Interactive digital avatar lets users chat with a virtual self

Synthesia, the video-generation startup, has introduced a new level of automation in public relations by allowing users to interact with virtual avatars trained on specific individuals. This technology moves beyond static text generation, creating digital twins capable of dynamic conversation. The initiative highlights the rapid integration of AI-generated content into corporate communications and media, raising questions about authenticity, trust, and the future of human interaction in journalism and business.

The Rise of Valued Digital Twin Startups

Synthesia, originally founded in the United Kingdom, has established itself as a leader in the digital avatar space alongside competitors such as D-ID, HeyGen, and Colossyan. The company achieved a significant milestone earlier this year, reaching a $4 billion valuation. Financial reports from last year indicated that Synthesia had already surpassed $100 million in annual recurring revenue (ARR). These figures underscore the growing market demand for AI-driven visual and auditory solutions in enterprise settings.

The platform enables organizations to create interactive training materials using AI avatars. A recent product launch, Roleplay Sessions, allows employees to practice scenarios like sales pitches. In these sessions, an interactive AI avatar responds to user input and provides scoring feedback, offering a realistic simulation for professional development. This functionality demonstrates how enterprises are leveraging synthetic media not just for broadcasting, but for two-way educational engagement.

Creating a Personal Digital Twin

The process of generating a personalized avatar involves capturing high-fidelity data from the subject. Participants enter a controlled studio environment where they undergo extensive photo capture and voice recording. For example, a two-minute audio sample is recorded to ensure accurate voice synthesis. Users must explicitly consent to the creation of these digital representations. The resulting assets include both personal avatars, which read pre-written scripts, and interactive avatars, which can listen and respond in real-time.

The technical architecture behind these avatars relies on a sophisticated stack of models. The system utilizes voice-to-text conversion to transcribe user speech, followed by an agentic language model that interprets the text and determines appropriate actions or responses. A text-to-voice model then converts the response into audio, while a proprietary video model animates the avatar’s facial movements and lip sync. While Synthesia uses its own models, it also integrates alternatives from providers like Cartesia, ElevenLabs, Google, and OpenAI. Enterprises retain flexibility in hosting these avatars, either on their preferred cloud infrastructure or through Synthesia’s managed services.

Testing the Limits of Deterministic AI

When tested, the distinction between scripted and interactive avatars becomes clear. A personal avatar simply recites provided text, such as reflections on seasonal changes. An interactive avatar, however, is constrained by its training data. In one instance, an avatar trained on a specific journalistic story about venture-backed startups refused to answer personal questions unrelated to that topic. It consistently redirected inquiries back to the source material, demonstrating a deterministic behavior designed to prevent hallucination or off-topic drift.

Feedback from non-technical observers varied. Some found the likeness accurate but unsettling, describing the experience as “creepy.” Others noted discrepancies in voice similarity compared to the original speaker. Despite these imperfections, the technology elicited strong reactions. One participant’s mother praised the result as “amazing,” even jokingly questioning the logic of her parents having multiple children after interacting with the digital twin. The inability of the avatar to engage in casual banter highlighted the current limitations of specialized AI agents in social contexts.

Implications for Journalism and Corporate Culture

The emergence of digital twins prompts serious consideration regarding the future of media and corporate communication. Journalists rely heavily on trust and human connection, elements that may be difficult to replicate or outsource to AI. While some investors remain skeptical about replacing human reporters with avatars, others see potential in augmentation rather than replacement. The concern over “AI slop” infiltrating news platforms suggests that audiences may resist fully synthetic media without clear disclosure.

Beyond journalism, the concept of cloning oneself for productivity offers intriguing possibilities. Employees could theoretically delegate routine inquiries to their digital counterparts, freeing up time for complex tasks. However, the psychological impact of interacting with deterministic entities remains uncertain. There is a risk of users developing unrealistic expectations or experiencing disorientation when engaging with avatars that lack genuine consciousness or spontaneity.

Generational differences may also influence adoption rates. Younger demographics, such as Gen Z, might view digital twins as inherently sci-fi or unnatural, potentially resisting widespread acceptance. Conversely, older generations accustomed to traditional media formats might find avatars less jarring than physical humanoids. Ultimately, the success of this technology depends on balancing innovation with ethical considerations and user comfort. As companies experiment with these tools, the line between human and machine interaction will continue to blur, requiring careful navigation of privacy, authenticity, and emotional resonance.