Picovoice Cobra VAD has quickly become the most accurate VAD engine in the market while having a negligible runtime footprint. Cobra shines the most when you compare its accuracy with the widely used WebRTC VAD developed by Google.

Voice Activity Detection (VAD) is a crucial component of many speech solutions. VAD detects the presence of a human voice within a stream of audio. Although simple to describe, this is a challenging task in the presence of non-stationary noise.

The V1.2 release of Cobra brought significant accuracy improvements compared to Its predecessor at no additional runtime cost! In the Figure below, one can observe that at %90 true positive rate V1.2 incurs less than half the miss rate of V1.1. If you are building with Cobra, make sure to update your SDK. Otherwise, read to try Cobra VAD live, inspect the open-source benchmark, and start building for free with Picovoice Free Plan.

Cobra VAD V1.2 compared to Cobra VAD V1.1

Live Demo

Probability of Voice

Open Benchmark

An open-source comparison between Cobra and WebRTC's VAD (developed by Google) is available in the Picovoice docs. The Figure below summarizes the comparison in the form of a ROC curve (a larger area under the curve is better).

Start Building

cobra = pvcobra.create(access_key)
while True:
is_voiced = cobra.process(audio_frame())