AI avatar startup Synthesia demonstrated its latest interactive avatar capabilities by creating a personalized digital twin for a technology journalist, built using its studio production process and multi-model tech stack. The startup, valued at $4 billion after surpassing $100 million in ARR last year, is expanding beyond simple script-reading avatars to interactive agentic systems designed for corporate environments.

The customized avatar relies on Synthesia’s proprietary video animation and voice models combined with third-party tech, such as speech-to-text, agentic language models, and text-to-speech engines. Customers can configure their tech stack using alternative providers like OpenAI, Anthropic, or ElevenLabs, and can choose between self-hosting or managed cloud infrastructure provided by Synthesia.

Synthesia’s product suite includes classic video creation platforms, personal script-reading avatars, and interactive training tools like its Roleplay Sessions product. Designed for enterprise applications like sales training, these interactive agents engage users in real-time conversations and offer automated feedback and scoring based on specific training materials.

Why it matters

  • Synthesia’s multi-vendor support allows developers to swap core AI components like text-to-speech or LLM backends based on performance needs.

  • Enterprise adoption of interactive avatars is expanding from static marketing videos into dynamic internal applications like automated sales coaching.

  • Integrations of real-time voice, vision, and LLM stacks are maturing, enabling higher-fidelity interactive human-computer interfaces.

Source: techcrunch.com