Back to Gap Map Model components / Speech & audio

Silero VAD

open source / Overall score: 2.8

A small pretrained voice activity detector that processes a 30 ms audio chunk in under a millisecond on one CPU thread. It ships as PyTorch and ONNX models with a Python package and runs wherever ONNX Runtime does, including the browser.

Published by Silero under the snakers4 GitHub account.

Openness

5 high confidence

5.0

license: MIT(OSI)
source: public
core features withheld: no

Silero VAD is MIT-licensed and, in the vendor's words, comes with no keys, registration or vendor lock; the published repository is the whole product.

https://raw.githubusercontent.com/snakers4/silero-vad/HEAD/LICENSE recorded 2026-09-27
LICENSE body: "MIT License" / "Copyright (c) 2020-present Silero Team".
https://raw.githubusercontent.com/snakers4/silero-vad/HEAD/README.md recorded 2026-09-27
"Published under permissive license (MIT) Silero VAD has zero strings attached - no telemetry, no keys, no registration, no built-in expiration, no keys or vendor lock."

Adoption

4 high confidence

4.0

PyPI downloads of silero-vad. Many engines embed the model without installing the package, and those uses are not counted.

https://pypistats.org/api/packages/silero-vad/recent recorded 2026-09-27
"last_month":1831401 for silero-vad.

Capability

2 medium confidence

2.0

Probably the most widely embedded VAD, but voice activity detection is a pipeline component: it produces no transcript or speech of its own.

https://raw.githubusercontent.com/snakers4/silero-vad/HEAD/README.md recorded 2026-09-27
"takes](https://github.com/snakers4/silero-vad/wiki/Performance-Metrics#silero-vad-performance-metrics) less than **1ms** to be processed on a single CPU thread."; "Silero VAD was trained on huge corpora that include over **6000** languages".

Verified 2026-09-27