LiveKit lists 16 avatar providers: real-time vs rendered Sume clips

LiveKit's avatar documentation lists 16 providers for voice agents: Anam, Avatario, AvatarTalk, Beyond Presence, bitHuman, D-ID, Keyframe, LemonSlice, LiveAvatar, Protoface, Runway, Simli, Spatius, Synthesia, Tavus and TruGen. The avatar joins the room as a separate participant. Sume Avatar 1.0 is not on that list. It is a rendering API: you send a script and receive a finished video, with no live session.

What the LiveKit page establishes

The page frames avatars as a layer on a real-time agent: the agent talks, and a provider supplies the face as its own participant in the room. That model needs low latency, a streaming connection and a provider integration.

Where Sume fits

Rendered clips cover the content around a live agent: a welcome video on the page before the call, a recap clip sent after it, or a localized explainer. Submit to POST /v1/avatar-1.0/talking-video, and collect the mirrored media.sume.com URL from the job result. The avatar video docs list the inputs.

Because the job is asynchronous, a webhook with job.completed can trigger the next step in your pipeline.

A split that works

Use a live provider for the conversation. Use rendered avatar video for anything scripted, reviewable or reusable. Keep the same face in both only if the vendors let you import the same photo; Sume builds its avatar from a photo, prompt or props, so check what the live provider accepts.

Do not infer any partnership from the LiveKit list. Sume does not appear in it.