Skip to content
cc-toolkit

scribe

scribe turns YouTube videos and local recordings into text, then into the summary, news digest, or notes you ask for. It saves each transcript and reuses it on later runs, so you pay for transcription once per video or file. This plugin was formerly called yt.

Install

Section titled “Install”
/plugin install scribe@cc-toolkit

It also needs:

  • Node.js 20 or later
  • ffmpeg
  • ELEVENLABS_API_KEY set in your shell, for the paid ElevenLabs Scribe API
  • yt-dlp, for YouTube videos only

scribe talks to you and asks its questions in Polish. Summaries and notes follow the language of your request, and fall back to Polish when it’s unclear.

What it does

Section titled “What it does”
Command or skill What it does
/scribe:yt Transcribes and processes one or more YouTube videos end to end. Starts on its own when you paste YouTube links with a request such as “summarize these videos.”
/scribe:local Transcribes and processes audio or video files on disk end to end. Accepts a file, a glob, or a folder. Starts on its own when you give file paths with a request.
/scribe:yt-transcribe Downloads and transcribes YouTube audio only, without processing.
/scribe:local-transcribe Transcribes local files only, without processing.
/scribe:transcript-process Runs a summary, news extraction, or your own prompt on an existing transcript.

Only yt and local save transcripts to transcripts/ and reuse them on later runs. They also write one report per run, such as yt-analysis-<date>-<time>.md, in the directory where you run Claude Code.

Usage

Section titled “Usage”
  • Paste a YouTube link and ask for a summary. scribe checks the cache, transcribes new videos, and asks how to process them.
  • Run /scribe:local Transcribe everything in ~/Recordings/. It picks up audio and video files that are direct children of the folder.
  • Run /scribe:yt with several links and ask what’s new in a field. It extracts news from that field for each video.
  • Ask again about a video you already transcribed. scribe asks whether to reuse the cached transcript.

The plugin README has every option.