The prophetic Season 6 opener of Black Mirror, Joan is Awful, propels viewers into a mirrorworld where every actor has been replaced with an AI likeness. Within months of the episode’s release, the fact betrayed by this fiction became a catalyst for a real world Hollywood strike. The actors’ union is protesting a contract clause they say would give big studios the right to use background actors’ digital doppelgangers for free, forever.
From dramatic examples like these to simpler cases like AI-generated articles and emails, it’s clear: Generative AI is remaking media as it’s remaking marketing. And some innovative marketers are already looking beyond the controversy and unlocking new creative possibilities with synthetic talent. In the process, they’re also uncovering some of the challenges and limitations of engaging real consumers with fake celebrities.
In this article, we’ll look at three examples of big brand campaigns that tap into the creative potential of synthetic media. But first, let’s cover the basics.
What is Synthetic Media?
In the broadest sense, synthetic media refers to any use of artificial intelligence to generate or manipulate digital content, such as images, videos, texts, or sounds, so that they appear to be authentic. By this definition, the article your content marketer generated in ChatGPT or Jasper fits the bill.
Here though — as we’re exploring the idea of digital spokespeople — we’re referring specifically to using deep learning neural networks to generate a convincing, photorealistic audio/visual likeness of a specific person, saying and doing things that they didn’t actually say or do. In colloquial terms, we’re talking about deepfakes.
The term deepfake itself carries an aura of malicious intent or — at a minimum — mischief. You likely heard the term when Mijourney images of Pope Francis wearing a Balenciaga puffer coat or Donald Trump’s NYC arrest went viral, earlier this year. Indeed, deepfake technology has been used in everything from online pornography to political misinformation campaigns. But like most innovations, the technology itself is neither good nor bad. Agency holding company WPP has used deepfake technology from one company to personlized employee training. Other companies — like Metaphysic.ai — target the high-end market with sophisticated celebrity recreations for big budget productions.
In one sense, AI-powered synthetic media is an evolution of the virtual influencer trend at play for several years. In my 2019 keynote at BrandManagecamp, I spoke about Lil Miquela — “one of a handful of inarguably artificial influencers, with millions of social media followers and a stable of lucrative brand sponsorships.” And I predicted “a future in which your brand may have its own customized, (eventually) AI-driven digital ambassador that represents your ideal customer and connects with your prospective customers in ways you wish today’s human influencers could.”
Today, given rapid advances in generative AI, brands are beginning to explore the potential of synthetic celebrity spokespeople to achieve previously impossible forms of creative innovation. Now, let’s take a look at three campaigns.
Virgin Voyages Jen AI
Virgin’s cruise line pairs superstar spokesperson Jennifer Lopez with generative AI techniques. A TV spot introduces Jen AI (get it?), a virtual spokesperson that “malfunctions” to expose the actors behind the avatar. It’s a clever spoof of deepfake technology that pokes fun at the current GenAI hype cycle while setting up the call-to-action: the viewer can “control” Jen AI too, and create a personalized cruise vacation promo video.
My sense is that the spot itself uses a living, breathing J Lo and some good old fashioned CGI rather than AI to produce its Jen AI effects. On the website that viewers use to create their custom promos, a synthetic J Lo is brought to life with an AI-powered video avatar created by Deeplocal and the star’s voice cloned by SpeakUnique.
The implementation is rudimentary (although I imagine it was neither simple nor cheap), allowing customization via just a few pointed text prompts and within the limits of tight constraints. The constraints are by no means all negative. In fact, they point to a sound understanding that generative AI is a nascent technology with inherent risks. Particularly for big brands working with star talent, its vital to bake strong guardrails into the brief. Jen AI won’t speak inappropriate language and the generated videos can’t be posted online or shared beyond the intended recipients.
A CNBC piece from late July offers a behind-the-scenes look at the campaign (conceived by VMLY&R), and reports that (at that time) the work had generated around 1,000 new bookings for the travel brand.
