Voice Activity Detection
Voice Activity Detection automatically detects whether an audio signal contains human speech – the foundation for efficient speech processing.
VAD detects speech in audio – indispensable for efficient ASR, voice agents, and real-time transcription.
Explanation
VAD segments audio into speech/non-speech sections. Modern VAD models like Silero VAD use neural networks. VAD reduces ASR costs, prevents hallucinations on silence, and enables turn-taking.
Marketing Relevance
Essential for voice agents, meeting transcription, and call center analysis. Without VAD, ASR processes unnecessary silence and hallucinates.
Common Pitfalls
Sensitive to background music. Whispering often not detected. Latency tradeoff in real-time applications.
Origin & History
Early VAD used energy thresholds (1970s). GMM-based VAD dominated 2000s. WebRTC VAD (Google) became widely used. Silero VAD (2021) brought neural VAD as open-source standard.
Comparisons & Differences
Voice Activity Detection vs. Speaker Diarization
VAD detects IF speech is present; diarization detects WHO is speaking – VAD is often the first step.
Voice Activity Detection vs. Noise Gate
Noise gates filter by volume; VAD specifically detects human speech, even at low volume.