The kind of avatar: talk for Talks (V2), clip for Clips (V3), expressive for Expressive (V4).
Talks (V2) and Clips (V3) agents stream over WebRTC and receive their events on the notifications web socket; Expressive (V4) agents connect to a real-time session instead, which is what makes the microphone, camera, client tool and data-channel methods of AgentManager available.
Either an AvatarType member or its string value — AvatarType.Talk and 'talk' are
both accepted, so an Agent restored from your own storage does not have to import the
enum.
OptionalvoiceThe voice the agent speaks with.
Optionallanguage?: stringLanguage of the configured voice. The SDK reports it in analytics; it does not change how the session behaves.
The avatar an agent speaks through: its rendering tier and the language of its voice.
The value of Agent.avatar. Its
typeis the single most important field on an agent: it decides which transport the session uses and therefore which SDK features work.