Deepfake avatars are going mainstream

While artificial intelligence is being used to support fake information, some pioneering companies are deploying simulated videos and voices for genuine objectives

At first glance the very concept of deepfake videos or voices appears criminal. Someone using deep learning artificial intelligence (AI) to simulate the face and voice of a person can’t be considered acceptable or legal.

Social media companies are struggling with various types of manipulated media which often are used to support fake information.

However, a pioneering set of companies is using simulated videos and voices for genuine objectives. Much like ethical hackers who work for good, these companies have created digital avatars for various applications, including public awareness campaigns and marketing initiatives, apart from education and training.

Soccer legend David Beckham lent his voice for creating a public service message in nine difference languages. AI was deployed to record an anti-malaria campaign in Beckham’s voice even though he didn’t actually say all the words in the message in languages that he probably didn’t understand.

US tech firm Synthesia, which developed the project with Beckham, has a wider set of AI offerings for training options. Companies can choose from over 40 avatars created by Synthesia for online training. After selecting an avatar, a company can add data and content. Within a short time, a “digital trainer” will be ready to deliver a training programme with realistic voice and video. Synthesia says that cameras, actors or actual trainers aren’t needed for creating videos.

While some companies create digital avatars with unique faces and voices, some other companies are using real life personalities. People rent their faces and images or allow the usage for companies to create digital avatars.

Such companies will pay a fee to people allowing the usage of their image to create marketing messages or training videos. A video of such a person can be used for various ads or promotions in different markets and languages.

Videos of historical characters or prominent personalities who are deceased can be created for educational purposes

Tel Aviv-based Hour One is doing leading-edge work with rented faces. Unlike a purely digital face, Hour One will record a few videos in high resolution with the person who has agreed to lend their image. Once it is recorded, AI takes over. Any text in any language can be mouthed by a moving image created by AI.

The technology for creating avatars and deepfakes is similar. The key difference is legality, permissions and disclosures. Most deepfakes found online have illegally used faces and images of known personalities. Digital platforms are in a constant battle to remove such illegal and malicious content.

Legally approved deepfakes are the way ahead for the synthetic media industry. In the movie business, computer-generated images have been totally acceptable. For shooting dangerous action scenes stunt doubles are used, too. With strides in deepfake, movie producers can use AI to make characters perform tricks without risking anyone’s safety. The younger version of Star Wars character Princess Leia was created using similar technology.

Videos of historical characters or prominent personalities who are deceased can be created for educational purposes. Several experimental projects are being implemented in various sectors by various startups.

Digitally generated images, videos and voices can be used for many applications. Such efforts will require a strong set of guidelines, if not regulation. At the centre of such guidelines should be disclosures to ensure that viewers are not misled. Viewers and consumers have a role to play by demanding disclosures and transparency from synthetic media creators. Approvals and permissions from individual whose images are being used must also be a part of the process.

While the entertainment industry has been an early adopter, the deployment of synthetic images has the potential of making a deep impact on the advertising, marketing and training industries.