hebrew-podcast-postproduction
Installation
SKILL.md
Hebrew Podcast Post-Production Kit
Problem
Hebrew podcasters ship every episode through 4-5 disconnected tools: one for transcription, one for show notes, another for clip-hunting, another for social captions, plus FFmpeg for the actual cutting. Generic tools (Kapwing, Maestra, Descript, Podsqueeze) mishandle RTL text, drop nikud, and produce broken subtitle burn-ins because they do not run FriBidi shaping for Hebrew. This skill consolidates the entire post-production workflow into one coherent package so the agent outputs show notes, chapters, clips, and captions from a single transcript input.
Instructions
The workflow takes one of two inputs and produces one markdown bundle:
| Input | Required Fields |
|---|---|
Whisper JSON (local openai-whisper, or the API with whisper-1 + verbose_json) |
language, segments[] with start, end, text |
| ElevenLabs Scribe JSON | language_code and words[], each with text, start, end, type. Scribe returns no segments[] array. |
| SRT file + optional audio file path | timestamped cues, HH:MM:SS,mmm format |
Two input-shape traps that break the workflow before it starts: