commit: cf1a8668e8e47a56c834fb567d227787d7480d08
parent 2dbb45ae82836699486d434cfb6d902920bab66e
Author: Joost Verdoorn <jpverdoorn@gmail.com>
Date: Thu, 19 Nov 2020 17:26:53 +0100
[Amara] Add new extractor (#20618)
* [Amara] Add new extractor
Diffstat:
2 files changed, 77 insertions(+), 0 deletions(-)
diff --git a/youtube_dl/extractor/amara.py b/youtube_dl/extractor/amara.py
@@ -0,0 +1,76 @@
+# coding: utf-8
+from __future__ import unicode_literals
+from .common import InfoExtractor
+
+
+class AmaraIE(InfoExtractor):
+ _VALID_URL = r'https?://(?:www\.)?amara\.org/(?:\w+/)?videos/(?P<id>\w+)'
+ _TESTS = [
+ {
+ 'url': 'https://amara.org/en/videos/jVx79ZKGK1ky/info/why-jury-trials-are-becoming-less-common/?tab=video',
+ 'md5': 'ea10daf2b6154b8c1ecf9922aca5e8ae',
+ 'info_dict': {
+ 'id': 'h6ZuVdvYnfE',
+ 'ext': 'mp4',
+ 'title': 'Why jury trials are becoming less common',
+ 'description': 'md5:a61811c319943960b6ab1c23e0cbc2c1',
+ 'thumbnail': r're:^https?://.*\.jpg$',
+ 'subtitles': dict,
+ 'upload_date': '20160813',
+ 'uploader': 'PBS NewsHour',
+ 'uploader_id': 'PBSNewsHour'
+ }
+ },
+ {
+ 'url': 'https://amara.org/en/videos/kYkK1VUTWW5I/info/vimeo-at-ces-2011',
+ 'md5': '99392c75fa05d432a8f11df03612195e',
+ 'info_dict': {
+ 'id': '18622084',
+ 'ext': 'mov',
+ 'title': 'Vimeo at CES 2011!',
+ 'description': 'md5:d41d8cd98f00b204e9800998ecf8427e',
+ 'thumbnail': r're:^https?://.*\.jpg$',
+ 'subtitles': dict,
+ 'timestamp': 1294649110,
+ 'upload_date': '20110110',
+ 'uploader': 'Sam Morrill',
+ 'uploader_id': 'sammorrill'
+ }
+ },
+ {
+ 'url': 'https://amara.org/en/videos/s8KL7I3jLmh6/info/the-danger-of-a-single-story/',
+ 'md5': 'd3970f08512738ee60c5807311ff5d3f',
+ 'info_dict': {
+ 'id': 'ChimamandaAdichie_2009G-transcript',
+ 'ext': 'mp4',
+ 'title': 'The danger of a single story',
+ 'description': 'md5:d769b31139c3b8bb5be9177f62ea3f23',
+ 'thumbnail': r're:^https?://.*\.jpg$',
+ 'subtitles': dict,
+ 'upload_date': '20131206'
+ }
+ }
+ ]
+
+ def get_subtitles_for_language(self, language):
+ return [{
+ 'ext': type,
+ 'url': language['subtitles_uri'].replace('format=json', 'format=' + type)
+ } for type in ['vtt', 'srt', 'json']]
+
+ def _real_extract(self, url):
+ video_id = self._match_id(url)
+ meta = self._download_json('https://amara.org/api/videos/%s/' % video_id, video_id, query={'format': 'json'})
+
+ video_url = meta.get('all_urls')[0]
+ subtitles = dict([(language['code'], self.get_subtitles_for_language(language)) for language in meta.get('languages', []) if language['published']])
+
+ return {
+ '_type': 'url_transparent',
+ 'url': video_url,
+ 'id': video_id,
+ 'subtitles': subtitles,
+ 'title': meta['title'],
+ 'description': meta.get('description'),
+ 'thumbnail': meta.get('thumbnail')
+ }
diff --git a/youtube_dl/extractor/extractors.py b/youtube_dl/extractor/extractors.py
@@ -36,6 +36,7 @@ from .afreecatv import AfreecaTVIE
from .airmozilla import AirMozillaIE
from .aljazeera import AlJazeeraIE
from .alphaporno import AlphaPornoIE
+from .amara import AmaraIE
from .amcnetworks import AMCNetworksIE
from .americastestkitchen import AmericasTestKitchenIE
from .animeondemand import AnimeOnDemandIE