The server granted an avatar: the model will send the agent's video.
OptionalVideoThe avatar's video carries the voice (a muxed MP4); otherwise the voice comes separately, as PCM. Default: whether the provider's video can carry the voice (RealtimeVideoConformanceTraits.VideoCanCarryVoice).
What the provider's server granted the session: an avatar or not. The kit adds the host's track request itself.