Services at a glance
YouTube has its own page. The other six services are smaller, but each has a quirk worth knowing before you open its code. All implement the same StreamingService contract; what differs is how they fetch and what they support.
SoundCloud (id 1)
Audio and comments. Talks to the api-v2.soundcloud.com REST API. The catch: there is no fixed API key, the client_id is scraped at runtime from SoundCloud's own frontend JavaScript (a ,client_id:"..." match in an sndcdn.com script) and appended to every call. Charts are a kiosk; it also has suggestions and subscription import. No video.
BiliBili (id 5)
Video, comments, bullet comments, and SponsorBlock, the most involved service. Stream URLs come from api.bilibili.com JSON, but most calls must be WBI-signed: it pulls a rotating permutation table from /x/web-interface/nav, builds a mixin key, and stamps requests with w_rid + wts. It also fabricates a cookie/ticket set (buvid3/4, b_lsid, bili_ticket, ...). Bullet comments (danmaku) arrive two ways: a zlib-compressed binary dump for VOD, and a WebSocket DANMU_MSG stream for live. Premium content uses a separate PGC endpoint. Adaptive (DASH) streams carry per-stream byte ranges (init + index), parsed from the SegmentBase Initialization / indexRange onto the audio and video tracks, so each plays as its own DASH representation. Expect the signing and table bits to need maintenance.
NicoNico (id 6)
Video and bullet comments (ordinary comments are disabled). Hybrid: it scrapes the watch page for a script#embedded-data JSON blob, then calls nvapi.nicovideo.jp with X-Frontend-Id headers. Playback is a negotiation: POST to .../access-rights/hls with the resolutions and audio you want, get back an encrypted HLS URL (gzip responses). Several content classes are login-gated.
PeerTube (id 3)
Video and comments, and the only federated one. There is no single backend: the service is bound to an instance (default framatube.org, user-configurable), and the instance URL is the API base (/api/v1/...). Plain JSON, no signing. Trending / Most-liked / Recent / Local are kiosks; channels expose CHANNELS and PLAYLISTS tabs.
Bandcamp (id 4)
Audio and comments. Mostly HTML scraping: album and track data come from a data-tralbum JSON attribute on the page. Two kiosks: Featured, and the weekly Radio show, the latter via a dedicated bcweekly API and its own radio stream extractor (selected by URL).
media.ccc.de (id 2)
Audio and video, no comments. A clean public JSON API (api.media.ccc.de/public/events/{id}). Content is organised around conferences: a conference is a channel, and the kiosks are conferences / recent / live. Two-stage fetch: event metadata first, then the parent conference via a URL in the event.
Quick reference
| Service | id | Media | Fetch | Watch out for |
|---|---|---|---|---|
| SoundCloud | 1 | audio, comments | api-v2 REST | dynamic client_id from JS |
| media.ccc.de | 2 | audio, video | public JSON API | conference-centric, no comments |
| PeerTube | 3 | video, comments | per-instance JSON | federated, instance = API base |
| Bandcamp | 4 | audio, comments | HTML scrape + JSON | data-tralbum blob, Radio show |
| BiliBili | 5 | video, comments, danmaku, SponsorBlock | api.bilibili.com JSON | WBI signing + cookie/ticket |
| NicoNico | 6 | video, danmaku | HTML + nvapi | login gate, HLS access-rights POST |