_target_: nemo.collections.tts.data.audio_trimming.VadAudioTrimmer

model_name: "vad_multilingual_marblenet"
vad_sample_rate: 16000
vad_threshold: 0.5
device: "cpu"
speech_frame_threshold: 3
trim_win_length: 4096
trim_hop_length: 1024
pad_seconds: 0.2