Skip to content
← Back to Research

industry · Fora Soft (June 2026) · Jun 2, 2026

Real-Time Avatar Lip-Sync in a Call: How It Works

Nikolay Sapunov

In plain English

A vendor explainer of the five-stage real-time avatar loop (speech recognition, LLM, text-to-speech, avatar synthesis, video transport) and its latency budget, arguing the avatar server should join the call as its own participant.

This industry source is not peer-reviewed. Trace specific figures to the original source before citing them.

real-time avatars
latency
vendor analysis
conversational ai
Read the source ↗

Related research