Deepfake technology has been widely criticized over the last couple of years and has recently been named one of the biggest threats to democracy. After having been stunned by political battles, the pandemic, and Big Tech scandals, we seem to have forgotten that generative AI technology itself can be used for both good and evil.
Let’s take a closer look at the more positive aspects of deepfake technology and how they are used to create constructive outcomes. Specifically, those that disrupt the marketing and advertising industry most.
The word deepfake is a combination of the two concepts "deep learning" and "fake." As you might guess, deepfake uses machine learning to create artificial videos that are indistinguishable from those that could have actually been filmed.
In video production, the swapping of one person's face with another is one of the most common deepfake use cases. Here is one example you’ve probably heard of — Bill Hader impersonates Arnold Schwarzenegger.
As you can see, this technology allows an unprecedented level of likeness between the fake actor and the original star. If the show's director was so inclined, you would be hard-pressed to distinguish Bill from the real Arnold.
Without going into too much detail, how a deepfake works can be summarized in four key points:
It all starts with inserting the original video of the target character into a neural network. Autoencoder and GAN algorithms go to work by analyzing the subject's facial expressions and main features.
Combining an autoencoder with GAN allows the algorithm to generate fake images until it can no longer distinguish them from the original.
The video with the stunt double is then inserted into the neural network. After having analyzed the facial characteristics of the target subject, the network can easily create a deepfake. The target subject’s face is then overlaid onto the video of the stunt double. Once this process is complete, their own mother won’t be able to identify the deepfake from the original subject.
To create the perfect deepfake, you will need to clone both the target's face and voice. With regards to AI voices, creating an authentic clone works in much the same way. To learn about how audio clones are synthesized, read our article about four of the most common synthetic speech problems and how to solve them.
Once you know how deepfake works, it's not hard to imagine how the technology can be used maliciously. The most common unethical uses of deepfake include:
Blackmail
Pornography
Politics and fake news
According to various estimates, up to 96% of deepfake videos circulating on the internet are either pornography or videos with political topics. The vast majority of such videos are quite successfully blocked by social networks and search engines. However, the technology maintains a tainted reputation.
This is largely due to the fact that the negative examples of deepfake gain more traction than the positive ones do. Successful educational projects, advertising campaigns, or civic initiatives that utilize the technology are recognized as incredible works of CGI whereas those that exist in pornography or fake news are slapped with the negative deepfake label.
You probably haven't heard anything about David Beckham's anti-malaria campaign, where his voice was artificially synthesized in 9 different languages. Or the legendary Vince Lombardi 2020 Super Bowl appearance. But you've most likely heard about revenge porn, which cost an American Senator her career.
Deepfakes are the most harmful form of artificial intelligence-based criminality, and we at Respeecher take AI ethics very seriously. On the Respeecher FAQ page, you can read about our voice cloning ethics code.
If you are running a business, there are some things to consider before trying a new solution like deepfake technology to give your marketing efforts a boost. Is it beneficial for your business? Can it deliver more value to your clients? In the end, is it ethical? Let's answer these questions by starting with the last one.
Well, is it ethical enough to use Facebook ads to promote your product or service? Do the principles of neurophysiology hacking meet the morality standards for use in creating effective advertising?
Several scientists have received Nobel Prizes for discoveries that made modern advertising for brands more effective. And the number of companies that utilize Facebook user data to create more effective customized ad targeting is in the millions. Technology isn’t evil, but the way a technology is used can certainly earn it that label.
If you are looking for ethical examples of synthetic media and deepfake in particular, there are plenty out there:
In movie production, you have Carrie Fisher's appearance in Star Wars: Rogue One or Mark Hamill's latest cameo in The Mandalorian. Both were made possible with the use of video and audio deepfake.
Agency GS&P used deepfake to resurrect Salvador Dali as the host for the Dali Museum in Florida.
JFK's historic speech at the Dallas Trade Mart that the former President was never able to deliver due to his assassination was dubbed using a synthetic copy of his original voice.
A Unicef and MIT collaboration, the Deep Empathy project, used deepfake to visualize cities such as London or Boston with the same level of destruction that Syrian cities are experiencing.
These are just a couple of the deepfake projects that have made a positive social impact. And there are many more use cases that exemplify positive business impacts where marketing is concerned.
The benefits of using generative AI media boil down to three points of interest:
Costs for AI content creation are reduced tenfold.
Branded video and audio communications can be tailored to be as personalized as possible.
Attracting media stars and creating unique ad campaigns is easier and cheaper than ever.
To this day, producing video content still comes with a high price tag attached. Recruiting artists, dubbing actors, and licensing fees for content rank among the highest costs for most advertising departments.
With deepfake technology, you can purchase a license for an actor's identity and use it to generate unlimited content.
Today, software companies like AI voice generator Respeecher or Synthesia help businesses to automate the process of generating video and audio content, including client outreach, sales pitches, product tutorials, and so on.
With synthetic voices, all content can be easily localized in almost any language. It can not only reduce production costs but eases entry into foreign markets.
With synthetic video and audio content generation, businesses can create genuinely personalized communication for their clients. Voice cloning allows for the creation of customized messaging for any industry or client, even if the video remains the same.
Imagine being able to create one video ad that can be quickly dubbed in multiple languages and deliver slightly different messages depending on who's watching. No expensive production required. You can ensure your clients receive the most relevant message without having to invest a fortune in ads.
If you're a well-established brand, deepfake tech makes working with big stars a breeze. For a company specializing in deepfake marketing production, only a couple of hours of high-quality original content are required to generate an unlimited amount of deepfake content.
That means that if it's a longer collaboration, someone like David Beckham doesn’t need to clear so much time in their tight schedule for your project. You can create all the content in-house and then just send it to them for final approval. Imagine the amount of time and money this saves both your business and the celebrity.
If you are a larger brand, now is the time to create your own library of digital humans. Imagine designing faces and AI voices in the same way you style your brand. The digital identities you develop can represent your business in the market for decades. No longer is it just a customer looking at a person on a product package or a website. This is a potential movie hero, music icon, or famous blogger representing your brand.
Brand marketers are already working with Respeecher via the use of our library of AI-synthesized voices available for licensing right now. If you are interested in a free consultation or demo, just contact us using the form on the website. In the meantime, feel free to subscribe to our newsletter. Every week we talk about news in the world of generative AI and its role in the business.