transcribing-youtube-videos
Made by JeffTranscribe YouTube videos and podcasts to timestamped, deep-linked text via yt-dlp + franken_whisper. Use when transcribing a video, podcast, or interview, or pulling clickable-timestamp quotes.
From Jeffrey's Skills.md: professionally crafted skills for any AI agent, installed and kept in sync by the jsm CLI. Subscription only, $20/month.
Works With
Quality and Utility come from canonical skill metadata. Community ratings appear separately as stars when subscriber reviews exist.
Subscribe to download
Public previews are available, but downloads, CLI sync, and premium content require an active subscription.
Sign In to Subscribe- Full skill content with deterministic versioning
- CLI sync + local search index
- Access to Jeffrey’s premium library
Save to your account
Saved skills live in your cloud list. Run jsm sync on each machine to pull them locally.
What's New with this Skill
1Updates, enhancements, and release notes for this skill.
New skill: transcribing-youtube-videos
Auto-captions hand you text with no speakers, no punctuation, every filler intact, and no way for a reader to check it against the source. This skill's unit of output is a verbatim quote plus a URL that opens the video at the second it was spoken. It re-transcribes with the native Rust Whisper stack, attributes speakers with in-process Sortformer diarization (no HuggingFace token, no Python), strips fillers without dropping content, and keeps the deep link attached throughout.
- The payoff is the audit trail: every quote carries a &t=<sec>s deep link, so a reader lands on the exact moment instead of taking your word for it
- Captions-only mode when the publisher shipped real captions (no model, no ASR), and yt-dlp reaches 1000+ sites, so this works well beyond YouTube
- The engine returns ASR and diarization on two unjoined timelines: segments[].speaker is null for most segments while the real signal sits in diarization.turns[], and fuse_diarization.py does the join at word granularity
- Overlap-voting alone misattributes roughly one word at every speaker change, which lands disproportionately on the short confirmations that turn edges are made of; snapping each change to sentence punctuation fixes it without a timing model
- --split-on-word gives the word-level timing the speaker fusion needs, but it makes every word its own segment, so feeding that run to the caption builder yields one deep link per word instead of per segment. Take links from the fused turn starts instead
- Long videos need two stage budgets raised rather than one: diarization is its own stage and the slower half of a run, so budgeting only the ASR stage leaves a two-hour podcast to die in diarization after the transcription succeeded
Covers both paths: fw youtube "<URL>" for the one-command case, and a yt-dlp + transcribe pipeline for non-YouTube sources and older builds. Captions-only mode skips ASR entirely when the publisher shipped real captions. Cleaning, speaker imputation from the video's own metadata, and the four-round audit are all self-contained in the skill.