kkroening · kkroening · Jan 14, 2018 · Jan 7, 2018 · Jan 7, 2018 · Jan 7, 2018
diff --git a/examples/requirements.txt b/examples/requirements.txt
@@ -0,0 +1 @@
+ffmpeg
diff --git a/examples/split_silence.py b/examples/split_silence.py
@@ -0,0 +1,141 @@
+#!/usr/bin/env python
+from __future__ import unicode_literals
+
+import argparse
+import errno
+import ffmpeg
+import logging
+import os
+import re
+import subprocess
+import sys
+
+
+logging.basicConfig(level=logging.INFO, format='%(message)s')
+logger = logging.getLogger(__file__)
+logger.setLevel(logging.INFO)
+
+DEFAULT_DURATION = 0.3
+DEFAULT_THRESHOLD = -60
+
+parser = argparse.ArgumentParser(description='Split media into separate chunks wherever silence occurs')
+parser.add_argument('in_filename', help='Input filename (`-` for stdin)')
+parser.add_argument('out_pattern', help='Output filename pattern (e.g. `out/chunk_{:04d}.wav`)')
+parser.add_argument('--silence-threshold', default=DEFAULT_THRESHOLD, type=int, help='Silence threshold (in dB)')
+parser.add_argument('--silence-duration', default=DEFAULT_DURATION, type=float, help='Silence duration')
+parser.add_argument('--start-time', type=float, help='Start time (seconds)')
+parser.add_argument('--end-time', type=float, help='End time (seconds)')
+parser.add_argument('-v', dest='verbose', action='store_true', help='Verbose mode')
+
+silence_start_re = re.compile(' silence_start: (?P<start>[0-9]+(\.?[0-9]*))$')
+silence_end_re = re.compile(' silence_end: (?P<end>[0-9]+(\.?[0-9]*)) ')
+total_duration_re = re.compile(
+    'size=[^ ]+ time=(?P<hours>[0-9]{2}):(?P<minutes>[0-9]{2}):(?P<seconds>[0-9\.]{5}) bitrate=')
+
+
+def _logged_popen(cmd_line, *args, **kwargs):
+    logger.debug('Running command: {}'.format(subprocess.list2cmdline(cmd_line)))
+    return subprocess.Popen(cmd_line, *args, **kwargs)
+
+
+def get_chunk_times(in_filename, silence_threshold, silence_duration, start_time=None, end_time=None):
+    input_kwargs = {}
+    if start_time is not None:
+        input_kwargs['ss'] = start_time
+    else:
+        start_time = 0.
+    if end_time is not None:
+        input_kwargs['t'] = end_time - start_time
+
+    p = _logged_popen(
+        (ffmpeg
+            .input(in_filename, **input_kwargs)
+            .filter_('silencedetect', n='{}dB'.format(silence_threshold), d=silence_duration)
+            .output('-', format='null')
+            .compile()
+        ) + ['-nostats'],  # FIXME: use .nostats() once it's implemented in ffmpeg-python.
+        stderr=subprocess.PIPE
+    )
+    output = p.communicate()[1].decode('utf-8')
+    if p.returncode != 0:
+        sys.stderr.write(output)
+        sys.exit(1)
+    logger.debug(output)
+    lines = output.splitlines()
+
+    # Chunks start when silence ends, and chunks end when silence starts.
+    chunk_starts = []
+    chunk_ends = []
+    for line in lines:
+        silence_start_match = silence_start_re.search(line)
+        silence_end_match = silence_end_re.search(line)
+        total_duration_match = total_duration_re.search(line)
+        if silence_start_match:
+            chunk_ends.append(float(silence_start_match.group('start')))
+            if len(chunk_starts) == 0:
+                # Started with non-silence.
+                chunk_starts.append(start_time or 0.)
+        elif silence_end_match:
+            chunk_starts.append(float(silence_end_match.group('end')))
+        elif total_duration_match:
+            hours = int(total_duration_match.group('hours'))
+            minutes = int(total_duration_match.group('minutes'))
+            seconds = float(total_duration_match.group('seconds'))
+            end_time = hours * 3600 + minutes * 60 + seconds
+
+    if len(chunk_starts) == 0:
+        # No silence found.
+        chunk_starts.append(start_time)
+
+    if len(chunk_starts) > len(chunk_ends):
+        # Finished with non-silence.
+        chunk_ends.append(end_time or 10000000.)
+
+    return list(zip(chunk_starts, chunk_ends))
+
+
+def _makedirs(path):
+    """Python2-compatible version of ``os.makedirs(path, exist_ok=True)``."""
+    try:
+        os.makedirs(path)
+    except OSError as exc:
+        if exc.errno != errno.EEXIST or not os.path.isdir(path):
+            raise
+
+
+def split_audio(
+    in_filename,
+    out_pattern,
+    silence_threshold=DEFAULT_THRESHOLD,
+    silence_duration=DEFAULT_DURATION,
+    start_time=None,
+    end_time=None,
+    verbose=False,
+):
+    chunk_times = get_chunk_times(in_filename, silence_threshold, silence_duration, start_time, end_time)
+
+    for i, (start_time, end_time) in enumerate(chunk_times):
+        time = end_time - start_time
+        out_filename = out_pattern.format(i, i=i)
+        _makedirs(os.path.dirname(out_filename))
+
+        logger.info('{}: start={:.02f}, end={:.02f}, duration={:.02f}'.format(out_filename, start_time, end_time,
+            time))
+        _logged_popen(
+            (ffmpeg
+                .input(in_filename, ss=start_time, t=time)
+                .output(out_filename)
+                .overwrite_output()
+                .compile()
+            ),
+            stdout=subprocess.PIPE if not verbose else None,
+            stderr=subprocess.PIPE if not verbose else None,
+        ).communicate()
+
+
+if __name__ == '__main__':
+    kwargs = vars(parser.parse_args())
+    if kwargs['verbose']:
+        logging.basicConfig(level=logging.DEBUG, format='%(levels): %(message)s')
+        logger.setLevel(logging.DEBUG)
+    split_audio(**kwargs)
diff --git a/examples/transcribe.py b/examples/transcribe.py
@@ -0,0 +1,62 @@
+#!/usr/bin/env python
+from __future__ import unicode_literals
+
+from google.cloud import speech
+from google.cloud.speech import enums
+from google.cloud.speech import types
+import argparse
+import ffmpeg
+import logging
+import subprocess
+import sys
+
+
+logging.basicConfig(level=logging.INFO, format='%(message)s')
+logger = logging.getLogger(__file__)
+logger.setLevel(logging.INFO)
+
+
+parser = argparse.ArgumentParser(description='Convert speech audio to text using Google Speech API')
+parser.add_argument('in_filename', help='Input filename (`-` for stdin)')
+
+
+def decode_audio(in_filename, **input_kwargs):
+    p = subprocess.Popen(
+        (ffmpeg
+            .input(in_filename, **input_kwargs)
+            .output('-', format='s16le', acodec='pcm_s16le', ac=1, ar='16k')
+            .overwrite_output()
+            .compile()
+        ),
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE
+    )
+    out = p.communicate()
+    if p.returncode != 0:
+        sys.stderr.write(out[1])
+        sys.exit(1)
+    return out[0]
+
+
+def get_transcripts(audio_data):
+    client = speech.SpeechClient()
+    audio = types.RecognitionAudio(content=audio_data)
+    config = types.RecognitionConfig(
+        encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
+        sample_rate_hertz=16000,
+        language_code='en-US'
+    )
+    response = client.recognize(config, audio)
+    return [result.alternatives[0].transcript for result in response.results]
+
+
+def transcribe(in_filename):
+    audio_data = decode_audio(in_filename)
+    transcripts = get_transcripts(audio_data)
+    for transcript in transcripts:
+        print(repr(transcript.encode('utf-8')))
+
+
+if __name__ == '__main__':
+    args = parser.parse_args()
+    transcribe(args.in_filename)
diff --git a/ffmpeg/_run.py b/ffmpeg/_run.py
@@ -131,22 +131,28 @@ def get_args(stream_spec, overwrite_output=False):
     return args
 
