Synthesia Builds First Journalist Avatar, Training 4 Versions on a 2-Minute Voice Sample
Updated
Updated · TechCrunch · Sep 26
Synthesia Builds First Journalist Avatar, Training 4 Versions on a 2-Minute Voice Sample
1 articles · Updated · TechCrunch · Sep 26
Summary
TechCrunch’s reporter said Synthesia created the company’s first digital avatar for a journalist, including two interactive versions limited to answering questions about her venture-fraud story.
A 2-minute voice recording and studio photo session at Synthesia’s New York office were enough for the startup to build four avatars in a couple of days — with and without glasses.
The interactive model is deterministic: it redirected off-topic questions from friends and family back to the article it was trained on, while the personal avatar simply read any script provided.
Synthesia said the avatar runs on voice-to-text, language, text-to-voice and video models, reflecting its broader push beyond scripted corporate videos into interactive training, surveys and roleplay tools.
The experiment left the reporter with mixed feelings about whether avatars could augment journalism, even as Synthesia — valued at $4 billion and past $100 million in ARR — bets they become routine at work.