When Alexandru Voica, head of corporate affairs at the video-generation startup Synthesia, sent me a link this summer to the newest addition to their public relations team, my first reaction was one of genuine surprise. It was an interactive virtual avatar of Voica himself, meticulously trained to field common press inquiries about the company, its operational mechanics, and its broader mission. The timing was particularly poignant; only the day before, I had participated in a professional panel where veteran PR practitioners asked me for my stance on pitches composed using AI-generated text. Yet, encountering Voica’s avatar felt like an escalation beyond mere text generation. It felt like the "final boss" of AI integration in the communications industry.
Synthesia, a powerhouse in the digital avatar space, has seen a meteoric rise. Originally founded in the United Kingdom, the company has become a centerpiece of the generative AI revolution, standing alongside competitors like D-ID, HeyGen, and Colossyan. Earlier this year, the company hit a significant milestone, securing a $4 billion valuation, following reports last year that it had crossed $100 million in Annual Recurring Revenue (ARR).
During a visit to Synthesia’s new office in New York this September, the team posed a question that would have been unthinkable just a few years ago: Would I like to create my own AI avatar? I did not hesitate. I accepted immediately. My outfit was professional, my hair was perfectly in place, and the prospect of possessing a digital twin was too intriguing to pass up.
Prior to this encounter, I had remained largely indifferent toward the proliferation of avatars. However, it has become increasingly clear that these digital personas are destined to become a permanent fixture of our daily online existence. I have observed individuals on platforms like Instagram crafting high-fidelity digital likenesses to streamline social media content production. It is a fascinating, if complex, evolution in digital identity. This curiosity is perhaps why I feel no hesitation in presenting my own digital twin to you today. This project marks the first time Synthesia has generated a custom digital avatar for a journalist—or indeed, for anyone outside of their own internal team. My avatar has been specifically trained on my recent reporting regarding why venture-backed startups are statistically more prone to committing fraud than their non-venture-backed counterparts. True to its programming, it will only engage in discourse related to that specific story.
The process of bringing my digital twin to life was a fascinating look behind the curtain. I stepped into a mini film studio tucked away within the Synthesia office, where the team captured a high-resolution series of photographs and a two-minute voice recording. I provided formal consent for the creation of these assets, and with that, "digital Dom" was born. The team produced a variety of versions: a personal avatar that reads scripts provided by the user (available both with and without my glasses), and two interactive, agentic avatars capable of listening and responding to live input.
The architecture behind this technology is a sophisticated stack. The interactive avatar is powered by a seamless integration of voice-to-text, natural language processing, text-to-voice, and video animation models. While my specific avatar utilizes Synthesia’s proprietary video and voice models, the company maintains an open architecture that allows enterprise clients to integrate alternative services from providers such as Cartesia, ElevenLabs, Google, or OpenAI. Companies also have the flexibility to choose their hosting environment, opting for their own cloud infrastructure or utilizing Synthesia’s managed hosting services.
At its core, the system functions through a logical progression: a voice-to-text model transcribes user input, an agentic language model interprets that text to determine the appropriate action, a text-to-voice model generates the spoken response, and a video model synchronizes the avatar’s facial movements to the audio output.
Currently, Synthesia organizes its offerings into three primary product categories. First, there is the standard video-creation and distribution platform, which features classic avatars that repeat user-provided scripts. Second, there is the agentic platform known as "Sessions," designed for surveys and roleplay—such as their recently launched product that allows employees to practice sales pitches with an AI that provides real-time feedback and scoring. Finally, there is an API platform, which allows developers to combine Synthesia’s models with external tech services to build bespoke interactive experiences.
It took the Synthesia team approximately two days to finalize my avatars. My initial tests were with the personal versions; I input a relatively generic script about the arrival of autumn in New York. The resulting audio was strikingly accurate, successfully avoiding the hoarseness present in my raw sample recording. When I showed these clips to friends who work outside the technology sector, the reaction was a blend of fascination and mild unease.
The response to the interactive version was even more telling. Because it is a deterministic model—meaning it is hard-coded to respond only to the data within my venture fraud story—it remained steadfastly on topic. When my friends attempted to ask off-script questions about my background or my personal life, the avatar politely but firmly redirected them back to the research. My mother, perhaps the most candid critic, found the technology "amazing." She and my father spent a significant amount of time attempting to probe the AI with personal questions only a family member would know, only to be met with the same consistent, professional redirection. "I don’t remember giving birth to two of you," she joked after the experiment.
This experience has forced me to confront the long-term implications for the future of journalism. Is the public prepared to consume news presented by an avatar? One investor I spoke with dismissed the idea instantly, citing the necessity of human presence. There is undoubtedly significant pushback against the "AI slop" currently infiltrating social media and news aggregators. Yet, others in the industry are less certain. Could these avatars eventually augment, or perhaps even replace, the traditional role of a journalist? Would a CEO prefer to be interviewed by an AI avatar rather than a human reporter?
The core value of my career lies in human connection, the nuance of writing, and the depth of investigative research. Trust is the currency of journalism, and that remains an element that seems impossible to outsource to an algorithm. However, outside of the media landscape, the utility of cloning oneself is undeniably attractive. The idea of a digital twin handling mundane inquiries while one is on vacation or off the clock represents a tangible shift in corporate productivity.
We are currently in the early stages of observing how avatar usage will evolve across corporate America. I admit to having mixed feelings. After the initial novelty wore off, I found myself staring at the avatar after it finished speaking, waiting for it to do something—blink, smile, or offer a flicker of awareness. I was waiting for it to show that it "knew." Because these models are deterministic, they will never do that. However, I can easily imagine how a user might slip into a form of "AI psychosis" when interacting with a non-deterministic model—one powered by a chatbot capable of open-ended, unpredictable pontification.
I told one investor that, regardless of where this technology goes, I suspect my generation, Gen Z, will struggle to fully normalize it. The experience feels inherently science-fictional, as if everything we once watched in movies has suddenly materialized in our offices. Yet, I find digital avatars significantly less jarring than humanoids. At the very least, with a digital avatar, if the interaction takes an uncomfortable turn, I always have the option to simply log off.
In the meantime, you can see the results for yourself. My personal avatar is ready to provide a briefing on this week’s top stories—a preview of what the future of information delivery might look like.
