Interactive Avatars

Build real-time, conversational avatar experiences on your own stack. You bring the LLM and speech-to-text; Synthesia renders the avatar, with optional built-in text-to-speech if you don't want to bring your own.

You'll need Python 3.10+ and a Synthesia API key with Interactive Avatar access to get started.

Get oriented

Concepts

The shared mental model: what runs where, and where your responsibilities begin and end.

Overview

How a bring-your-own integration fits together end to end, from your app to the session.

Plugin reference

Every parameter, method, event, and exception in the livekit-plugins-synthesia package.

Build it with an agent

A Claude Code and Cursor skill that wires up the LiveKit integration for you, automatically.

Quickstarts

Start with the Minimal quickstart.

Minimal quickstart

The smallest working real-time avatar session. Start here.

Building something more specific?

RAG quickstart

Ground your avatar's responses in your own knowledge base.

Tools quickstart

Give your avatar tools it can call during a live session.

Or integrate directly

Not using the Python plugin? You can call the API yourself. You still connect to a LiveKit room, but you mint the token and manage the session instead of the plugin doing it for you.

Create Interactive Avatar

POST — Create an interactive avatar you can start sessions with.

Create Interactive Avatar Session

POST — Start a real-time session with an existing avatar.

Troubleshooting

Errors

Error types and codes, and what each one means for your integration.

Operational trust

Pricing, limits, known limitations, and production guidance.

API updates · Status