| name | sofit-kit |
| description | Generate the text kit — chapters, Hebrew show notes, and pull-quotes — for a Hebrew podcast episode (Weekly Sync) with sofit, formatted for YouTube / Spotify / podcast-app descriptions. Use when the user wants chapters, show notes, or quotes for an episode (not the video clips). Part of the sofit tool (see /sofit for the full pipeline). |
sofit-kit
Turns a (cached) transcript into text you paste into the episode description.
Home
HC=/Users/navotv/src/hebrew-chapters # repo dir (name kept; brand is "sofit")
SOFIT="$HC/.venv/bin/sofit" # CLI
Needs the transcript cached first — run /sofit-transcribe if it isn't (this step is fast).
Generation uses --titler claude-cli (shells out to claude -p, uses the Claude Code
subscription, no API key).
Run
"$SOFIT" "<episode.mp4>" --format spotify --shownotes --quotes --titler claude-cli
"$SOFIT" "<episode.mp4>" --format youtube --titler claude-cli
Formats: youtube (≥10s gaps), spotify (≥30s gaps, for Spotify/Megaphone episode
descriptions), podcast (Podcasting 2.0 JSON for RSS), md/txt (read), plus
--embed-into AUDIO (markers into a copy of the audio, Apple Podcasts).
Gotchas:
- Megaphone controls its own feed → PC2.0 JSON won't attach; the only path is pasting
the
spotify timestamp block into the episode description. Dynamic ad insertion desyncs later timestamps.
- YouTube output must NOT contain LRM/bidi marks (breaks YouTube's parser — the format handles this).
- Chapter timestamps come from Whisper, never the LLM (Claude picks the boundary quote; code locates + validates it).
