mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-23 19:35:19 +01:00
33 lines
1.1 KiB
Python
33 lines
1.1 KiB
Python
|
# coding: utf-8
|
||
|
from __future__ import unicode_literals
|
||
|
|
||
|
from .common import InfoExtractor
|
||
|
|
||
|
|
||
|
class HungamaIE(InfoExtractor):
|
||
|
_VALID_URL = r'https?://(?:www\.)hungama\.com/song/[\w\d-]+/(?P<id>[0-9]+)'
|
||
|
_TEST = {
|
||
|
'url': 'https://www.hungama.com/song/kitni-haseen-zindagi/2931166/',
|
||
|
'md5': '396fa7e8e7e67aa25da0edc4cac9b785',
|
||
|
'info_dict': {
|
||
|
'id': '2931166',
|
||
|
'ext': 'mp4',
|
||
|
'title': 'Kitni Haseen Zindagi',
|
||
|
}
|
||
|
}
|
||
|
|
||
|
def _real_extract(self, url):
|
||
|
video_id = self._match_id(url)
|
||
|
webpage = self._download_webpage(url, video_id)
|
||
|
|
||
|
player_data = self._download_json('https://www.hungama.com/audio-player-data/track/%s?_country=IN' % video_id, video_id)[0]
|
||
|
title = player_data.get('song_name') or self._og_search_title(webpage)
|
||
|
track_data = self._download_json(player_data['file'], video_id)
|
||
|
media_url = track_data['response']['media_url']
|
||
|
|
||
|
return {
|
||
|
'id': video_id,
|
||
|
'title': title,
|
||
|
'formats': self._extract_m3u8_formats(media_url, video_id, ext='mp4'),
|
||
|
}
|