pytube: [BUG] KeyError 'start' when getting captions from a video

Apparently YouTube changed their captions format.

Here’s my version of the function for the function in captions.py:

def xml_caption_to_srt(self, xml_captions: str) -> str:
      """Convert xml caption tracks to "SubRip Subtitle (srt)".

      :param str xml_captions:
          XML formatted caption tracks.
      """
      segments = []
      root = ElementTree.fromstring(xml_captions)[1]
      i=0
      for child in list(root):
          if child.tag == 'p':
              caption = ''
              if len(list(child))==0:
                  continue
              for s in list(child):
                  if s.tag == 's':
                      caption += ' ' + s.text
              caption = unescape(caption.replace("\n", " ").replace("  ", " "),)
              try:
                  duration = float(child.attrib["d"])/1000.0
              except KeyError:
                  duration = 0.0
              start = float(child.attrib["t"])/1000.0
              end = start + duration
              sequence_number = i + 1  # convert from 0-indexed to 1.
              line = "{seq}\n{start} --> {end}\n{text}\n".format(
                  seq=sequence_number,
                  start=self.float_to_srt_time_format(start),
                  end=self.float_to_srt_time_format(end),
                  text=caption,
              )
              segments.append(line)
              i += 1
      return "\n".join(segments).strip()

@maksimbolonkin I changed a bit your code, so now it works for me as well 👌 The issue I’ve found was caused by the fact that some captions are located inside of <p> tag

    def xml_caption_to_srt(self, xml_captions: str) -> str:
        """Convert xml caption tracks to "SubRip Subtitle (srt)".

        :param str xml_captions:
        XML formatted caption tracks.
        """
        segments = []
        root = ElementTree.fromstring(xml_captions)
        i=0
        for child in list(root.iter("body"))[0]:
            if child.tag == 'p':
                caption = ''
                if len(list(child))==0:
                    # instead of 'continue'
                    caption = child.text
                for s in list(child):
                    if s.tag == 's':
                        caption += ' ' + s.text
                caption = unescape(caption.replace("\n", " ").replace("  ", " "),)
                try:
                    duration = float(child.attrib["d"])/1000.0
                except KeyError:
                    duration = 0.0
                start = float(child.attrib["t"])/1000.0
                end = start + duration
                sequence_number = i + 1  # convert from 0-indexed to 1.
                line = "{seq}\n{start} --> {end}\n{text}\n".format(
                    seq=sequence_number,
                    start=self.float_to_srt_time_format(start),
                    end=self.float_to_srt_time_format(end),
                    text=caption,
                )
                segments.append(line)
                i += 1
        return "\n".join(segments).strip()

So to fix this bug we can just replace xml_caption_to_srt inside of pytube/captions.py/Caption class with current code. Hope it’s gonna work for everyone 👍

I was trying to download videos from here: https://www.youtube.com/watch?v=gqaHkPEZAew&list=PLoROMvodv4rOSH4v6133s9LFPRHjEmbmJ

The code snippet above didn’t work out for me. So I had to modify it a little. Here’s the code that worked for me just in case anyone needs it:

def xml_caption_to_srt(self, xml_captions: str) -> str:
    """Convert xml caption tracks to "SubRip Subtitle (srt)".

    :param str xml_captions:
        XML formatted caption tracks.
    """
    segments = []
    root = ElementTree.fromstring(xml_captions)[0]
    i=0
    for child in list(root):
        if child.tag == 'p':
            caption = child.text
            caption = unescape(caption.replace("\n", " ").replace("  ", " "),)
            try:
                duration = float(child.attrib["d"])/1000.0
            except KeyError:
                duration = 0.0
            start = float(child.attrib["t"])/1000.0
            end = start + duration
            sequence_number = i + 1  # convert from 0-indexed to 1.
            line = "{seq}\n{start} --> {end}\n{text}\n".format(
                seq=sequence_number,
                start=self.float_to_srt_time_format(start),
                end=self.float_to_srt_time_format(end),
                text=caption,
            )
            segments.append(line)
            i += 1
    return "\n".join(segments).strip()

I have the same problem. Any solutions yet?

@maksimbolonkin thanks for the suggestion, I am using it.

I edited the code very minimally to make it work. This seems to work for me. Please Check for any issues. -> Added .find(“body”) -> Changed the keys from “dur” to “d” and “start” to “t” -> Divided the time by 1000. (to get seconds from miliseconds.)

def xml_caption_to_srt(self, xml_captions: str) -> str:
    """Convert xml caption tracks to "SubRip Subtitle (srt)".

    :param str xml_captions:
        XML formatted caption tracks.
    """
    segments = []
    root = ElementTree.fromstring(xml_captions).find('body')
    for i, child in enumerate(list(root)):
        text = child.text or ""
        caption = unescape(text.replace("\n", " ").replace("  ", " "),)
        try:
            duration = float(child.attrib["d"])/1000
        except KeyError:
            duration = 0.0
        start = float(child.attrib["t"])/1000
        end = start + duration
        sequence_number = i + 1  # convert from 0-indexed to 1.
        line = "{seq}\n{start} --> {end}\n{text}\n".format(
            seq=sequence_number,
            start= self.float_to_srt_time_format(start),
            end= self.float_to_srt_time_format(end),
            text=caption,
        )
        segments.append(line)
    return "\n".join(segments).strip()

邮件已收到，通常我在1～2天内回复。如有急事，请直接电话联系 ----我的微信 haijun-data

@maksimbolonkin I changed a bit your code, so now it works for me as well 👌 The issue I’ve found was caused by the fact that some captions are located inside of <p> tag

    def xml_caption_to_srt(self, xml_captions: str) -> str:
        """Convert xml caption tracks to "SubRip Subtitle (srt)".

        :param str xml_captions:
        XML formatted caption tracks.
        """
        segments = []
        root = ElementTree.fromstring(xml_captions)
        i=0
        for child in list(root.iter("body"))[0]:
            if child.tag == 'p':
                caption = ''
                if len(list(child))==0:
                    # instead of 'continue'
                    caption = child.text
                for s in list(child):
                    if s.tag == 's':
                        caption += ' ' + s.text
                caption = unescape(caption.replace("\n", " ").replace("  ", " "),)
                try:
                    duration = float(child.attrib["d"])/1000.0
                except KeyError:
                    duration = 0.0
                start = float(child.attrib["t"])/1000.0
                end = start + duration
                sequence_number = i + 1  # convert from 0-indexed to 1.
                line = "{seq}\n{start} --> {end}\n{text}\n".format(
                    seq=sequence_number,
                    start=self.float_to_srt_time_format(start),
                    end=self.float_to_srt_time_format(end),
                    text=caption,
                )
                segments.append(line)
                i += 1
        return "\n".join(segments).strip()

So to fix this bug we can just replace xml_caption_to_srt inside of pytube/captions.py/Caption class with current code. Hope it’s gonna work for everyone 👍

Thx… works for me