def read(self, content, lang='en-US'): caption_set = CaptionSet() lines = content.splitlines() start_line = 0 captions = [] while start_line < len(lines): if not lines[start_line].isdigit(): break caption = Caption() end_line = self._find_text_line(start_line, lines) timing = lines[start_line + 1].split('-->') caption.start = self._srttomicro(timing[0].strip(' \r\n')) caption.end = self._srttomicro(timing[1].strip(' \r\n')) for line in lines[start_line + 2:end_line - 1]: # skip extra blank lines if not caption.nodes or line != '': caption.nodes.append(CaptionNode.create_text(line)) caption.nodes.append(CaptionNode.create_break()) # remove last line break from end of caption list caption.nodes.pop() captions.append(caption) start_line = end_line caption_set.set_captions(lang, captions) return caption_set
def _translate_lang(self, language, sami_soup): captions = [] milliseconds = 0 for p in sami_soup.select('p[lang|=%s]' % language): milliseconds = int(float(p.parent['start'])) start = milliseconds * 1000 end = 0 if captions != [] and captions[-1].end == 0: captions[-1].end = milliseconds * 1000 if p.get_text().strip(): self.line = [] self._translate_tag(p) text = self.line styles = self._translate_attrs(p) caption = Caption() caption.start = start caption.end = end caption.nodes = text caption.style = styles captions.append(caption) if captions and captions[-1].end == 0: # Arbitrarily make this last 4 seconds. Not ideal... captions[-1].end = (milliseconds + 4000) * 1000 return captions
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)
def _translate_p_tag(self, p_tag): start, end = self._find_times(p_tag) self.nodes = [] self._translate_tag(p_tag) styles = self._translate_style(p_tag) caption = Caption() caption.start = start caption.end = end caption.nodes = self.nodes caption.style = styles return caption
def _convert_to_caption(self, buffer, start): # check to see if previous caption needs an end-time if self.scc and self.scc[-1].end == 0: self.scc[-1].end = start # initial variables caption = Caption() caption.start = start caption.end = 0 # Not yet known; filled in later self.open_italic = False self.first_element = True # split into elements (e.g. break, italics, text) for element in buffer.split('<$>'): # skip empty elements if element.strip() == '': continue # handle line breaks elif element == '{break}': self._translate_break(caption) # handle open italics elif element == '{italic}': # add italics caption.nodes.append(CaptionData.create_style(True, {'italics': True})) # open italics, no longer first element self.open_italic = True self.first_element = False # handle clone italics elif element == '{end-italic}' and self.open_italic: caption.nodes.append(CaptionData.create_style(False, {'italics': True})) self.open_italic = False # handle text else: # add text caption.nodes.append(CaptionData.create_text(' '.join(element.decode("utf-8").split()))) # no longer first element self.first_element = False # close any open italics left over if self.open_italic == True: caption.nodes.append(CaptionData.create_style(False, {'italics': True})) # remove extraneous italics tags in the same caption self._remove_italics(caption) # only add captions to list if content inside exists if caption.nodes: self.scc.append(caption)
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)
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()
def __init__(self, raw_caption: Caption): self.raw_caption = raw_caption self.raw_text = raw_caption.get_text() self.lines = self.raw_text.split('\n') assert len(self.lines) > 0
def read(self, content, lang="en"): captions = CaptionSet() inlines = content.splitlines() start_line = 0 subdata = [] while start_line < len(inlines): if not inlines[start_line].isdigit(): break caption = Caption() end_line = self._find_text_line(start_line, inlines) timing = inlines[start_line + 1].split("-->") caption.start = self._srttomicro(timing[0].strip(" \r\n")) caption.end = self._srttomicro(timing[1].strip(" \r\n")) for line in inlines[start_line + 2 : end_line - 1]: caption.nodes.append(CaptionData.create_text(line)) caption.nodes.append(CaptionData.create_break()) caption.nodes.pop() # remove last line break from end of caption list subdata.append(caption) start_line = end_line captions.set_captions(lang, subdata) return captions