Пример #1
0
def run_pipeline(url=None,
                 hmm=None,
                 lm=None,
                 dict=None,
                 caption_format='webvtt',
                 out_file=None):
    if url is None:
        raise Exception('No URL specified!')
    pipeline = Gst.parse_launch('uridecodebin name=source ! audioconvert !' +
                                ' audioresample ! pocketsphinx name=asr !' +
                                ' fakesink')
    source = pipeline.get_by_name('source')
    source.set_property('uri', url)
    pocketsphinx = pipeline.get_by_name('asr')
    if hmm:
        pocketsphinx.set_property('hmm', hmm)
    if lm:
        pocketsphinx.set_property('lm', lm)
    if dict:
        pocketsphinx.set_property('dict', dict)

    bus = pipeline.get_bus()

    # Start playing
    pipeline.set_state(Gst.State.PLAYING)

    cap_set = CaptionSet()
    captions = []

    # Wait until error or EOS
    while True:
        try:
            msg = bus.timed_pop(Gst.CLOCK_TIME_NONE)
            if msg:
                #if msg.get_structure():
                #    print(msg.get_structure().to_string())

                if msg.type == Gst.MessageType.EOS:
                    break
                struct = msg.get_structure()
                if struct and struct.get_name() == 'pocketsphinx':
                    if struct['final']:
                        c = Caption()
                        c.start = struct['start_time'] / Gst.USECOND
                        c.end = struct['end_time'] / Gst.USECOND
                        c.nodes.append(
                            CaptionNode.create_text(struct['hypothesis']))
                        captions.append(c)
        except KeyboardInterrupt:
            pipeline.send_event(Gst.Event.new_eos())

    # Free resources
    pipeline.set_state(Gst.State.NULL)

    cap_set.set_captions('en-US', captions)
    writer = SRTWriter() if caption_format == 'srt' else WebVTTWriter()
    caption_data = writer.write(cap_set)
    if out_file is not None:
        codecs.open(out_file, 'w', 'utf-8').write(caption_data)
    else:
        print(caption_data)
Пример #2
0
stories = codecs.open('story.txt', 'r', 'utf-8').readlines()


def microsec(t):
    return t * 1000000


offset = 0.0
captions = []
for line in sys.stdin:
    if line.startswith(' '):
        continue
    tokens = line.split()
    if len(tokens) != 3:
        continue
    dirname = tokens[0]
    index = int(dirname.split('/')[-1]) - 1
    duration = float(tokens[2])
    print duration
    text = stories[index]
    cap = Caption(microsec(offset), microsec(offset + duration),
                  [CaptionNode.create_text(text)])
    offset += duration
    captions.append(cap)

caps = CaptionSet({'en': captions})

srt = codecs.open('output.srt', 'w', 'utf-8')
srt.write(SRTWriter().write(caps))
srt.close()