Video agents with a face and voice that hold a live conversation.
What it does
On this page
Tavus builds conversational video: an agent with a face and a voice that listens and answers in real time. The platform exposes an API for the full pipeline — perception, language model, speech and the video render — plus a maker studio for people who do not write code.
How it works
Face and voice are created or cloned, then attached to a conversation flow with a knowledge base, memories and guardrails. The company's newer models add perception of expression and tone, and the pipeline can run audio-only when video is not needed. Sessions can join Google Meet, Zoom or Teams.
Who it is for
Teams shipping onboarding, sales or support conversations where a person on screen matters more than a chatbot bubble, and researchers testing human-agent interaction.
Pricing
A free tier, minute-based plans and custom enterprise contracts. The pricing page does not publish plan amounts, so compare through the account or sales.
Limits
Latency and cost rise as video and perception features are switched on. Custom face creation requires consent checks, and the perception models are still in preview.
Source: https://www.tavus.io/

