Origin nodes master canonical captions (speech recognition, CEA-608/708, pushCaptions) into signed MUXL WebVTT text tracks via muxl's per-GoP text hook. Every node extracts canonical cues from validated segments into the caption hub; nodes caption for accessibility when policy allows and share sidecars with relays over the segment websocket. Transcript records follow the live session, VOD captions merge MUXL text tracks with records, and the web and app players, settings, dashboard, VOD tools, docs, Playwright and Maestro flows cover the feature. muxl is pinned to the opus55/text-tracks commit until it is released.