Since breaking into the mainstream in late 2017, deepfakes have gained increasing notoriety, thanks in part to malicious use hitting headlines and a collective uncertainty of how their use will influence consumers day-to-day.
The term ‘deepfake’ has become more commonly used, being applied to any number of image manipulation techniques and further clouding the discussion. A true deepfake is created by using sophisticated artificial intelligence methods to produce an image, video, or audio clip of a person. These deepfakes are meant to be convincing enough to not only fool your average person, but also trick software specifically designed to detect them in the first place. However, while the technology continues to rapidly improve it is still in its relative infancy. Because of this, we should explore beyond the immediate negative and look to how deepfakes can be used to create positive experiences for everyday people.
A Storied Past
Naturally when the topic of deepfakes is mentioned, the first things to come to mind are impersonation or fraud. Thanks to this technology, from professionals and amateurs alike, we’ve now witnessed everything from President Barack Obama voicing support of comic book villains to a baby Elon Musk dressed as Yoda, not to mention the more unsavory applications that originally gave deepfakes their notoriety.
However, what often flies beneath the radar is the creative ways in which the technology is being leveraged to create media. Last year, IPG’s R/GA and Ridley Scott Associates partnered with A.I. startup Synthesia to showcase David Beckham speaking nine languages in their Malaria Must Die campaign. While this particular campaign used multiple voices to emphasize its message, the technology has also been used to replicate human speech, allowing us to hear a never-before-heard speech by John F. Kennedy.
A Potentially Bright Future
As more companies begin to explore the potential of this new technology, we can expect to see it applied with varying degrees of success in the advertising industry.
Above we touched on examples of a man speaking through other’s voices and hearing a speech that was never spoken aloud. In the future, we can expect that deepfakes will become sophisticated enough to replicate one’s own voice into numerous languages, possibly dialects, and then manipulate the video for a seamless experience.
That isn’t to say there won’t be challenges and questions along the way, as new technologies often lead us to ask not only “can we?” but “should we?” Key to how the technology works is learning from recordings of the subject, so if you want to produce a video of David Beckham speaking in Mandarin it helps to have a lot of past recordings to work with. However, as the technology advances, it’s not only believable but can be expected that it could not only learn the voice patterns of one person, but also those common in any regional dialect.
Ultra-Personalize and Immersive Ads
For years the advertising world has been moving further towards personalizing ads for the individual, and deepfake technology is another potential tool for achieving the task. In recent years, a flurry of apps such as FaceApp have created highly realistic transformations of human faces, from swapping faces with friends to catching a glimpse of themselves in their twilight years. This same technology is already moving into the consumer-focused world, allowing people to ‘try on’ clothes at online retailers before making their purchase.
Even taking this a step further are those who wish to put you, or at least your face, in their advertisements. While this is an interesting approach to explore, it raises questions of how consumers will react to seeing themselves in advertisements and, at least for myself, it may lead to a feeling of apprehension.
Complete Dynamic Creative Optimization (DCO)
Dynamic Creative Optimization is not a new concept, and for years has been used in our industry with varying degrees of complexity in content configuration, targeting, and measurement. Despite major advancements, there remains a limitation on the possible creatives based on the preconfigured assets a human is able to input.
While this may be far off, deepfake technology can be trained to generate every asset in the advertisement in a convincing manner. An infinite number of people, pets, or even art can be combined to tailor the image to the specific individual and even guarantee the same person never sees the same ad twice.
Future Potential, But Still Present Concerns
As with many new technologies, deepfakes can be used with malicious intent, however we’re also already seeing their potential to revolutionize all aspects of media. Doors previously closed due to practicality or cost can now be opened, leading to new and exciting opportunities. We’re optimistic that some of these positive applications of deepfake technology could create bespoke, beneficial experiences for consumers as the technology continues to evolve.