Convert googles audio transcription xml to sbv

youtube_transcribe.py

#!/usr/bin/env python
# encoding: utf-8

import sys
import os.path

import HTMLParser
from xml.sax.saxutils import unescape
from xml.dom import pulldom

MINS = 60

def fmt_time(t):
    h = t / 3600
    t = t % 3600
    m = t / 60
    t = t % 60
    return "%d:%02d:%02.3f" % (h, m, t)


def read_elements(filename):
    """ yields (start_second, end_second, text) """
    events = pulldom.parse(filename)
    for event, node in events:
        if event == 'START_ELEMENT' and node.tagName=='text':
            text = ""

            start = float(node.getAttribute('start'))
            dur = float(node.getAttribute('dur'))
            end = start + dur

        elif event == 'CHARACTERS':
            text += node.data

        elif event == 'END_ELEMENT':
            yield start, end, text


def convert_times(elements):
    for s, e, txt in elements:
        yield fmt_time(s), fmt_time(e), txt


def write_elements(elements, outfile):
    p = HTMLParser.HTMLParser()
    with open(outfile, 'w') as f:
        for s, e, txt in elements:
            f.write(s)
            f.write(",")
            f.write(e)
            f.write("\n")
            f.write(p.unescape(txt))
            f.write("\n\n")



TEST_FILE = "testfile_transscript.xml"

def read_url(url, outfile=None):
    import tempfile
    import requests

    if not outfile:
        temp = temfile.TemporaryFile()
        outfile = temp.name

    with open(outfile, 'w') as f:
        f.write(requests.get(url).text)

    return outfile


if __name__ == '__main__':
    args = sys.argv
    outfile = "outfile.sbv"

    if len(args) > 1:
        arg = args[1]
        if arg.startswith("http"):
            #filename = read_url(arg)
            filename = TEST_FILE
        else:
            filename = arg

        if len(args) > 2:
            outfile = args[2]

        if os.path.exists(filename):
            elems = read_elements(filename)
            elems = convert_times(elems)
            write_elements(elems, outfile)
        else:
            print "No such file: " + filename

    else:
        print """Usage: youtube_trascribe.py filename.xml [outfile.sbv]"""