Join us on November 9th to learn how to successfully innovate and achieve efficiencies by educating and scaling citizen developers at the Low-Code/No-Code Summit. Register here.
Natalie Monbiot, head of strategy at Hour One, a synthetic media company, doesn’t like the word “deepfakes.”
“Deepfake implies unauthorized use of synthetic media and generative artificial intelligence — we’re authorized from the start,” she told VentureBeat.
She described the Tel Aviv and New York-based Hour One as an AI company that has also “created a legal and ethical framework for how to work with real people to create their likeness in digital form.”
Authorized versus unauthorized. This is an important distinction at a time when deepfakes, or synthetic media in which one person in an existing image or video is replaced with another person’s likeness, has received a boatload of bad press — not surprising, given the longstanding association of Revenge porn deepfakes and fake news. The term “deepfake” can be traced back to a 2017 Reddit user named “Deepfakes” who, along with others in the community, shared videos, many of which contained celebrity faces swapped onto actresses’ bodies in pornographic videos became.
Learn how to easily build, scale, and manage low-code programs to create success for all on November 9th. Sign up for your free pass today.
And deepfake threats are looming, according to a recent research paper by Eric Horvitz, Microsoft’s Chief Science Officer. These include interactive deepfakes, which give the illusion of talking to a real person, and compositional deepfakes, where bad actors create many deepfakes to put together a “synthetic story”.
Recently, the news about celebrity deepfakes has been piling up. There’s Wall Street Journal coverage of deepfakes of Tom Cruise, Elon Musk, and Leonardo DiCaprio appearing in unauthorized ads, as well as rumors of Bruce Willis giving away the rights to his deepfake likeness (not true).
The business side of the deepfake debate
But there’s another side to the deepfake debate, say several vendors specializing in synthetic media technology. What about authorized deepfakes used for business video production?
Most use cases for deepfake videos are fully authorized. They can be in corporate environments – for example for employee training, education and e-commerce. Or they can be created by users, such as celebrities and corporate executives, who wish to take advantage of synthetic media to “offload” it to a virtual twin.
In these cases, the idea is to use synthetic media — in the form of virtual humans — to address the expensive, complex, and non-scalable challenges of traditional video production, especially at a time when the hunger for video content seems insatiable. Hour One, for example, claims to have made 100,000 videos in the last three and a half years, with clients like language learning guide Berlitz and media companies like NBC Universal and DreamWorks.
At a moment when generative AI has become part of the mainstream cultural zeitgeist, the future for enterprise use cases of deepfakes looks bright. Forrester recently released its top AI predictions for 2023, one of which says 10% of Fortune 500 companies will create content using AI tools. The report mentions startups like Hour One and Synthesia that are “using AI to accelerate video content generation.”
Another report predicts that up to 90% of digital media could be synthetically generated in the next five to seven years.
“That sounded very optimistic…probably even to me,” Monbiot said. “But when the technology matures and massive players enter this space, we see disruption.”
The business side is a “very underestimated” part of the deepfakes debate, points out Victor Riparbelli, CEO of London-based Synthesia, which bills itself as an “AI video creation company”. Founded in 2017, the company has more than 15,000 customers, a team of 135 employees and is “growing double digits every month”. His clients include fast-food giants like McDonald’s, research firm Teleperformance and global advertising holding company WPP.
“It’s very interesting how narrow the lens was with all the bad things you can do with this technology,” Riparbelli said. “I think what we’ve seen is just an increasing interest in this topic and more and more use cases.”
A living video that you can edit at any time
Accessing quality content is difficult, and most companies don’t have the skills to enable quality content creation, Monbiot said.
“Most companies don’t have employees who have skills that enable content creation, especially high-quality content creation with real talent, nor do they have the ability to edit videos or have those types of resources in-house.” She explained . Hour One is a no-code platform, so even users with no prior knowledge of content creation can choose from a range of virtual humans or become one themselves.
Berlitz, one of Hour One’s first corporate clients, had to make a digital transformation after 150 years of offering face-to-face classes. “To keep the instructor updated, they do live video conferencing, but that doesn’t really scale,” Monbiot said. “Even if they had all the production resources in the world, the cost and the investment and managing all those files is just insane.” She added that with AI, the content can be continually updated and refreshed. In the meantime, Berlitz has created over 20,000 videos in various languages with Hour One.
Meanwhile, Synthesia said its AI is trained on real actors. It provides the actors’ images and voices as virtual characters that clients can choose from to create training, learning, compliance and marketing videos. Actors are paid per video created with them.
For enterprise customers, this becomes a “living video” that they can access and edit at any time, Riparbelli explained.
“I think we now work for almost all of the biggest fast food chains in the world,” he said. “You have to train hundreds of thousands of people every year on everything… how to be safe on the job, how to handle a customer complaint, how to operate the fryer.”
In the past, he said, a company might have recorded a few videos, but they’re very high-profile and evergreen. All other training would likely be via PowerPoint slides or PDFs. “It’s not a great way of training, especially for the younger generation,” he said. Instead, they’re now creating video content – to replace not the original video footage but the text options.
Entitlement agreements are crucial
Hour One walks users through the process to get the highest quality video capture in front of a green screen. The base material becomes the training data for the AI.
“We’re basically creating a digital twin of that person — say, a CEO,” Monbiot said. “The CEO would sign an agreement that would allow us to capture the footage and create a virtual twin.” Another part of the agreement would specify who would have permission to create content using the virtual twin.
“We want people to have a very positive, comfortable and enjoyable experience with our virtual human content,” she said. “When people feel a little confused or uncomfortable, that breeds distrust, and that goes against why we’re doing what we’re doing.”
According to Synthesia, this type of authorization is common in all types of existing license agreements.
“Kim Kardashian literally licensed her likeness to app developers to create a game that made billions of dollars,” Riparbelli said. “Every actor or celebrity licenses their likeness.”
Offer influencers their images on a large scale
One synthetic media company, Deepcake, focuses less on the corporate space and more on the business of authorized deepfakes used by celebrities and influencers for brand endorsements. For example, the company created a “digital twin” of Bruce Willis to be used in an advertisement for Russian telecom company MegaFon. This led to the rumor that Deepcake owns the rights to Willis’ digital twin (which they don’t).
“We work directly with stars and talent management agencies to develop digital twins ready to be inserted into any type of content, such as: B. Commercials for TikTok,” said CEO Maria Chmir. “This is a new way of producing content without traditional resources such as constant scouting of locations and a very long and expensive post-production process.”
There are also fully synthetic people who can become brand ambassadors for a few tens of dollars, she added. Users simply type the text that these characters have to say.
“Obviously you can’t clone charisma and have a person improvise, but we’re working on it,” she said.
The future of authorized deepfakes
Synthesia says it will be adding emotion and gesture to its videos in the coming months. Hour One recently released 3D environments to create a “truly immersive” experience.
“If you think about the maturity of AI technology, every time we get to that scale, we’re unlocking more use cases,” Riparbelli said. “So I think next year we’re going to see a lot of marketing content, like Facebook ads. We will generally be seeing a lot less text and a lot more video and audio in the communications we consume every day.”
The enterprise use cases around synthetic media “deepfakes” are just beginning, said Monbiot, who added, “But this economy has already started.”
VentureBeat’s mission is intended to be a digital marketplace for technical decision makers to acquire knowledge about transformative enterprise technology and to conduct transactions. Discover our briefings.