Deploy via LiveKit

The fastest path to production — drop a bitHuman avatar into any LiveKit agent worker with the cloud plugin. Managed runtime, ~5-minute setup.

The cloud plugin

livekit-plugins-bithuman runs the avatar on bitHuman’s servers — no model files, no GPU on your side. It’s the fastest production path (~5 minutes) and charges per active minute against your account. Works with both Essence and Expression.

Install

pip install livekit-plugins-bithuman

Python 3.9+. The plugin pulls bithuman + livekit-agents.

Set your environment

export BITHUMAN_API_SECRET="your_api_secret"
export BITHUMAN_AGENT_ID="A78WKV4515"        # your agent code from the Library
export OPENAI_API_KEY="sk-..."
export LIVEKIT_URL="wss://your-project.livekit.cloud"
export LIVEKIT_API_KEY="APIxxxx"
export LIVEKIT_API_SECRET="xxxx"

Wire into an agent worker

from livekit.agents import Agent, JobContext
from livekit.plugins import bithuman
import os

async def entrypoint(ctx: JobContext):
    await ctx.connect()
    await ctx.wait_for_participant()
    avatar = bithuman.AvatarSession(
        avatar_id=os.environ["BITHUMAN_AGENT_ID"],
        api_secret=os.environ["BITHUMAN_API_SECRET"],
    )
    # ...attach the avatar to your AgentSession and start it.

To select Expression, pass model="expression" to the session. To point at your own Essence server instead of bitHuman’s cloud, pass api_url= — see self-hosted deployment.

What you get

• Managed avatar runtime — no GPU to provision, no Docker to operate.
• LiveKit Cloud-compatible — works with both LiveKit Cloud and self-hosted LiveKit servers.
• WebRTC delivery — video streamed via LiveKit’s media pipeline to any client.

Billing

Each session bills at the cloud or self-hosted rate depending on whether the avatar GPU is yours or bitHuman’s — 2 cr/min cloud Essence, 4 cr/min cloud Expression.

Runnable examples

Six complete LiveKit agents ship in the examples repo — each with .env.example, requirements.txt, and a docker-compose.yml full stack (LiveKit + agent + web UI). Clone, fill .env, then docker compose up:

Example	Model · where
cloud-essence	Essence · bitHuman cloud — start here
cloud-expression	Expression · bitHuman cloud
local-essence	Essence · your server (CPU)
local-expression-gpu	Expression · your NVIDIA GPU (local LiveKit + Redis)
local-expression-gpu-livekit-cloud	Expression · your GPU, WebRTC via LiveKit Cloud
local-expression-mac	Expression · Apple Silicon M3+

See the LiveKit Agents docs for the broader agent-worker model.

Where to go next

• Self-hosted GPU — run Expression on your own NVIDIA hardware.
• Embed widget — drop an iframe on any page.
• Pricing — the cloud-vs-self-hosted credit breakdown.
• API reference — agents, speak, dynamics, tokens.