OpenAIRealtimeServerVad

Server VAD turn-detection configuration for OpenAI Realtime transcription sessions.

Server VAD turn-detection configuration for OpenAI Realtime transcription sessions.

Remarks

Server VAD detects the start and end of speech based on audio volume and commits the input buffer automatically at each turn boundary.

Property	Type	Description	Defined in
`prefixPaddingMs?`	`number`	Amount of audio (in milliseconds) to include before detected speech. Default `300 (OpenAI server default)`	src/providers/stt/openai/OpenAIRealtimeSTT.ts:37
`silenceDurationMs?`	`number`	Duration of silence (in milliseconds) that ends a turn. Shorter values finalize utterances faster but may cut off slow speakers. Default `500 (OpenAI server default)`	src/providers/stt/openai/OpenAIRealtimeSTT.ts:43
`threshold?`	`number`	Activation threshold for VAD (0.0 to 1.0). Higher values require louder audio and can perform better in noisy environments. Default `0.5 (OpenAI server default)`	src/providers/stt/openai/OpenAIRealtimeSTT.ts:32
`type`	`"server_vad"`	Simple volume-based voice activity detection.	src/providers/stt/openai/OpenAIRealtimeSTT.ts:26