A normalized request maps through the family mapper with its eagerness withheld, then gets the
Eleven-vocabulary eagerness (see MapElevenRealtimeEagerness). With no request the block
is omitted so ElevenLabs' own turn-taking applies — the product's turn model is a large part of
why it is chosen. Meeting mode still needs create_response: false; semantic_vad is used for
it because it is the closer of the two documented modes to ElevenLabs' native turn model.
Builds Eleven Realtime's
audio.input.turn_detectionblock.A normalized request maps through the family mapper with its eagerness withheld, then gets the Eleven-vocabulary eagerness (see MapElevenRealtimeEagerness). With no request the block is omitted so ElevenLabs' own turn-taking applies — the product's turn model is a large part of why it is chosen. Meeting mode still needs
create_response: false;semantic_vadis used for it because it is the closer of the two documented modes to ElevenLabs' native turn model.