[ie/mlbtv] Fix extractor (#10515)
[yt-dlp.git] / yt_dlp / postprocessor / sponsorblock.py
blob6cf9ab62eab5d43a0eb59edbfaf7314a716adc82
1 import hashlib
2 import json
3 import re
4 import urllib.parse
6 from .ffmpeg import FFmpegPostProcessor
9 class SponsorBlockPP(FFmpegPostProcessor):
10 # https://wiki.sponsor.ajay.app/w/Types
11 EXTRACTORS = {
12 'Youtube': 'YouTube',
14 POI_CATEGORIES = {
15 'poi_highlight': 'Highlight',
17 NON_SKIPPABLE_CATEGORIES = {
18 **POI_CATEGORIES,
19 'chapter': 'Chapter',
21 CATEGORIES = {
22 'sponsor': 'Sponsor',
23 'intro': 'Intermission/Intro Animation',
24 'outro': 'Endcards/Credits',
25 'selfpromo': 'Unpaid/Self Promotion',
26 'preview': 'Preview/Recap',
27 'filler': 'Filler Tangent',
28 'interaction': 'Interaction Reminder',
29 'music_offtopic': 'Non-Music Section',
30 **NON_SKIPPABLE_CATEGORIES,
33 def __init__(self, downloader, categories=None, api='https://sponsor.ajay.app'):
34 FFmpegPostProcessor.__init__(self, downloader)
35 self._categories = tuple(categories or self.CATEGORIES.keys())
36 self._API_URL = api if re.match('^https?://', api) else 'https://' + api
38 def run(self, info):
39 extractor = info['extractor_key']
40 if extractor not in self.EXTRACTORS:
41 self.to_screen(f'SponsorBlock is not supported for {extractor}')
42 return [], info
44 self.to_screen('Fetching SponsorBlock segments')
45 info['sponsorblock_chapters'] = self._get_sponsor_chapters(info, info.get('duration'))
46 return [], info
48 def _get_sponsor_chapters(self, info, duration):
49 segments = self._get_sponsor_segments(info['id'], self.EXTRACTORS[info['extractor_key']])
51 def duration_filter(s):
52 start_end = s['segment']
53 # Ignore entire video segments (https://wiki.sponsor.ajay.app/w/Types).
54 if start_end == (0, 0):
55 return False
56 # Ignore milliseconds difference at the start.
57 if start_end[0] <= 1:
58 start_end[0] = 0
59 # Make POI chapters 1 sec so that we can properly mark them
60 if s['category'] in self.POI_CATEGORIES:
61 start_end[1] += 1
62 # Ignore milliseconds difference at the end.
63 # Never allow the segment to exceed the video.
64 if duration and duration - start_end[1] <= 1:
65 start_end[1] = duration
66 # SponsorBlock duration may be absent or it may deviate from the real one.
67 diff = abs(duration - s['videoDuration']) if s['videoDuration'] else 0
68 return diff < 1 or (diff < 5 and diff / (start_end[1] - start_end[0]) < 0.05)
70 duration_match = [s for s in segments if duration_filter(s)]
71 if len(duration_match) != len(segments):
72 self.report_warning('Some SponsorBlock segments are from a video of different duration, maybe from an old version of this video')
74 def to_chapter(s):
75 (start, end), cat = s['segment'], s['category']
76 title = s['description'] if cat == 'chapter' else self.CATEGORIES[cat]
77 return {
78 'start_time': start,
79 'end_time': end,
80 'category': cat,
81 'title': title,
82 'type': s['actionType'],
83 '_categories': [(cat, start, end, title)],
86 sponsor_chapters = [to_chapter(s) for s in duration_match]
87 if not sponsor_chapters:
88 self.to_screen('No matching segments were found in the SponsorBlock database')
89 else:
90 self.to_screen(f'Found {len(sponsor_chapters)} segments in the SponsorBlock database')
91 return sponsor_chapters
93 def _get_sponsor_segments(self, video_id, service):
94 video_hash = hashlib.sha256(video_id.encode('ascii')).hexdigest()
95 # SponsorBlock API recommends using first 4 hash characters.
96 url = f'{self._API_URL}/api/skipSegments/{video_hash[:4]}?' + urllib.parse.urlencode({
97 'service': service,
98 'categories': json.dumps(self._categories),
99 'actionTypes': json.dumps(['skip', 'poi', 'chapter']),
101 for d in self._download_json(url) or []:
102 if d['videoID'] == video_id:
103 return d['segments']
104 return []