colab_record_example.ipynb

Kernel: Python 3

Dependencies and inputs

#!apt install ffmpeg
!pip -q install pydub
from google.colab import output
from base64 import b64decode, b64encode
from io import BytesIO
import numpy as np
from pydub import AudioSegment
from IPython.display import HTML, display
import torch
import matplotlib.pyplot as plt
import moviepy.editor as mpe
from matplotlib.animation import FuncAnimation, FFMpegWriter
import matplotlib
matplotlib.use('Agg')

torch.set_num_threads(1)

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

def int2float(audio):
    samples = audio.get_array_of_samples()
    new_sound = audio._spawn(samples)
    arr = np.array(samples).astype(np.float32)
    arr = arr / np.abs(arr).max()
    return arr

AUDIO_HTML = """
<script>
var my_div = document.createElement("DIV");
var my_p = document.createElement("P");
var my_btn = document.createElement("BUTTON");
var t = document.createTextNode("Press to start recording");

my_btn.appendChild(t);
//my_p.appendChild(my_btn);
my_div.appendChild(my_btn);
document.body.appendChild(my_div);

var base64data = 0;
var reader;
var recorder, gumStream;
var recordButton = my_btn;

var handleSuccess = function(stream) {
  gumStream = stream;
  var options = {
    //bitsPerSecond: 8000, //chrome seems to ignore, always 48k
    mimeType : 'audio/webm;codecs=opus'
    //mimeType : 'audio/webm;codecs=pcm'
  };
  //recorder = new MediaRecorder(stream, options);
  recorder = new MediaRecorder(stream);
  recorder.ondataavailable = function(e) {
    var url = URL.createObjectURL(e.data);
    // var preview = document.createElement('audio');
    // preview.controls = true;
    // preview.src = url;
    // document.body.appendChild(preview);

    reader = new FileReader();
    reader.readAsDataURL(e.data);
    reader.onloadend = function() {
      base64data = reader.result;
      //console.log("Inside FileReader:" + base64data);
    }
  };
  recorder.start();
  };

recordButton.innerText = "Recording... press to stop";

navigator.mediaDevices.getUserMedia({audio: true}).then(handleSuccess);


function toggleRecording() {
  if (recorder && recorder.state == "recording") {
      recorder.stop();
      gumStream.getAudioTracks()[0].stop();
      recordButton.innerText = "Saving recording..."
  }
}

// https://stackoverflow.com/a/951057
function sleep(ms) {
  return new Promise(resolve => setTimeout(resolve, ms));
}

var data = new Promise(resolve=>{
//recordButton.addEventListener("click", toggleRecording);
recordButton.onclick = ()=>{
toggleRecording()

sleep(2000).then(() => {
  // wait 2000ms for the data to be available...
  // ideally this should use something like await...
  //console.log("Inside data:" + base64data)
  resolve(base64data.toString())

});

}
});

</script>
"""

def record(sec=10):
    display(HTML(AUDIO_HTML))
    s = output.eval_js("data")
    b = b64decode(s.split(',')[1])
    audio = AudioSegment.from_file(BytesIO(b))
    audio.export('test.mp3', format='mp3')
    audio = audio.set_channels(1)
    audio = audio.set_frame_rate(16000)
    audio_float = int2float(audio)
    audio_tens = torch.tensor(audio_float)
    return audio_tens

def make_animation(probs, audio_duration, interval=40):
    fig = plt.figure(figsize=(16, 9))
    ax = plt.axes(xlim=(0, audio_duration), ylim=(0, 1.02))
    line, = ax.plot([], [], lw=2)
    x = [i / 16000 * 512 for i in range(len(probs))]
    plt.xlabel('Time, seconds', fontsize=16)
    plt.ylabel('Speech Probability', fontsize=16)

    def init():
        plt.fill_between(x, probs, color='#064273')
        line.set_data([], [])
        line.set_color('#990000')
        return line,

    def animate(i):
        x = i * interval / 1000 - 0.04
        y = np.linspace(0, 1.02, 2)

        line.set_data(x, y)
        line.set_color('#990000')
        return line,
    anim = FuncAnimation(fig, animate, init_func=init, interval=interval, save_count=int(audio_duration / (interval / 1000)))

    f = r"animation.mp4"
    writervideo = FFMpegWriter(fps=1000/interval)
    anim.save(f, writer=writervideo)
    plt.close('all')

def combine_audio(vidname, audname, outname, fps=25):
    my_clip = mpe.VideoFileClip(vidname, verbose=False)
    audio_background = mpe.AudioFileClip(audname)
    final_clip = my_clip.set_audio(audio_background)
    final_clip.write_videofile(outname,fps=fps,verbose=False)

def record_make_animation():
  tensor = record()
  print('Calculating probabilities...')
  speech_probs = []
  window_size_samples = 512
  speech_probs = model.audio_forward(tensor, sr=16000)[0].tolist()
  model.reset_states()
  print('Making animation...')
  make_animation(speech_probs, len(tensor) / 16000)

  print('Merging your voice with animation...')
  combine_audio('animation.mp4', 'test.mp3', 'merged.mp4')
  print('Done!')
  mp4 = open('merged.mp4','rb').read()
  data_url = "data:video/mp4;base64," + b64encode(mp4).decode()
  display(HTML("""
  <video width=800 controls>
        <source src="%s" type="video/mp4">
  </video>
  """ % data_url))

  return speech_probs

Record example

speech_probs = record_make_animation()