silero-vad-onnx

Silero VAD (ONNX)

Silero Voice Activity Detection (VAD) model in ONNX format. Detects when a person is speaking in an audio stream.

Original source: snakers4/silero-vad

This is an exact copy of onnx-community/silero-vad.

Model variants

Usage

With ONNX Runtime

import numpy as np
import onnxruntime as ort

session = ort.InferenceSession("onnx/model_int8.onnx")
input_name = session.get_inputs()[0].name
sr_name = session.get_inputs()[1].name

audio = np.random.randn(1, 512).astype(np.float32)
sample_rate = np.array([16000], dtype=np.int64)

outputs = session.run(None, {input_name: audio, sr_name: sample_rate})
speech_prob = outputs[0]

With silero-vad library

import torch
torch.set_num_threads(1)

model, utils = torch.hub.load(
    repo_or_dir="snakers4/silero-vad",
    model="silero_vad",
)
get_speech_timestamps, _, read_audio, _, _ = utils

wav = read_audio("audio.wav")
speech_timestamps = get_speech_timestamps(wav, model, return_seconds=True)

With faster-whisper (built-in VAD)

from faster_whisper import WhisperModel

model = WhisperModel("small", device="cuda", compute_type="int8", vad_filter=True)
segments, _ = model.transcribe("audio.mp3")

Key features

• Fast: ~1ms per 30ms audio chunk on a single CPU thread
• Lightweight: ~2 MB model size
• Accurate: trained on large corpora covering 6000+ languages
• Flexible: supports 8kHz and 16kHz sampling rates
• Portable: runs anywhere ON Runtime is available

License

MIT License (same as snakers4/silero-vad)

Repository

Hugging Face: mijuanlo/silero-vad-onnx