especificos-videos/back/audio_timings.py

#!/usr/bin/env python3

from datetime import timedelta
import pyaudio
import wave
import sys
import audioop
import requests
import subprocess
import os
import stat

def parse_file(fname, CHUNK=44100):
    outfile = "/tmp/asd"
    if os.path.isfile(outfile) or stat.S_ISFIFO(os.stat(outfile).st_mode):
        os.remove(outfile)
        print("Deleted")
    
    os.mkfifo(outfile)
    
    p = subprocess.Popen(["ffmpeg", "-nostdin", "-loglevel", "error",
                          "-i", fname, 
                          "-c:a", "pcm_s16le", "-vn", "-f", "wav", "-y", outfile])
    wf = wave.open(outfile, 'rb')
    
    p = pyaudio.PyAudio()
    width = wf.getsampwidth()
    data = wf.readframes(CHUNK)
    
    chunks = 0
    count = 0
    initial = 0
    ret = []
    while len(data) > 0:
        data = wf.readframes(CHUNK)
        rms = audioop.rms(data, width)
    
        # we need to get at least 60 values with rms < 700
        # that's 60 seconds
        if rms < 750:
            count += 1
        else:
            if count > 80:
                itime = timedelta(minutes=initial // 60, seconds=initial % 60)
                etime = timedelta(minutes=chunks // 60, seconds=chunks % 60)
                ret.append(itime)
                ret.append(etime)
            count = 0
            initial = chunks
    
        chunks += 1
    
    
    itime = timedelta(minutes=initial // 60, seconds=initial % 60)
    etime = timedelta(minutes=chunks // 60, seconds=chunks % 60)
    ret.append(itime)
    ret.append(etime)
    
    ret = ret[1:-1] # chop the initial and final blank time
    
    ret = sorted(list(set([t.total_seconds() for t in ret ])))
    p.terminate()
    return ret

if __name__ == '__main__':
    print(parse_file("processed/movie_20171006_093228.flv"))