Navigating the Challenges of Creating Realistic CGI Humans in Film and TV

Film and television have long leveraged the power of computer-generated imagery (CGI) to bring visions to life on screen. However, creating realistic-looking CGI humans has proven to be a daunting challenge. This article delves into the intricacies of why human faces are so difficult to capture accurately with CGI, the phenomenon known as the "uncanny valley," and the future of animation technology.

The Human Brain's Expertise in Facial Recognition

Humans are inherently social creatures, and our brain is exceptionally adept at recognizing human faces and mannerisms. The level of discrimination needed for such recognition is remarkable, with the ability to detect even minute differences in facial geometry, which can vary between individuals by as little as millimeters. This high degree of discrimination served humans well for millennia, allowing us to distinguish friend from foe even in obscured views, such as through mud or smog.

The Uncanny Valley: A Hurdle for CGI

As society advances technologically, we've created sophisticated tools that enable us to achieve high levels of photo realism in CGI. Yet, achieving true realism presents a challenge due to the human brain's exquisite facial recognition capabilities. Any deviation, no matter how small, can trigger unease or discomfort, often referred to as the "uncanny valley." This term describes the phenomenon where a digital human or character closely resembles a real person but is slightly off, prompting a sense of revulsion or discomfort. This discomfort arises because the brain is being confused in ways it has not evolved to handle. It's similar to the experience of motion sickness, where the brain receives conflicting signals and cannot process them correctly. In some ways, the uncanny valley is a product of our biological limitations.

Blade Runner 2049: A Case Study in Realistic CGI

One notable example of the challenges and potential solutions is the de-aged performance of actress Sean Young in Blade Runner 2049. The entire process took an entire VFX company six weeks to perfect a mere minute of screen time. The meticulous work included tracking skin pores to ensure the character's appearance was as lifelike as possible. This level of detail is crucial in making the character believable and memorable. However, creating a still frame that looks like a real person is one thing. Creating an animated character that can move and act like a real person is significantly more challenging.

Difficulties in Matching Human Behavior

Even when a digital human's appearance is nearly perfect, it may still fall short of realism because of the complex behavior associated with human interaction. The ability to capture and replicate the subtle nuances of human behavior, such as facial expressions, body language, and vocal inflections, is incredibly difficult. Each of these elements combines to create a convincing portrayal of a real person. Still frames require less complexity to look authentic, but as soon as they start to move and speak, the distance to perfect realism becomes vast. The recreation of Paul Walker after his passing in Furious 7 is a remarkable achievement, but it is still not at the level where the human audience would not be able to distinguish between the actor and the digital character without careful attention.

Solving the Uncanny Valley Through Realism and Contract with Audiences

One strategy to mitigate the challenges of the uncanny valley is to establish a clear line between digital characters and real actors. Characters that do not have to live up to the standard of absolute verisimilitude often succeed in feeling more realistic. For example, characters like the Navi in Avatar, Gollum, or Baby Yoda are successful precisely because the audience understands they are not meant to be real. The audience can enjoy these characters as characters, rather than as facsimiles of actors playing roles.

The Future of CGI: Deep Fake and Traditional Techniques Combined

Advancements in technology, particularly deep fake technology, may help us traverse the uncanny valley. By combining deep fake techniques with traditional CGI, we can achieve more realistic portrayals of human characters. While we may never fully replicate the complexity of human life and behavior, continuing improvements in technology will undoubtedly push the boundaries of what is possible.

As we move towards a future where digital characters become increasingly lifelike, it is crucial to understand and navigate the challenges presented by the human brain's profound capacity for facial recognition and the emotions that arise from digital representations that do not quite fit the bill.