TruGen and Avatario avatar IDs in LiveKit vs a Sume avatar handle

TruGen and Avatario both choose a face in LiveKit with an avatar_id, usually picked from a stock list or dashboard. Sume instead gives you a avatar_handle that you name when you create the avatar, so your code can say product_host and get the same face every time. Use a handle when brand consistency matters more than catalog breadth.

How the two LiveKit plugins pick a face

TruGen's page lists avatar_id as an optional parameter for a stock avatar model, plus avatar_participant_identity (default "trugen-avatar") and avatar_participant_name. It installs with uv add "livekit-agents[trugen]~=1.8" or, for Node.js, pnpm add @livekit/agents-plugin-trugen, and reads TRUGEN_API_KEY.

Avatario's page needs AVATARIO_API_KEY and AVATARIO_AVATAR_ID, which you choose in the Avatario dashboard. Its pattern is to start the avatar session before the agent session. It is a Python plugin installed with uv add "livekit-agents[avatario]~=1.8".

How Sume names a face

POST /v1/avatar-1.0/generate takes a top-level avatar_handle and an input union: a prompt, structured props (ethnicity, sex, age) or a photo with a public HTTPS image_url. Sume normalizes the handle and stores it without a leading @. When the creation job completes, the avatar is a reusable resource in your workspace.

Then every talking video names that handle, so the face is a stable key in your own code rather than an opaque vendor ID.

When a handle is the better key

A stock avatar ID is quick for a prototype. A handle you create is better when the presenter is part of your brand, because a teammate or an agent can reuse it without looking up a catalogue. Creating an avatar costs $0.95 once, and each later video only pays for its own seconds.

If your face has to appear in a live call, stay with the LiveKit plugin. If it has to appear in a dozen scheduled videos, create the handle once, as one avatar across twelve videos shows.