 
+@output_operator()
+def compile(stream_spec, cmd='ffmpeg', **kwargs):
+    """Build command-line for ffmpeg."""
+    if isinstance(cmd, basestring):
+        cmd = [cmd]
+    elif type(cmd) != list:
+        cmd = list(cmd)
+    return cmd + get_args(stream_spec, **kwargs)
+
+
 @output_operator()
 def run(stream_spec, cmd='ffmpeg', **kwargs):
     """Run ffmpeg on node graph.
 
     Args:
         **kwargs: keyword-arguments passed to ``get_args()`` (e.g. ``overwrite_output=True``).
     """
-    if isinstance(cmd, basestring):
-        cmd = [cmd]
-    elif type(cmd) != list:
-        cmd = list(cmd)
-    args = cmd + get_args(stream_spec, **kwargs)
-    _subprocess.check_call(args)
+    _subprocess.check_call(compile(stream_spec, cmd, **kwargs))
 
 
 __all__ = [
+    'compile',
     'get_args',
     'run',
 ]
diff --git a/ffmpeg/tests/test_ffmpeg.py b/ffmpeg/tests/test_ffmpeg.py
@@ -219,6 +219,12 @@ def _get_drawtext_text_repr(text):
 #    subprocess.check_call(['ffmpeg', '-version'])
 
 
+def test_compile():
+    out_file = ffmpeg.input('dummy.mp4').output('dummy2.mp4')
+    assert out_file.compile() == ['ffmpeg', '-i', 'dummy.mp4', 'dummy2.mp4']
+    assert out_file.compile(cmd='ffmpeg.old') == ['ffmpeg.old', '-i', 'dummy.mp4', 'dummy2.mp4']
+
+
 def test_run():
     stream = _get_complex_filter_example()
     ffmpeg.run(stream)