Files
voice_linux/config.ini
T

32 lines
946 B
INI

# voice_linux runtime configuration
# Hotkey format supports Ctrl, Alt, Shift, Super + final key symbol
# Example: Ctrl+Alt+V
model_path=./models/ggml-base.en.bin
use_gpu=1
audio_device=alsa_input.usb-Shure_Inc_Shure_MV6_MV6_5-ed2d37fa14d3e65d8d477cacf910d93c-01.mono-fallback
sample_rate=16000
language=en
hotkey=Ctrl+Alt+V
type_delay_us=8000
max_record_seconds=30
# always-on VAD controls
always_on=1
# legacy window size (unused by new VAD pipeline, kept for compatibility)
always_on_window_ms=1400
# live level threshold (0.0 - 1.0)
vad_peak_threshold=0.006
# level must stay above threshold this long before speech starts
vad_trigger_ms=100
# level must stay below threshold this long before speech ends
vad_release_ms=300
# include this much pre-trigger audio in transcript segment
vad_preroll_ms=500
# discard segments shorter than this
vad_min_speech_ms=250
# hard cap for one speech segment
vad_max_speech_ms=15000
autotype_enabled=1