- The CEO of UK-based AI video startup Synthesia, Victor Riparbelli, thinks we’ll see photorealistic computer-generated humans in the next three years.
- Synthesia is the company behind a recent video of David Beckham where the soccer player appears to be speaking with the voices of nine different malaria survivors.
- Riparbelli told Business Insider that while internally the company doesn’t like to use the term “deepfake,” it can’t really escape it.
- Visit Business Insider’s homepage for more stories.
The CEO of a UK startup pioneering deepfake technology thinks we’re just three years away from having computer-generated versions of actors that are so good, they’re indistinguishable from real humans.
Victor Riparbelli, 27, cofounded Synthesia two years ago. The company made its first big splash in 2018 when it used its technology to make a BBC news anchor appear to be speaking Spanish, Mandarin and Hindi.
More recently the company applied its tech to soccer legend David Beckham. In collaboration with the campaign Malaria Must Die, Synthesia manipulated Beckham’s facial features so that nine malaria survivors were able to speak through him — in nine different languages.
You can watch the video here:
Riparbelli told Business Insider that the actual filming was almost identical to a regular day on set. The only difference is before they shot Beckham delivering his lines, they had to train Synthesia’s algorithm on his face.
To do this, Beckham just had to talk into the camera. There’s no need for a script, although Riparbelli said they often give prompts to help people think of what to say. One example they suggest people talk about: what they had for breakfast that morning.
This footage is then translated into training data, teaching the algorithm how Beckham’s face moves so it can create a digital model of him. The whole process takes about three to four minutes. “Once you’ve done that, it doesn’t require any special hardware or cameras or anything like that,” Riparbelli said.
You can watch more here about how Synthesia made the video:
Bridging the uncanny valley
Synthesia this week announced that it has raised $3.1 million in funding, some of which came from early investor Mark Cuban. Riparbelli told Business Insider that while at the moment the company’s focused on advertising, the goal is to break into the world of TV and film special effects.
“As the company moves forward we are going to expand our platform and the plan is to start working with film and entertainment and make ideas come to life much [more easily] than they are today,” he said. He added that the tech Synthesia is developing is the same process that is already used in Hollywood films, “we’re just doing it with neural networks which make the process completely automatic.”
In recent years we’ve increasingly seen CGI versions of actors and actresses crop up in films — think “Bladerunner” or “Star Wars: Rogue One.” However, many of these digital actors fall into what’s known as the “uncanny valley” — too realistic to be cute, not realistic enough to be totally convincing. There’s just something a little off.
Riparbelli thinks we’re very close to getting rid of the uncanny valley.
“I think in the next three years we will see a significant improvement in how we can create digital humans,” he said. He added that Synthesia can already make photorealistic humans, “we just can’t do it with films yet.”
Business Insider asked whether the company has any interest in applying its tech to video games. “Right now it’s not our focus, our focus is on producing photorealistic video, but who knows maybe one day in the future,” he said.
Synthesia “can’t escape” deepfakes’ bad reputation
Synthesia’s success has mirrored the rise of deepfakes, videos where AI software is used to map people’s faces onto other people’s bodies in a realistic and often disturbing way.
The tech has gained notoriety, as it has been used to harass women by grafting their faces onto porn videos, and has stirred more generalised fears around disinformation and fake news. Internally, Synthesia prefers to use the word “synthesis” so as to escape the negative connotations.
Nonetheless, Riparbelli is resigned to the fact that the outside world will keep on describing their tech as deepfakes. “We can’t escape it, and we’re also fine with that.”
“Would I wish that we don’t always get associated with that term? Yes, but in another way everybody knows what you’re talking about when you talk about ‘deepfake’ right, because there really isn’t any other word,” he added.
Riparbelli doesn’t find the world of deepfakery totally reprehensible. He said his favourite is the deepfake video of Steve Buscemi’s face grafted onto Jennifer Lawrence, which went viral earlier this year.
I’ve gone down a black hole of the latest DeepFakes and this mashup of Steve Buscemi and Jennifer Lawrence is a sight to behold pic.twitter.com/sWnU8SmAcz
— Mikael Thalen (@MikaelThalen) January 29, 2019