Branding Out of Thin Air

"Synthetic media" is poised to transform marketing and communications

By Chuck Kapelke

In an ad for Malaria Must Die, an aged version of soccer star David Beckham announces that "we have ended malaria." The video then morphs into the present-day Beckham, who reminds viewers that the fight against the disease is far from over. The "deepfake" technology, or synthetic media, used to create that spot holds both promise and peril for marketers. Malaria Must Die/YouTube

In a video released as part for Malaria Must Die, a global campaign to eradicate malaria, soccer star David Beckham, aged with gray hair, beard, and wrinkles, announces that "we have ended malaria." The video then morphs into Beckham as he appears today, reminding viewers that the fight against malaria goes on. The spot was created with "deepfake" technology, merging Beckham's face with a digital creation of his older self, with stunning realism. Sans makeup or motion capture technologies, the production required just 45 minutes of Beckham's time. "We let machine learning do a lot of the heavy lifting to get us into that happy medium area," says Kevin Lau, executive creative director of advertising, games, and new media at Digital Domain, which created the effect. "There are a lot of ways in which this technology can be utilized. It's a new tool in our toolkit for creating images and stories."

Deepfake videos are just one example of "synthetic media." No longer nascent, synthetic media is a process in which machine learning systems and generative adversarial networks (GANs) are "trained" to create new content based on prior examples. These technologies have advanced rapidly in recent years, spawning a range of applications that could have profound implications for marketers.

"Generative AI can learn from existing artifacts to create new artifacts that reflect the characteristics of the training data, but don't repeat them," says Jason McNellis, senior director and analyst at Gartner and author of the report, "CMOs: Deliver on the Promise of AI with Deep Learning." "It can produce a variety of novel content, such as images, videos, speech, tech, software code, and product designs."

Deepfake videos can be used to put words into other people's mouths — or their own mouths, as when Sonantic released a recording of an AI-generated voice modeled after the actor Val Kilmer, who lost his voice in 2015 following throat cancer surgery.

"It enables customization not just of how ads are delivered, but of the ads themselves, including the characters or the background," says Pattie Maes, a professor at the MIT Media Lab. "There are already sites that can generate virtual people, so instead of having to find a real human model who fits the bill, you can just pick the right mix of attributes you need."

"Right now, the fight is harder than ever," says David Beckham in an ad for a Malaria Must Die campaign spot titled "A World Without Malaria," referring to the battle against the disease. The video was created with Beckham and uses deepfake technology, or synthetic media, to show an aged version of the soccer star. The method leverages techniques from machine learning and AI to manipulate visual and audio content. Malaria Must Die/YouTube

For all its potential, however, synthetic media poses thorny ethical and legal questions for CMOs. Creators of a documentary about Anthony Bourdain were heavily criticized when the late epicure's voice was deepfaked to simulate the reading of an email he had written — without permission from his estate. In more extreme cases, deepfakes have been used for superimposing people's faces onto pornographic images and faking the voices of politicians and business leaders.

As synthetic media inexorably moves toward the mainstream, how should marketers respond? What are the potential ethical and legal ramifications? Will machines eventually replace ad creatives? Here are some takeaways to consider.

Second Life

In 2020, MyHeritage, a family genealogy service, scored a hit with Deep Nostalgia, an application that lets people upload pictures that "come to life" with motion rendered by deepfake technology. The app shot to No. 1 in app stores around the world, with 100 million images generated.

D-ID, whose technology powers the Deep Nostalgia app, has since enabled users to insert themselves into a trailer for the sci-fi thriller Reminiscence, and created an app for Mondelēz's mooncake brand, Kinh Do, that let users animate photos of Vietnam's Mid-Autumn Moon Festival. "If a brand wants to do something fresh and original and different, this technology is great for that," says Matthew Kershaw, VP of marketing and growth at D-ID.

Key takeaway: Deepfake videos are still a novelty, so early adopters have a chance to make a splash while gaining experience that could prove valuable down the road. "Make sure the share buttons are there so it can go viral," Kershaw says. "You also need to put a bit of spend behind it, whether through a PR stunt or a digital buy."

Localizing Content Quickly

By enabling real-time translation of copy paired with a synthetic voice, synthetic media could have a major effect on global brands. Media outlets like Sky News and Business Insider, for example, use Papercup, a service that deploys AI and synthetic voices to automatically dub English-language videos into seven other languages (with another 10 languages slated for 2022).

"Creators can localize their videos at scale and use automated dubbing in parallel with other localization to speak the language of their audience across all their content types," says Amir Jirbandey, head of growth at Papercup.

Key takeaway: Machine translation and dubbing are far from perfect — Papercup uses a "human-in-the-loop" system to verify for accuracy — but synthetic media levels the playing field for how brands can reach multiple audiences at very low cost.

Virtual Spokespeople?

Imagine being able to interact with characters like Johnnie Walker or Lucky the Leprechaun when there's no human controlling the character. Advances in synthetic media could make it possible for people to interact with AI-based characters that do their own "thinking."

As a proof of concept, Digital Domain has developed an autonomous digital human named Douglas, that nearly looks and sounds like a person and can hold conversations on its own. "You can plug an AI brain into a computer-generated character," Lau says. "It's still science fiction, but it's getting a lot closer to fact."

Key takeaway: As marketers begin to grapple with the metaverse, an immersive online world, how soon before virtual spokespeople materialize? "We expect that brands will be moving very strongly into these metaverse worlds, and everyone will be claiming space," Kershaw says.

The Acid Test

Marketers must ensure that data used to train their AI systems does not introduce bias, and that they stay abreast of regulations concerning "explainable AI," which requires developers to explain why their AI systems behave as they do.

Companies should get involved with industry efforts to set strong standards for transparency and disclosure of synthetic content. "The real risk is that the media landscape will become so swamped with AI-generated media that determining whether a message is real will become difficult," says Christopher Penn, co-founder and chief data scientist at Trust Insights Inc., which specializes in marketing and analytics. "A lot of it comes down to ethics and laws."

Key takeaway: Aside from playing a role in establishing industry standards, marketers also need to anticipate how AI-based systems could render their jobs obsolete. "Marketers need to up-level their skills and hone their craft," Penn says. "They need to constantly be training themselves through professional development to stay ahead of the machines."



You must be logged in to submit a comment.