An interactive AI twin turns a reporter's article into a conversation

Synthesia built Dominic-Madori Davis a digital avatar that can discuss one of her TechCrunch stories but redirects personal questions.

1 Source, 1h ago, first seen 1h ago

TLDR

Synthesia created personal and interactive AI avatars of TechCrunch reporter Dominic-Madori Davis using a photo session and two minutes of recorded speech. The interactive version can discuss her article about fraud at venture-backed startups, but it redirects questions outside that subject. Synthesia combines speech recognition, a language model, voice synthesis and video animation to produce the exchange. Davis found the result impressive and unsettling, and concluded that a digital representative may answer routine questions but cannot replace the trust and human connection at the center of journalism.

Synthesia created an interactive digital version of TechCrunch reporter Dominic-Madori Davis that can listen to a question and answer on video, but only within the boundaries of one of her published articles.

Davis tested the avatar for TechCrunch after visiting Synthesia's New York office. The company photographed her in a small studio and recorded two minutes of her voice, then produced scripted personal avatars and interactive versions with and without glasses.

Four model stages create the exchange

The interactive avatar was trained around Davis' reporting on why venture-backed startups commit more fraud than non-VC-backed companies. It can answer questions about the research, her reason for covering it and the paper's findings.

Synthesia's pipeline turns a user's voice into text, sends that text to an agentic language model, converts the response back into speech and animates the avatar with the company's video model. Customers can use Synthesia's voice and video systems or combine them with services from other providers.

The version Davis tested was deliberately constrained. When family and friends asked where she had worked or where she lived, it redirected them to the fraud article instead of improvising an answer. That limit made the interaction less personal, but it also prevented the system from pretending to know facts outside its assigned material.

A useful stand-in is not the same as a journalist

Davis found the personal avatar's synthetic voice fairly accurate, while friends thought the interactive version's voice and likeness were less convincing. Their responses ranged from impressed to creeped out.

The experiment points to practical corporate uses, including training, sales practice and answering routine questions when a human is unavailable. Davis was less persuaded that an avatar could replace a reporter. Reporting depends on research, relationships and trust, qualities that a visual and vocal replica does not create on its own.

Her digital twin can explain one completed story. It cannot form new judgments about the people and institutions that story covers, and its most important safeguard may be knowing when not to answer.

Today's Rank

#14

Combined views

12.7K

1 Source, first seen 1h ago

An interactive AI twin turns a reporter's article into a conversation

Synthesia built Dominic-Madori Davis a digital avatar that can discuss one of her TechCrunch stories but redirects personal questions.

1 Source, 1h ago, first seen 1h ago

TLDR

Four model stages create the exchange

A useful stand-in is not the same as a journalist

Her digital twin can explain one completed story. It cannot form new judgments about the people and institutions that story covers, and its most important safeguard may be knowing when not to answer.

An interactive AI twin turns a reporter's article into a conversation

TLDR

Four model stages create the exchange

A useful stand-in is not the same as a journalist

Combined views

An interactive AI twin turns a reporter's article into a conversation

TLDR

Four model stages create the exchange

A useful stand-in is not the same as a journalist

1 Source

Combined views

Four model stages create the exchange

A useful stand-in is not the same as a journalist

Four model stages create the exchange

A useful stand-in is not the same as a journalist

Sentiment

Summary

Sentiment

Summary

Sentiment

Summary

1 Source

1 Source

1 Source

Sentiment

Summary