FFmpeg silence detect wrapper - prints human-readable timecodes

silencedetect.py

#!/usr/bin/env python3

import argparse
import re
import subprocess
import sys


def format_time(s):
    (h, s) = divmod(s, 3600)
    (m, s) = divmod(s, 60)
    h = int(h)
    m = int(m)
    # NOTE: Seconds are zero-padded to 8 characters
    # to account for the decimal point and fractional part
    return f'{h:02}:{m:02}:{s:08.5f}'


parser = argparse.ArgumentParser()
parser.add_argument(
    'file',
    help='input file',
)
parser.add_argument(
    '-n',
    '--noise',
    default=0.001,
    help='noise tolerance',
)
parser.add_argument(
    '-d',
    '--duration',
    default=2,
    help='silence duration until notification',
)
parser.add_argument(
    '-m',
    '--mono',
    action='store_true',
    help='process each channel separately',
)

args = parser.parse_args()

if args.mono:
    mono_arg = 'y'
else:
    mono_arg = 'n'

p = subprocess.Popen([
    'ffmpeg',
    '-i', args.file,
    '-af', f'silencedetect=noise={args.noise}:duration={args.duration}:mono={mono_arg}',
    '-f', 'null',
    '-',
], encoding='utf-8', errors='replace', stderr=subprocess.PIPE)

stderr = []

for line in p.stderr:
    line = line.rstrip('\n')
    stderr.append(line)

    if not line.startswith('[silencedetect'):
        continue

    m = re.search(r'silence_start: ([\d.]+)', line)
    if m:
        start = format_time(float(m.group(1)))
        print(f'silence: {start} - ', end='')

    m = re.search(r'silence_end: ([\d.]+) \| silence_duration: ([\d.]+)', line)
    if m:
        end = format_time(float(m.group(1)))
        duration = format_time(float(m.group(2)))
        print(f'{end} (duration: {duration})')

p.wait()
if p.returncode != 0:
    print('\n'.join(stderr))
    print(f'ffmpeg exited with return code {p.returncode}', file=sys.stderr)
    sys.exit(1)