Examples and Dependencies

• Dependencies
• Examples

Dependencies

We are keeping the colab examples up-to-date, but you can manually manage your dependencies:

• pytorch >= 1.9.0
• torchaudio >= 0.9.0 (used only for examples, IO and resampling, can be omitted in production)

The provided JIT-models can be run with other torch backends as well.

Examples

Imports

SAMPLE_RATE = 16000

import glob
import torch
torch.set_num_threads(1)

from IPython.display import Audio
from pprint import pprint

model, utils = torch.hub.load(repo_or_dir='snakers4/silero-vad',
                              model='silero_vad',
                              force_reload=True)

(get_speech_timestamps,
 save_audio,
 read_audio,
 VADIterator,
 collect_chunks) = utils

files_dir = torch.hub.get_dir() + '/snakers4_silero-vad_master/files'

Speech timestapms from full audio

wav = read_audio(f'{files_dir}/en.wav', sampling_rate=SAMPLE_RATE)
# get speech timestamps from full audio file
speech_timestamps = get_speech_timestamps(wav, model, sampling_rate=SAMPLE_RATE)
pprint(speech_timestamps)

# merge all speech chunks to one audio
save_audio('only_speech.wav',
           collect_chunks(speech_timestamps, wav), sampling_rate=16000) 
Audio('only_speech.wav')

Stream imitation example

## using VADIterator class

vad_iterator = VADiterator(double_model)
wav = read_audio((f'{files_dir}/en.wav', sampling_rate=SAMPLE_RATE)

window_size_samples = 1536 # number of samples in a single audio chunk
for i in range(0, len(wav), window_size_samples):
    speech_dict = vad_iterator(wav[i: i+ window_size_samples], return_seconds=True)
    if speech_dict:
        print(speech_dict, end=' ')
vad_iterator.reset_states() # reset model states after each audio

## just probabilities

wav = read_audio((f'{files_dir}/en.wav', sampling_rate=SAMPLE_RATE)
speech_probs = []
window_size_samples = 1536
for i in range(0, len(wav), window_size_samples):
    speech_prob = model(wav[i: i+ window_size_samples], SAMPLE_RATE).item()
    speech_probs.append(speech_prob)
model.reset_states()
pprint(speech_probs[:100])