This short film may not feature internationally recognized A-list movie stars.
But with the help of machine learning AI algorithms, our actors may seem a bit familiar.
A cautionary science fiction satire, FACE SWAP is arguably the first narrative film using the “deepfakes” process as a storytelling tool.
A guy convinces his wife to try out a new AI technology to spice up their sex life, but gets a bit more spice than he bargained for.
FACE SWAP takes a satirical look at the imminent new reality ushered in by the advent of face swapping apps known as "Deepfakes".
It poses the question: In today's world, do our faces still belong to us?
Inspired by the broader ramifications of newly accessible machine learning algorithms and face recognition AI, we set out to create a parody portraying one way things could evolve.
In an age in which the reliance on our eyes and ears for the truth has become exponentially more fragile, we hope our film serves as a social commentary on the looming implications of such technologies - the good, the bad and the ugly (or beautiful, in the case of our face swappers).
ABOUT THE TECHNOLOGY
In late 2017, someone started a Reddit thread titled /r/deepfakes/ and invited the GitHub coding community to join a new software development project, called deepfakes/faceswap. This actively developing software uses a popular processing workflow known as a "deep neural network" that, through repetitive optimization, gets better at predicting how an individual person's facial patterns would change were they a different person.
This is how long it actually takes my computer to convert an entire scene:
Thanks to recent advances in computer hardware by manufacturers like NVIDIA and INTEL, Machine Learning and Deep Neural Networks have graduated from theoretic to practical applications. The deepfakes/faceswap code uses this new technology, made even more accessible by the Tensor-flow framework originally developed by researchers and engineers from the Google Brain Team.