A Clips (V3) agent: a pre-recorded presenter, driven by the D-ID Clips pipeline.
An Expressive (V4) avatar: a real-time avatar that also supports microphone and camera publishing, client tools and interruption.
A Talks (V2) agent: a still photo animated to speak.
The kind of avatar an agent is built on, which is what decides how its video is produced.
It is the
typeof the agent's avatar (Agent.avatar) and so determines which SDK features a session has: talk is Talks (V2), clip is Clips (V3), and expressive is Expressive (V4).