Generative video is easy to frame as entertainment technology. Synthesia's founders took the opposite route. Victor Riparbelli, Steffen Tjerrild, Matthias Niessner and Lourdes Agapito built around a simple enterprise question: how many training, onboarding and internal communication videos would companies create if cameras, studios and actors were no longer required for each one?

The founding team combined entrepreneurship with academic expertise from UCL, Stanford, Cambridge and TUM. Riparbelli became the public face and chief executive, while the company's research roots helped it push realistic avatar generation before text-to-video became a mainstream AI category.

The wedge was boring on purpose

Synthesia's early customer proposition was not cinematic creativity. It was operational efficiency. A global company that needs the same training module in twenty languages can either coordinate presenters, translations and reshoots, or generate localised versions programmatically.

That wedge gave Synthesia a path into recurring enterprise budgets and created a natural expansion route into more interactive avatars, workplace learning and AI-assisted knowledge sharing. It also reduced dependence on consumer novelty, where switching costs are lower and usage can be volatile.

Enterprise distribution became the moat

By 2025 Synthesia said it had passed $100 million in annual recurring revenue. In January 2026 it raised $200 million at a $4 billion valuation, with GV leading the round and a long list of existing investors returning.

That funding is noteworthy because the company had already crossed a commercial threshold many generative-AI startups are still chasing: meaningful recurring revenue from organisations. The investor case was therefore not simply that video generation would improve, but that Synthesia had built a distribution layer inside large companies before the market became crowded.

London matters to the story

Synthesia moved into a larger London headquarters in 2025 and has become one of the clearest examples used by policymakers when arguing that Britain can build globally significant AI companies rather than merely host research labs for American groups.

The caveat is that frontier-model economics remain demanding and synthetic media creates its own trust problems. Synthesia has invested heavily in consent, content controls and AI-safety staffing. For an enterprise vendor, those controls are not only an ethical layer. They are part of the product.