Realtime STT + Speaker ID

VAD Configuration ▶

Parameter Help

positiveSpeechThreshold: Probability above which a frame is classified as speech
negativeSpeechThreshold: Probability below which a frame is classified as silence
minSpeechMs: Minimum speech duration before VAD confirms real speech
preSpeechPadMs: Audio to prepend before detected speech start
REDEMPTION_START_MS: Initial grace period after speech end before flushing
REDEMPTION_END_MS: Final grace period (shrinks as segment grows)
MIN_DURATION_MS: Minimum segment duration — speech end before this enters protected zone
MAX_DURATION_MS: Maximum segment duration before forced seamless split
OVERLAP_CHUNKS: Number of audio chunks overlapped during seamless split

Speech Detection (applied on next Start)

Positive Threshold: 0.6

Negative Threshold: 0.5

Min Speech (ms): 500

Pre-Speech Pad (ms): 1500

Segment Cutting (applied immediately)

Redemption Start (ms): 250

Redemption End (ms): 30

Min Duration (ms): 3000

Max Duration (ms): 9000

Overlap Chunks: 4