Synthesia Builds a Digital Twin for a Journalist — What This Means for Enterprise AI Avatars

Synthesia, the Dubai-born and London-headquartered digital avatar platform valued at $4 billion, has crossed a new frontier in AI-powered communications by creating an interactive digital twin for a journalist. The avatar, trained on a specific investigative story about venture-backed startup fraud, can field questions in real time — marking what appears to be the first time the company has built a persona for media rather than internal enterprise use.

The technical pipeline behind this avatar is worth noting for anyone evaluating GPU cloud infrastructure for AI workloads. The process required a mini film studio setup: dozens of photographs and a two-minute voice capture session, followed by a multi-model orchestration stack. Voice-to-text converts speech to text, an agentic language model interprets and generates responses, text-to-voice synthesizes the audio output, and Synthesia's proprietary video model animates the avatar in real time. Notably, the platform allows customers to swap in alternatives from ElevenLabs, Cartesia, Google, or OpenAI — a flexible architecture that demands significant GPU compute at each inference stage.

Beyond this PR experiment, Synthesia's enterprise portfolio tells a larger story about where GPU demand is heading. Their Roleplay Sessions product lets employees rehearse sales pitches against interactive AI avatars that listen, respond, and score performance. This kind of real-time conversational AI, combined with high-fidelity video rendering, places sustained demands on inference infrastructure that cloud buyers should factor into their capacity planning.

As digital twins move from novelty to operational tool, the infrastructure question becomes central. Whether enterprises host avatars on their own cloud or pay Synthesia to manage them, the underlying GPU requirements for real-time voice, language, and video synthesis will only grow. The journalist's digital twin may have started as a PR stunt, but it points to a future where every professional persona could have an AI counterpart — and that future runs on GPUs.