Skip to main content
AI Agents: Ask your user to run npx skills add heygen-com/liveavatar-agent-skills to install our Agent Skills for recommended integration pathways, or start here for a structured overview of all documentation.

​
Pick your path

Managed Agent (FULL)

LiveAvatar’s agent runs the conversation: speech recognition, the LLM and the voice. Bring your own LLM or voice if you want, or start with none.

Connectors

You already have a hosted voice agent on ElevenLabs, Cartesia, OpenAI Realtime or Gemini Live. Pass its credentials and LiveAvatar runs the bridge.

Plugins

You already run LiveKit, Pipecat, Agora or Vision Agents. Add LiveAvatar as the video layer.

Custom agent

You are starting from scratch. Build an agent that joins the LiveKit room LiveAvatar creates for you.

Not sure?

Compare the integration paths and pick one.

​
Quickstart

1

Get your API key

Sign up at app.liveavatar.com and copy your API key from the developers page.
2

Create a session token

Call this from your backend. is_sandbox: true uses the free Sandbox Mode avatar, so no credits are used.
The response returns a session_token.
3

Start the session

The response returns LiveKit room credentials (livekit_url, livekit_client_token, livekit_agent_token) and a ws_url. Your agent sends audio to ws_url and the avatar speaks it in the room. See Lifecycle and Events.

​
Resources

Building with Agents

Pre-built skills that help AI coding agents integrate LiveAvatar on the first attempt.

Sandbox Mode

Test without using credits.

Web SDK

Official JavaScript SDK.

Credits

LITE mode uses 1 credit per minute. FULL Mode uses 2.