Expand description
Client-side Voice Activity Detection (VAD).
WaveKat-backed VAD with the previous dual-threshold energy detector retained as a fallback for unsupported sample rates or frame sizes. Complements Gemini’s server-side VAD:
- Bandwidth savings: Don’t send silence over the network
- Latency reduction: Signal
activityStartbefore server detects it - Barge-in pre-emption: Flush jitter buffer locally before server confirms
Structs§
- VadConfig
- VAD configuration parameters.
- Voice
Activity Detector - Voice Activity Detector with adaptive noise floor.