channel: "video" gives a voice agent a face: an Anam avatar joins the call, consumes the agent’s audio and publishes lip-synced video. Everything else — pipeline (cascade or realtime), voices, tools, handoffs — behaves as on voice.
Switching to video fills in a default avatar; switching away clears it.

​
Picking a face

Each entry has id, name, variant, image_url, active_version and render_style. Filter with active_version (Cara generation) and render_style; the response lists the available values as active_versions and render_styles. offset/limit page it, up to 100. These are gallery filters, not models — Anam renders each face at its own default version. With your Anam key saved, the gallery is your Anam account’s (workspace_key: true).

​
Requirements and limits

  • Your own Anam key. Publishing without one is refused: no Anam API key for this workspace - add one under Provider keys. See provider keys.
  • No fallback to voice. If the avatar has not joined within 30 seconds the call ends with avatar_start_failed; if it leaves mid-call, avatar_disconnected. See close reasons.
  • Web only. A video agent cannot be assigned to a phone number or dialled outbound; start calls with POST /v1/calls/token — see web calls.
  • resume_false_interruption is forced off — see turn handling.
  • Handoffs keep the same face. A video agent hands off only to another video agent on the same pipeline with the same avatar provider, model, avatar_id and name.

​
What it costs

Avatar time is metered as wall-clock seconds from session start to close, idle included, at the catalog’s Anam rate, on your Anam account, shown as its own line in the call’s cost breakdown. Talqing’s platform fee is higher for video than voice — see pricing and credits. Rendering the avatar track in your page: web call features.

​
Next

Web calls

Call tokens and the browser integration.

Channels

What voice, video and text each have.