Synthesia

Synthesia is a company that uses AI to make videos of people talking. Founded in London in 2017, it produces the kind of video where "someone explains something on camera" — like corporate training videos or internal announcements — without ever needing an actual camera shoot.

It's easy to lump this company in with general-purpose video generation models like Sora or Veo, but what they make is different. Those models are general-purpose video generators that can create almost anything, while Synthesia builds avatar models focused specifically on talking people. And its avatars aren't imagined faces either — they're filmed with real actors who give their consent to be used.

Synthesia trains its own models for voice, facial expression, and gesture. For tasks that involve handling text, like writing scripts, it relies on outside models.

In July–August 2026, the company rolled out seven new features at once, including roleplay avatars that act as practice partners and automatic dubbing. The direction is clear: expanding training videos from something you "watch" to something you "interact with."

Even "AI-generated video" comes in different flavors. Synthesia focuses on making talking-person videos, which puts it in a different lane from models built for cinematic scene generation.