Show HN: Navigate by speaker in YouTube videos (zanshin.sh)

Zanshin is a new media player/extension that lets you navigate video and audio by speaker: it visualizes who speaks when and for how long, lets you jump or skip speaker segments, set different playback speeds per speaker, and auto-skip unwanted voices. It works with YouTube videos and local media files and is powered by Senko, described as a very fast speaker diarization pipeline. The interface emphasizes a speaker timeline that makes it easy to skim interviews, podcasts, press briefings, and panel discussions without scrubbing blindly. For the AI/ML community this is a practical consumer use of speaker diarization: accurate, low-latency speaker segmentation enables new interaction patterns (per-speaker speed control, segment jumping, and automatic filtering) that improve information consumption and accessibility. Key technical implications include reliance on robust diarization and clustering (speaker change detection, embedding consistency), transcript alignment if captions are used, and real-time segment playback control. Tools like Zanshin showcase how advances in diarization can be productized for better UX and hint at extensions — speaker ID, searchable speaker-based transcripts, summarization per speaker — while also raising deployment considerations around accuracy, latency, and privacy when processing hosted video.