03 Aug 2022
As cinema predicted, the robots are taking over!
Perhaps not the entire world just yet, but they spring up in all areas of entertainment. Artificial intelligence is rife, not just on the internet but in television shows, music videos, and on the silver screen. And it’s all about deepfake technology.
Deepfake technology is changing the way that artificial intelligence can be used to create realistic moments without real-life actors.
But can such artificial intelligence become the next big thing in entertainment when it’s so fraught with controversy? Let’s take a look.
How does deepfake work?
Deepfake is a type of image manipulation that uses machine learning techniques to alter facial features or expressions. It can also be used to mimic voices.
Its origins can be traced to the work of William Goodfellow, who looked into generative adversarial networks (GANs). These generate realistic images using an algorithm whereby two GANs try to trick each other into thinking that the image is real.
This all sounds rather complicated, but deepfake technology has come a long way these days and can be practiced by any adept digital artist. A GAN can already recreate a realistic human face from one single image, with the propensity to ‘rebuild’ an uncanny likeness of a whole human being in the future.
How AI enhances performance
Deepfake has gained worldwide fame in many forms of artistic expression. Most notably, a deepfake artist superimposed faces of actors including Tom Cruise and Seth Rogen in an interview with comedian Bill Hader to enhance his impressions. Rapper Kendrick Lamar used deepfake in a 2022 music video to ‘become’ influential celebrities conveying the meanings of his lyrical content. Elsewhere, the revitalised Star Wars franchise very cleverly implemented deepfake to bring back characters such as Princess Leia and Moff Tarkin despite their respective actors having passed away. The technique can similarly ‘age’ faces, giving a whole new dimension to realistic storytelling through cinema.
Elsewhere, deepfake technology has been utilised by a malaria charity to alter David Beckham’s voice to speak nine languages, and has made it easy to generate AI avatars without the need for actors during lockdowns in the Covid-19 pandemic.
The problems posed by deepfake tech
Then again, there are problems that deepfake technology can cause. Taking the ‘fake’ term very literally, the images are not real but can easily be construed as such, often for nefarious means.
Take for instance the rising cases of ‘fake news’ that promote unreal events often through social media; deepfake can be used to place real people’s faces into pictures to tell a make-believe news story. Politicians have found themselves mimicked through the technology – potential voters could easily be duped. There are also cases of identity misuse when deepfaked faces of celebrities are put into pornographic contexts. This has extended into artificially generated nude models of real people, done without their permission.
The technology can be used for identity fraud through facial recognition or voice activation services. The viral enthusiasm of the technology, and the means by which it is spread, makes it tough for social media networks to regulate.
Deepfake vs mocap: what is the future of film?
Deepfake could be mistaken for a type of motion capture technology, but this is not completely true. It remains a form of artificial intelligence or machine learning that creates manipulated images and voices. Motion capture technology instead is a visual effect technology: it uses the real-time movements, facial expressions and voices of an actor with the ability to transpose computer generated imagery (CGI) onto them.
Using the examples of de-aging or regenerating late actors, deepfakes look uncanny as they are not able to track actual movement. Mocap on the other hand is advanced in its capturing of an actor’s movements and expressions, and uses that data to animate CGI with realistic precision.
Perhaps there will be a time when deepfake’s cinematic advantages can be used in conjunction with visual effects and motion capture. Mocap can be used as the foundation of by recording and analysing an actor’s movement in performance, with help from deepfake technology to recreate faces or age actors for continuity efforts.
Deepfake is certainly useful as an artistic endeavour, with an unfortunate array of misuse. While there are ways to use GANs to detect fake images and lower its more negative connotations, nothing can replicate the real; deepfakes are currently far from creating a more ideal image that can be gained from mocap.
At Motion Analysis, we think mocap remains the future of the film industry to create unforgettable animated experiences. Explore our Cortex software to find out why.