Choose your integration
Two ways to put an Atmee avatar in front of your users: let Atmee run the whole conversation and only join the room, or bring your own LiveKit agent and let Atmee render the face.
Atmee offers two integration paths. Both use the same API key, the same avatars and the same credits; they differ in who runs the conversation.
| Session API (hosted conversation, room only on your side) | LiveKit Agents plugin (bring your own agent) | |
|---|---|---|
| Who runs the conversation | Atmee: speech recognition, the avatar's persona and model, its voice and the video. | You: your own LiveKit agent with your speech recognition, model and voice. Atmee only renders. |
| What you build | One server endpoint that calls POST /v1/session, and a page that joins the room it returns. | A LiveKit agent in Python (livekit-agents) that adds atmee.AvatarSession before it starts. |
| Where the room lives | On Atmee's LiveKit. The response carries serverUrl and a one-room userToken. | In your LiveKit project. Your agent mints the avatar's join token; Atmee never sees your LiveKit secret. |
| What the avatar needs | A finished, ready-to-chat avatar: portrait, voice and persona. | A portrait. Render-only avatars are ready the moment they are created; a full avatar works too. |
| Avatar generation | v1 avatars. | v1 avatars: a talking head rendered from a single portrait. |
| Billing | Per minute at the standard rate, from your credits. | Per minute at the render rate, lower than a hosted session, from your credits. Your key's spend cap and limits apply to both. |
| Languages, tools, knowledge | Whatever you configure on the avatar in the studio. | Whatever your agent does; Atmee has no say in what is said. |
| Client | Any client that speaks LiveKit: web (livekit-client), iOS, Android, Unity. | Any client already in your LiveKit room. |
| Endpoints | POST /v1/session | POST /v1/avatars/{avatarId}/avatar_sessions, GET /v1/avatar_sessions/{id}, POST /v1/avatar_sessions/{id}/end |
Pick the Session API when
- you want a talking avatar on your site or in your app without running any AI yourself,
- the avatar's persona, knowledge and voice are set up once in the studio and should stay in Atmee's hands,
- your side is a page or a mobile client that only needs to join a room and show video.
Start with the Quickstart; the full request and response are in the Reference.
Pick the LiveKit Agents plugin when
- you already run a voice agent on LiveKit Agents, with your own speech recognition, model and voice, and want to give it a face,
- the conversation must stay in your infrastructure: your LiveKit project, your prompts, your tools, your logs,
- you need a video avatar for something that is not a chat: a narrator, a presenter, an agent that reads from your own systems.
Start with the LiveKit Agents plugin page.
Using both
The two paths share everything but the conversation. One API key with the sessions scope
serves both; the avatar you built for hosted sessions renders just as well from the plugin, and
a portrait-only avatar created for the plugin becomes a hosted avatar as soon as you add a voice
(PUT /v1/avatars/{avatarId}/voice). Sessions of either kind count against the same spend cap
and concurrency limit of the key.
Next steps