Synthesia created a TechCrunch journalist’s interactive AI avatar, trained only on the writer’s article arguing that venture-backed startups commit more fraud than non-VC-backed ones. It redirected unrelated questions; the writer said it left them with mixed feelings.
The reporter consented, posed for photos and recorded two minutes of speech; Synthesia took a couple of days to make the avatars. The journalist also received a noninteractive version that reads typed scripts. The interactive system converts speech to text, generates a response, turns it into audio and animates the avatar. Synthesia also offers role-play tools for employee training.
The reporter’s friends found the avatars interesting and somewhat creepy; they said the interactive version’s voice and likeness were less convincing than the script-reading avatar. The writer questioned whether avatars could present or replace journalists, saying journalism depends on trust that, in the writer’s view, does not seem possible to outsource to AI.
