I created an interactive digital avatar of myself — and you can talk to it

When Synthesia sent over an interactive avatar of its own head of corporate affairs, Alexandru Voica, the pitch was already weird enough. Then the company offered one for a reporter too. Not a generic PR bot, but a digital twin trained on a single TechCrunch story and built to answer only questions about it. That’s where this stops being a novelty and starts looking like a test case for where AI is headed in media and corporate comms.

Synthesia, originally based in the U.K. and now opening a New York office, has become one of the better-known avatar companies in the space, alongside names like D-ID, HeyGen, and Colossyan. The company said it hit a $4 billion valuation earlier this year and crossed $100 million in ARR last year. It sells tools for making training videos with AI avatars, plus a product called Roleplay Sessions that lets employees practice things like sales pitches with an avatar that responds and scores them.

For this experiment, the company took photos, recorded two minutes of voice, and built several versions: a personal avatar that just reads a script, and interactive versions that can listen and talk back. The interactive one is deterministic, which means it stays inside the bounds of what it was trained on. Mine was trained on a story about why venture-backed startups commit more fraud than non-VC-backed startups, and no amount of poking would get it to wander off script.

The tech behind it is a stack of voice-to-text, language, text-to-voice and video models. Synthesia uses its own video and voice models, but customers can also plug in alternatives from companies like Cartesia, ElevenLabs, Google or OpenAI. They can host the avatar wherever they want, or let Synthesia do it. That flexibility is the point: the company is trying to sell avatars not just as a polished front end, but as something enterprises can actually wire into other systems.

What stood out most was not the resemblance, but the limits. The personal avatar was close enough to feel eerie. The interactive one kept snapping back to its assigned story, which made it feel safer and stranger at the same time. And that may be the real question here: not whether a digital twin can be built, but whether people will trust one when it starts speaking for them.