TechCrunch Reporter Builds AI Avatar Twin to Answer Questions About Her Work

Synthesia, a digital avatar startup valued at $4 billion, recently created an interactive AI clone of a TechCrunch journalist—marking the first time the company has built such a tool for someone outside its own team. The avatar was trained to field questions about one specific article on venture-backed startup fraud and demonstrates how synthetic media technology is evolving beyond simple video generation into conversational agents.

How the Avatar Works#

The technology combines several AI components working in concert. When someone asks the avatar a question, voice-to-text software converts spoken words into written form. An agentic language model then processes that text and formulates responses rooted in the training material. A text-to-voice engine transforms the response back into audio, and Synthesia's proprietary video model animates the avatar as it speaks. The entire system was built using a mix of Synthesia's own models and third-party tools from providers like OpenAI and ElevenLabs.

Creating the avatar required a mini photoshoot and a two-minute voice recording session at Synthesia's New York office. The company produced four versions: two personal avatars (one with glasses, one without) that recite whatever script is fed to them, and two interactive variants capable of listening and responding. The build process took approximately two days.

From Novelty to Uncanny Valley#

Initial reactions were mixed. The personal avatar, tested with a generic script about autumn in New York, sounded reasonably accurate and avoided picking up on audio artifacts from the original recording. Friends found it "interesting and creepy." The interactive avatar, while faithful enough to be unsettling, didn't quite capture the journalist's voice or appearance as convincingly—though it was close enough to feel strange.

The interactive avatar operates under strict constraints. Because it was trained exclusively on one article, it refuses to answer off-topic questions, consistently redirecting conversations back to the venture fraud story. This deterministic behavior meant family members couldn't trick it with personal questions only they would know.

Implications for Journalism and Beyond#

The experiment raises uncomfortable questions about the future of media. Would audiences accept news anchors who are AI avatars? Is there appetite for digital doubles handling routine corporate queries? Industry reactions vary. Some investors dismissed the idea outright, while others saw potential. The journalist notes that avatar adoption may depend on context—skepticism toward AI content on social media contrasts with possible corporate utility.