yt_dlp/postprocessor/sponsorblock.py

   1 import hashlib
   2 import json
   3 import re
   4
   5 from .ffmpeg import FFmpegPostProcessor
   6 from ..compat import compat_urllib_parse_urlencode
   7
   8
   9 class SponsorBlockPP(FFmpegPostProcessor):
  10     # https://wiki.sponsor.ajay.app/w/Types
  11     EXTRACTORS = {
  12         'Youtube': 'YouTube',
  13     }
  14     POI_CATEGORIES = {
  15         'poi_highlight': 'Highlight',
  16     }
  17     CATEGORIES = {
  18         'sponsor': 'Sponsor',
  19         'intro': 'Intermission/Intro Animation',
  20         'outro': 'Endcards/Credits',
  21         'selfpromo': 'Unpaid/Self Promotion',
  22         'preview': 'Preview/Recap',
  23         'filler': 'Filler Tangent',
  24         'interaction': 'Interaction Reminder',
  25         'music_offtopic': 'Non-Music Section',
  26         **POI_CATEGORIES,
  27     }
  28
  29     def __init__(self, downloader, categories=None, api='https://sponsor.ajay.app'):
  30         FFmpegPostProcessor.__init__(self, downloader)
  31         self._categories = tuple(categories or self.CATEGORIES.keys())
  32         self._API_URL = api if re.match('^https?://', api) else 'https://' + api
  33
  34     def run(self, info):
  35         extractor = info['extractor_key']
  36         if extractor not in self.EXTRACTORS:
  37             self.to_screen(f'SponsorBlock is not supported for {extractor}')
  38             return [], info
  39
  40         self.to_screen('Fetching SponsorBlock segments')
  41         info['sponsorblock_chapters'] = self._get_sponsor_chapters(info, info['duration'])
  42         return [], info
  43
  44     def _get_sponsor_chapters(self, info, duration):
  45         segments = self._get_sponsor_segments(info['id'], self.EXTRACTORS[info['extractor_key']])
  46
  47         def duration_filter(s):
  48             start_end = s['segment']
  49             # Ignore entire video segments (https://wiki.sponsor.ajay.app/w/Types).
  50             if start_end == (0, 0):
  51                 return False
  52             # Ignore milliseconds difference at the start.
  53             if start_end[0] <= 1:
  54                 start_end[0] = 0
  55             # Make POI chapters 1 sec so that we can properly mark them
  56             if s['category'] in self.POI_CATEGORIES.keys():
  57                 start_end[1] += 1
  58             # Ignore milliseconds difference at the end.
  59             # Never allow the segment to exceed the video.
  60             if duration and duration - start_end[1] <= 1:
  61                 start_end[1] = duration
  62             # SponsorBlock duration may be absent or it may deviate from the real one.
  63             return s['videoDuration'] == 0 or not duration or abs(duration - s['videoDuration']) <= 1
  64
  65         duration_match = [s for s in segments if duration_filter(s)]
  66         if len(duration_match) != len(segments):
  67             self.report_warning('Some SponsorBlock segments are from a video of different duration, maybe from an old version of this video')
  68
  69         def to_chapter(s):
  70             (start, end), cat = s['segment'], s['category']
  71             return {
  72                 'start_time': start,
  73                 'end_time': end,
  74                 'category': cat,
  75                 'title': self.CATEGORIES[cat],
  76                 '_categories': [(cat, start, end)]
  77             }
  78
  79         sponsor_chapters = [to_chapter(s) for s in duration_match]
  80         if not sponsor_chapters:
  81             self.to_screen('No segments were found in the SponsorBlock database')
  82         else:
  83             self.to_screen(f'Found {len(sponsor_chapters)} segments in the SponsorBlock database')
  84         return sponsor_chapters
  85
  86     def _get_sponsor_segments(self, video_id, service):
  87         hash = hashlib.sha256(video_id.encode('ascii')).hexdigest()
  88         # SponsorBlock API recommends using first 4 hash characters.
  89         url = f'{self._API_URL}/api/skipSegments/{hash[:4]}?' + compat_urllib_parse_urlencode({
  90             'service': service,
  91             'categories': json.dumps(self._categories),
  92             'actionTypes': json.dumps(['skip', 'poi'])
  93         })
  94         for d in self._download_json(url) or []:
  95             if d['videoID'] == video_id:
  96                 return d['segments']
  97         return []