney Research stated.
Disney Research’s solution opens up a wide range of potential uses in the TV and film industry, such as portraying a character at a younger age, putting an actor’s face onto a stunt double, or even using an actor that is not available “or is perhaps even long deceased”. This technology could also be used to create convincing deep-fake videos.
Previously, face-swapping solutions in the film industry have typically been labour-intensive, using computer graphics methods.
A key component of Disney Research’s method is increasing the number of identities presented to the network, to achieve higher fidelity of the swapped expressions. The full Disney Research ‘High-Resolution Neural Face Swapping for Visual Effects’ whitepaper can be read here.