Merge remote-tracking branch 'origin/master' into stream_selectors

examples/requirements.txt

@@ -0,0 +1,2 @@
+ffmpeg-python
+google-cloud-speech

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)

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)

ffmpeg/_filters.py

@@ -52,6 +52,11 @@ def split(stream):
     return FilterNode(stream, split.__name__)
 
 
+@filter_operator()
+def asplit(stream):
+    return FilterNode(stream, asplit.__name__)
+
+
 @filter_operator()
 def setpts(stream, expr):
     """Change the PTS (presentation timestamp) of the input frames.
@@ -421,6 +426,7 @@ __all__ = [
     'drawbox',
     'drawtext',
     'filter_',
+    'filter_multi_output',
     'hflip',
     'hue',
     'overlay',

ffmpeg/_view.py

@@ -78,7 +78,7 @@ def view(stream_spec, **kwargs):
             downstream_node_id = str(hash(edge.downstream_node))
             graph.edge(upstream_node_id, downstream_node_id, **kwargs)
 
-    graph.view(filename)
+    graph.view(filename, cleanup=True)
 
     return stream_spec
 

ffmpeg/nodes.py

@@ -280,7 +280,7 @@ class FilterNode(Node):
     def _get_filter(self, outgoing_edges):
         args = self.args
         kwargs = self.kwargs
-        if self.name == 'split':
+        if self.name in ('split', 'asplit'):
             args = [len(outgoing_edges)]
 
         out_args = [escape_chars(x, '\\\'=:') for x in args]

ffmpeg/tests/test_ffmpeg.py

@@ -180,6 +180,41 @@ def test_map_same_effect_as_output():
                                                       TEST_OUTPUT_FILE1]
 
 
+def _get_complex_filter_asplit_example():
+    split = (ffmpeg
+        .input(TEST_INPUT_FILE1)
+        .vflip()
+        .asplit()
+    )
+    split0 = split[0]
+    split1 = split[1]
+
+    return (ffmpeg
+        .concat(
+            split0.filter_("atrim", start=10, end=20),
+            split1.filter_("atrim", start=30, end=40),
+        )
+        .output(TEST_OUTPUT_FILE1)
+        .overwrite_output()
+    )
+
+
+def test_filter_asplit():
+    out = _get_complex_filter_asplit_example()
+    args = out.get_args()
+    assert args == [
+        '-i',
+        TEST_INPUT_FILE1,
+        '-filter_complex',
+        '[0]vflip[s0];[s0]asplit=2[s1][s2];[s1]atrim=end=20:start=10[s3];[s2]atrim=end=40:start=30[s4];[s3]'
+        '[s4]concat=n=2[s5]',
+        '-map',
+        '[s5]',
+        TEST_OUTPUT_FILE1,
+        '-y'
+    ]
+
+
 def test_filter_normal_arg_escape():
     """Test string escaping of normal filter args (e.g. ``font`` param of ``drawtext`` filter)."""
     def _get_drawtext_font_repr(font):

setup.py

@@ -2,7 +2,7 @@ from setuptools import setup
 from textwrap import dedent
 import subprocess
 
-version = '0.1.9'
+version = '0.1.10'
 download_url = 'https://github.com/kkroening/ffmpeg-python/archive/v{}.zip'.format(version)
 
 long_description = dedent("""\