The

min_speech_duration

parameter is used in the Silero Voice Activity Detection (VAD) system to specify the minimum duration of speech required to start a new speech chunk. By default, it is set to 0.05 seconds. This parameter can be set when loading the VAD model or updated later using the

update_options

method. It helps the system distinguish actual speech from brief noise or artifacts. Sources: livekit.plugins.silero API documentation