A journalist’s AI double can discuss one story, raising questions about trust

A TechCrunch journalist has tested an interactive avatar made in their likeness, trained to answer questions about a story on venture-backed startups and fraud. The account offers a narrow demonstration of the technology: the avatar redirected questions outside that subject back to the article.

The journalist said Synthesia used photos and a two-minute voice recording to create personal and interactive avatars, a process that took a couple of days. The interactive version combined voice, video and language models. The journalist described it as deterministic: it would only respond within the material it had been trained on.

Friends and family had mixed reactions. Some found the avatars interesting and unsettling; the journalist’s parents tested the interactive version with personal questions, which it did not answer. The account reports no technical assessment of the avatar’s accuracy or likeness.

The experience prompted questions about whether avatars could present news, augment journalists or become a way for executives to interact with reporters. The journalist framed these as open questions, while arguing that trust remains central to journalism and may be difficult to outsource to AI.

Synthesia makes avatar products for enterprise use, including interactive roleplay sessions for employees. TechCrunch reported that the company reached a $4 billion valuation earlier this year and said it exceeded $100 million in annual recurring revenue last year; those figures were not independently verified in the supplied account.