Cadbury “NotJustACadburyAd”
The Mondelez-owned confectionary brand combined generative AI, a holiday promotion, and corporate social responsibility in a Diwali campaign designed to sell chocolate and drive shoppers to local mom-and-pop stores. “NotJustACadburyAd” allows local purveyors to create store-specific versions of a video spot featuring Bollywood star Shah Rukh Khan. The campaign uses synthetic media technology from Rephrase.ai to recreate Khan’s face and voice, promoting each store by name to encourage Diwali holiday shopping.
This campaign wrap-up video explains the strategy and execution.
According to a statistic on Rephrase’s website, Cadbury saw its biggest Diwali sales spike in a decade. But what set this campaign apart was the way using generative AI and synthetic media allowed small shops (many of whom were still struggling financially in post-COVID India) to reap the benefits of big budget ad production and the power of celebrity endorsement with nothing more than a prompt.
This said, some consumers took issue with the campaign. One Twitter users called the ad “problematic,” and worried that it highlighted the availability and ease of use for a technology that could be used by bad actors to generate discriminatory, malicious deepfakes. While pushback was minor in this case, this does highlight the need for brands and agencies to anticipate ethical objections, have a plan to mitigate any reputational harms that may arise, and even consider playing an active role in educating consumers about appropriate vs. inappropriate usage of generative AI tools.
Volkswagen do Brasil Resurrects “Elis Regina”
In a Brazilian video ad marking the automaker’s 70th anniversary, musical icon Elis Regina and her adult daughter Maria Rita (a Grammy-winner herself) sing a soul-stirring duet while driving alongside one another in VW vans. The catch? Elis Regina passed away in 1982 at the age of 36. Her likeness is AI-generated, using facial recognition to map her features to the movements of a background actor. The effect is utterly convincing and shows just how good synthetic media can be, even at this relatively early stage of the game.
Millions of consumers who remembered Elis or knew of her by reputation found themselves moved by the emotional video. But others complained that the spot raises significant ethical concerns about AI, its use in entertainment, and its potential societal impact.
While Volkswagen had the family’s permission and participation for the production, the question remains: Under what circumstances is it right to bring a deceased celebrity back to life in a media production through artificial intelligence? To further complicate the matter, Elis staunchly opposed the military dictatorship that governed Brazil at the time of her death — a regime Volkswagen notoriously supported. It’s unclear whether the singer herself would have agreed to this particular brand collaboration if she were alive today.
In the age of AI immortality, who owns your digital likeness? Is resurrecting a late celebrity through AI any more problematic than featuring them through archival footage? Any more objectionable than representing them with a look-alike impersonator?
These may sound like philosophical questions. On one level they are, of course. But they’re also practical considerations. Brazil’s ad industry watchdog, Conar, is investigating VW for a possible breach of ethics after receiving complaints about the campaign. According to The Guardian, Conar is looking into “whether [the use of such techniques] might cause some to confuse fiction with reality, above all children and teenagers.” Clearly, it’s incumbent on brands to take their guidance from a strong ethical AI north star. It’s just as clear that considerations around reputational and regulatory risk must feature prominently in every creative decision — especially when marketers are just beginning to think through the right strategies, use cases, applications, and guardrails for generative AI.
A Real Opportunity for Responsible Synthetic Media
Generative AI is a rapidly developing technology that has the potential to change media and marketing. Creatives can now generate synthetic media that looks convincingly real. Companies like Virgin Voyages, Cadbury, and Volkswagen Brazil have used this technology to create successful campaigns — in the process proving that synthetic media provides a powerful vehicle for marketing innovation.
However, brands need to be aware of the risks of using this technology; must engage in active upfront conversations about unforeseen implications; and should create guardrails and governance to mitigate reputational harms and regulatory risks. Despite concerns, generative AI can fuel creative innovation and create competitive advantage for brands that are strategic and ethical in its application.
