# Voice Activity Detection (VAD)

> A check that tells when someone is talking and when they stopped.

**Field:** Artificial Intelligence  
**Category:** Multimodal  
**Also known as:** VAD, Speech Detection, Turn Detection

## Think of it like

It is like a light that turns on only when someone is talking. When they stop, it goes off.

## Definition

Voice activity detection, or VAD, tells apart speech from silence and noise in audio. It marks where talking starts and where it stops. Voice agents use it to know when you have finished, so they can reply. Speech-to-text tools use it to skip the quiet parts. If it waits too long, the reply feels slow. If it is too quick, it cuts you off mid-thought.

## Example

A voice agent waits for 0.5 seconds of quiet before it replies. Emma pauses to think, and the agent jumps in too soon, so she raises the wait.

## Why it matters

Every voice app has to know when it is its turn to talk. VAD is the part that decides.

## Common mistake

Setting the silence wait the same for everyone. People pause in different ways, and a busy room changes what counts as quiet.

## Three principles

- **Speech or not.** Each slice of sound is marked.
- **Start and stop.** It finds where talking ends.
- **Tune the wait.** Too short cuts in, too long lags.

## Related terms

- [Voice Agent](https://learn.benibix.com/term/ai/voice-agent)
- [Speech-to-Text (Speech Recognition)](https://learn.benibix.com/term/ai/speech-to-text)
- [Wake Word](https://learn.benibix.com/term/ai/wake-word)
- [Speaker Diarization](https://learn.benibix.com/term/ai/speaker-diarization)
- [Speech-to-Speech AI](https://learn.benibix.com/term/ai/speech-to-speech)

## Told apart from

- [Wake Word vs Voice Activity Detection](https://learn.benibix.com/compare/wake-word-vs-voice-activity-detection): People think a wake word and voice activity detection are one check, since both listen. A wake word listens for one set phrase and ignores all other talk. Voice activity detection does not care what is said. It only marks when someone starts and stops talking, so the device knows when its turn comes.

![Top row: a voice goes through three boxes, speech to text, a language model, and text to speech, then out as a voice. Bottom row: a voice goes through one box and straight out as a voice. Under the top row, small tags mark the wake word, voice detection and who is speaking.](https://learn.benibix.com/diagrams/ai5-speech-pipeline.svg)

---

Source: https://learn.benibix.com/term/ai/voice-activity-detection
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-17
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
