Synthesia, a UK-originated video-generation startup, recently opened a New York office and announced a $4 billion valuation after reporting more than $100 million in annual recurring revenue. During a visit, the company’s head of corporate affairs offered a technology journalist a custom AI avatar, marking the first time the firm has produced a digital twin for an external individual.
The avatar creation took place in a small studio inside the New York office, where the journalist posed for numerous photographs and recorded a two-minute voice sample. After providing consent, Synthesia generated both a simple reading avatar and a conversational version, each offered with and without glasses, and trained the latter on a specific article about venture-backed startup fraud.
The conversational avatar relies on a pipeline that first converts spoken input to text, then processes the text with an agentic language model, produces spoken output via a text-to-voice system, and finally animates the visual representation using Synthesia’s proprietary video model. Customers may replace any component with alternatives from providers such as Cartesia, ElevenLabs, Google, or OpenAI, and can host the service on their own cloud infrastructure or use Synthesia’s hosting.
Synthesia’s product suite is divided into three categories: a classic video-creation platform where users supply scripts for static avatar playback, an agentic Sessions platform that enables interactive role-play or survey experiences, and an API offering that lets developers combine Synthesia’s video and voice models with external services to build custom interactive solutions.
Testing the avatar revealed a voice that matched the recorded tone without reproducing any hoarseness, while the visual likeness was close enough to feel unsettling to some observers. The interactive version was deterministic, responding only with information from the trained fraud story, and redirected personal queries back to that content. The journalist’s parents tried personal questions, but the system consistently refused to answer.
The experience sparked a debate about the future role of AI avatars in journalism. Some investors expressed immediate doubt about replacing news presenters with synthetic counterparts, citing concerns over credibility, while others saw potential for augmenting reporting. The journalist emphasized that trust remains a core pillar of the profession, suggesting that full substitution by avatars is unlikely without significant cultural shifts.
Beyond media, the notion of a personal digital twin could appeal to professionals seeking continuous presence, allowing a virtual version to handle routine inquiries during vacations. As corporate America experiments with such technology, Synthesia’s rapid avatar production and flexible deployment options position it as a notable player in the emerging market for personalized AI representations.