This was not what they expected ‘Her Majesty’ to say in a televised Christmas speech, but she did and did it with a twist. Channel 4 digitally created a fake of the Queen to warn viewers to question “whether what we see and hear is always what it seems.”
The ‘real’ British Queen traditionally delivers a Christmas Day speech which is aired around the world. Every year, Channel 4 follows the Queen’s speech with an “alternative Christmas message.” Former Iran president Mahmoud Ahmadinejad, US whistleblower Edward Snowden, Jesse Jackson and many more have delivered the “alternative Christmas message” earlier. This year, however, Channel 4 hired VFX studio ‘Framestore’ to create a fake of the Queen, who opened up her personal life on TV.
The Queen is shown joking about Harry and Meghan, Canadians and Prince Andrew in the ‘Christmas message’, laced with typical British humour.
A digitally-created Artificial Intelligence (AI) ‘deepfake’ version of the Queen, played by actress Debra Stephenson, offered a stark warning about misinformation and fake news and how the problem needs to be addressed now.
The Queen then goes on to take a dig at Prime Minister Boris Johnson. She says: “2020 has also been a year of heroes, such as our brave NHS workers, many of whom were forced to take incredible risks – like treating Boris Johnson – knowing at any time, as a result, they too could become pregnant.”
The fake Queen also jokes about the toilet paper crisis that unfolded during Covid-19. She went on to do a TikTok dance video before becoming all serious and advising people to be wary of what they see online and on TV, ‘because not everything may be as it appears’.
The fake has already invited an outrage across the UK with a section of citizens calling it “distasteful and disrespectful”.
Also read: Dangerous deepfakes blur line between real and sham
The makers of the 3 minute 45 second video claimed the fake was actually a warning. “Deepfake technology is the frightening new frontier in the battle between misinformation and truth,” Channel 4 Director of Programmes Ian Katz said in a statement.
An internet user said: “It’s fine to expose people to deepfakes, but we shouldn’t be escalating the rhetoric to claim we’re surrounded by them.”
Several data science experts say this particular deep fake isn’t exactly top quality. “They could have done it better,” said a leading data scientist with an American Animation company.
What is Deepfake?
Deep fakes are fake videos or audio recordings that look and sound just like the real thing.
The main ‘actor’ in any deep fake video is actually machine learning (ML), which has made it possible to produce the fakes much faster and at a lower cost. To make a deep fake of someone, the video producer would first need to “train” a neural network on many hours of real video footage of the person to give it a realistic “understanding” of what he or she looks like from different angles and under different lighting. The software then combines these “learnings” with the computer-graphics techniques to superimpose a copy of the person onto a different actor.
The addition of Artificial Intelligence (AI) or ML makes this process faster than it ever could have been. However, it still takes time for this process to yield a realistic composite that places a person into an entirely fictional situation. The ML programmer must also manually tweak many of the trained program’s parameters to avoid bloopers.
How to build your own Deepfake video?
There is a method to this madness of making a face-swap video.
First, the data scientists collect information on thousands of face shots of the two people through an AI algorithm called an encoder. The encoder finds and learns similarities between the two faces, and reduces them to their ‘GCD’ (greatest common denominator) of shared common human features. In this process the images are deeply compressed. A second Machine Learning (AI) algorithm which is called a decoder is then taught to recover the faces from the compressed data. Since the faces are essentially different, the engineer “trains” one decoder to recover the first person’s face, and another decoder to recover the second person’s face.
To perform the swap, the ML engineer merely feeds encoded images into the “wrong” decoder and you get the deep fake video – be it Putin on Trump or Mark Zuckerberg saying “he controls the world”. Her Majesty, the Queen of England, is the newest addition to the list.