An 8M-parameter multilingual turn-detection model, a Whisper Tiny encoder with a linear head, that predicts from raw audio whether the speaker has finished; the int8 CPU ONNX build is about 8.7 MB.
onnx-community/silero-vad
Indexed from Hugging FaceLinked to its home on Hugging Face. FrontierOS indexes it; nothing is hosted or mirrored here.ModelVoice activity & turnsEst. Single-board computerEstimated minimum compute for inference: Raspberry Pi 5 class CPU. Not a benchmark.MITDesk device
About
ONNX exports of the Silero voice activity detector in fp32, fp16 and 8-bit forms (the int8 file is about 0.6 MB) that mark which short audio frames contain speech.
Why it matters for physical AI
VAD is the always-on gate in front of every speech pipeline, and this one is cheap enough to run continuously on a Pi or a device's laptop host.
Summary written by FrontierOS in our own words. The upstream card and license govern use. Metadata checked 27 Sep 2026. Read the full card on Hugging Face
Files, weights and version history live on Hugging Face. FrontierOS indexes this repo; it does not host or mirror it.
hf download onnx-community/silero-vadUses the Hugging Face CLI (pip install -U huggingface_hub). This downloads the whole repo; add --include "<pattern>" to fetch only the files you need.
What's the lowest-latency fully local voice loop you've built?
A typical cascade is voice activity detection, a turn detector, streaming speech-to-text, a small language model and a fast text-to-speech voice.