Synthesia has not always been regarded as a leading player in the generative AI industry. For six years, Riparbelli and his co-founders worked diligently behind the scenes to fulfill their mission of creating video content without the need for camera equipment. Riparbelli, who is currently 31, recalls that back in 2017, there were very few investors who found this idea interesting. However, everything changed with the arrival of ChatGPT. This catapulted the Danish CEO into London’s thriving AI elite, alongside founders of renowned companies such as DeepMind, owned by Alphabet since 2014, and Stability AI, the startup behind Stable Diffusion, an image generator.
In June, Synthesia announced a funding round that valued the company at $1 billion. Although not reaching the $29 billion valuation OpenAI received in May, this still represents a significant increase of $700 million compared to two years ago, during the last investment opportunity for Synthesia.
I had the opportunity to meet Riparbelli via Zoom. He joined the call from his family’s vacation home on a Danish island, with his childhood bunk bed visible in the background. Growing up in Copenhagen, Riparbelli developed an interest in computers through gaming and electronic music. Looking back, he believes that being able to create techno music solely with his laptop, in a place like Denmark that is not widely known for its music industry, greatly influenced his current work. He stated, “It was more about who could produce great music and upload it to SoundCloud or YouTube, rather than where they lived or who their parents were in the music industry.” In contrast, he believes that the video industry still has a long way to go, as it requires substantial equipment and is inherently limited and expensive.
After completing his studies, Riparbelli immersed himself in the Danish startup scene, developing what he describes as “conventional” technologies, such as accounting software. However, unsatisfied with this path, he relocated to London in search of something more futuristic. After exploring projects related to cryptocurrency and virtual reality, he came across the concept of deepfakes and became fascinated by their potential. In 2017, he teamed up with fellow Dane Steffen Tjerrild, along with computer vision professors Lourdes Agapito and Matthias Niessner, to launch Synthesia.
Throughout the past six years, the company has amassed an extensive library of avatars. These avatars encompass various genders, skin tones, and occupations, ranging from hipsters to call center workers. Even Santa Claus is available in different ethnicities. Within Synthesia’s platform, clients have the ability to customize their avatars’ language, accents, and gestures in a script. Riparbelli expresses his fondness for a specific avatar named Alex, who is conventionally attractive with mid-length brown hair and appears to be in her mid-twenties. Interestingly, there is a real-life person resembling Alex who roams the streets somewhere. Synthesia trains its algorithms by analyzing footage of actors filmed in their own production studios.
The ownership of this valuable dataset is a significant attraction for investors. Philippe Botteri, a partner at venture capital firm Accel, which led Synthesia’s most recent funding round, explains that the algorithms rely on 3D data to understand how people move and converse. He points out that this specific data is not readily available, hence its importance.