Greek startupper causes mayhem in Hollywood

A description of up to 200 characters is enough to produce a scene from a short film, video clip or advertising campaign (text to video) in less than two minutes. The figures, effects, movements, sounds etc, as they are produced through a simple written description, look so realistic that they are in stark contrast to their surreal filming process. Why is that? Entire short films, even scenes from films such as the seven-Oscar-winning “Everything Everywhere all at Once,” have been created using Runway’s generative AI software, the brainchild of Greek 32-year-old Anastasis Germanidis and Chileans Cristobal Valenzuela and Alejandro Matamala Ortiz. By tapping into technology similar to the one that made ChatGPT popular, the New York-based company launched its latest model in March 2023, while last summer it raised $141 million from Nvidia, Google and Salesforce. Today its market capitalization is at $1.5 billion, as reported by international media.

‘Our goal is not to replace the artists but to work with them’

“It is a co-production tool that works with the creator,” clarifies Germanidis, speaking to Kathimerini about the company that has intrigued Hollywood. “When we started Runway in 2018, one of our visions was to build text-to-video technology, which took us four to five years to complete.” Germanidis, a Hellenic-American Educational Foundation graduate, studied computer science at Wesleyan University in the US. He then earned a master’s degree at New York University, exploring new technologies in a creative way. During graduate school, he also met the other founders. “We were trying to see how machine learning and artificial intelligence can help the work of directors, designers etc. So we realized artificial intelligence can speed up the production of many different things based on the creator’s ideas. We saw an opportunity in the sector and seized it,” he says.

Speaking about the technology, he explains that “with Runway’s AI model, you can describe anything and it will create it; the description can range from objects, to the style you want the video, movie, frame etc. In other words, the director’s thought, when captured in just a few words, can be transformed into a movie scene. At the moment, the AI model can produce a four-second scene within one minute, which of course we are constantly improving.”

“Our technology is used by Hollywood directors as well as advertising companies. As a tool it is often used for previsualization, that is, to have a visual draft of a film before shooting it,” he adds. As for its benefits, “it’s not necessary for the entire film to have been shot with AI. Some use it to create something that is difficult or dangerous to shoot in the real world. What’s more, everything is done much faster and at a lower cost.” Germanides cites the example of director Paul Trillo, who launched the short film “Thank You for Not Answering” using solely the company’s technology. The AI model “is trained by us. It is constantly training by watching videos and then producing them itself. That’s why we invest in research and suitable human resources. Nvidia and Google became AI expert companies only after they realized how important teams with specific expertise are in order to build such models.”

The two-hour film goal

How far can this technology reach? “We are still at the beginning. Our goal is to produce a two-hour film with our technology, which I predict will be possible by the end of 2024,” he says. Germanidis emphasizes that the company launched after the release of its model’s latest version, and notes that they currently host “millions” of users, without revealing their precise number. With artificial intelligence growing rapidly, fear of replacement by AI’s digital copies is looming over the film industry. “Our goal is not to replace the artists but to work with them. When Photoshop came along, everyone thought it would alter reality, but in the end it enhanced creativity, allowing creators to produce their ideas faster. The same is expected to happen with generative AI models,” he estimates. “Man is and always will be at the center. Anything that is produced with technology is based on the ideas, the vision and the stories you want to tell,” he says, adding that Runway enables low-budget creators to “do what they couldn’t do before because they couldn’t access the right tools.”