tubescribe – OpenClaw Skill

---

name: tubescribe description: "YouTube video summarizer with speaker detection, formatted documents, and audio output. Use when user sends a YouTube URL or asks to summarize/transcribe a YouTube video."

TubeScribe 🎬

Turn any YouTube video into a polished document + audio summary in seconds.

Drop a YouTube link → get a beautiful transcript with speaker labels, key quotes, timestamps that link back to the video, and an audio summary you can listen to on the go.

💸 100% Free & Local

• No subscription — runs entirely on your machine
• No API keys required — works out of the box
• No data leaves your computer — your content stays private
• No usage limits — summarize as many videos as you want

✨ Features

• 🎯 Smart Speaker Detection — Automatically identifies who's talking in interviews, podcasts, and conversations
• 📝 Clickable Timestamps — Every quote links directly to that moment in the video
• 📄 Clean Documents — Export as HTML, DOCX, or Markdown
• 🔊 Audio Summaries — Listen to the key points (MP3/WAV)
• 🚀 Zero Config — Works out of the box, upgrades available for power users

🎬 Works With Any Video

• Interviews & podcasts (multi-speaker detection)
• Lectures & tutorials (single speaker)
• Music videos (lyrics extraction)
• News & documentaries
• Any YouTube content with captions

Quick Start

When user sends a YouTube URL, run the full pipeline automatically:

# 1. Extract transcript
python skills/tubescribe/scripts/tubescribe.py "YOUTUBE_URL"

This creates:

• /tmp/tubescribe_{video_id}_source.json — metadata + transcript
• /tmp/tubescribe_{video_id}_output.md — path for output

Then process with sub-agent (see workflow below).

First-Time Setup

Run setup to check dependencies and configure defaults:

python skills/tubescribe/scripts/setup.py

This checks: summarize CLI, pandoc/python-docx, ffmpeg, Kokoro TTS

Full Workflow

Step 1: Extract Transcript

python skills/tubescribe/scripts/tubescribe.py "https://youtube.com/watch?v=VIDEO_ID"

Step 2: Process with Sub-Agent

Spawn a sub-agent to analyze and format:

sessions_spawn(
    task="""Read /tmp/tubescribe_{video_id}_source.json and create formatted output.

**Output to:** /tmp/tubescribe_{video_id}_output.md

**Format:**
1. # Title (from metadata)
2. ## Participants — identify speakers from context
3. ## Summary — 3-5 paragraphs covering main topics
4. ## Key Quotes — 5 best quotes with timestamps [[MM:SS]](https://youtu.be/{video_id}?t=SECONDS)
5. ## Full Transcript — ALL segments with:
   - Speaker labels (**Name:** ) when identifiable
   - Clickable timestamps: [[0:42]](https://youtu.be/{video_id}?t=42)
   - Convert MM:SS to seconds for links

**Speaker Detection:**
- Use context clues (questions vs answers, explicit names, speaking patterns)
- For single-speaker videos, use narrator label or skip speaker labels
- For interviews: host asks questions, guest gives longer answers
""",
    label="tubescribe",
    runTimeoutSeconds=600,
    cleanup="delete"
)

Step 3: Create Document

Convert markdown to final format:

# HTML (no dependencies beyond Python)
python skills/tubescribe/scripts/html_writer.py /tmp/tubescribe_{video_id}_output.md output.html

# DOCX with pandoc (best formatting)
pandoc /tmp/tubescribe_{video_id}_output.md -o output.docx

# Markdown (just copy the file)
cp /tmp/tubescribe_{video_id}_output.md output.md

Step 4: Generate Audio Summary (Optional)

Extract summary section and generate TTS:

# Read summary from output markdown
# Generate audio using Kokoro (preferred) or built-in TTS
# Save to {output_dir}/{title}_summary.wav or .mp3

Step 5: Open Results

open output.html  # or .docx or .md
open -a "QuickTime Player" output_summary.wav

Configuration

Config file: ~/.tubescribe/config.json

{
  "output": {
    "folder": "~/Documents/TubeScribe",
    "open_folder_after": true
  },
  "document": {
    "format": "docx"
  },
  "audio": {
    "enabled": true,
    "format": "mp3",
    "tts_engine": "kokoro"
  }
}

Options:

• output.folder: Where to save files (default: ~/Documents/TubeScribe)
• document.format: html (default, no deps), docx (with pandoc/python-docx), md (raw markdown)
• audio.format: mp3 (with ffmpeg), wav (default without ffmpeg)
• audio.tts_engine: builtin (macOS say), kokoro (high quality)

Output Structure

~/Documents/TubeScribe/
├── {Video Title}.html         # Formatted document (or .docx / .md)
└── {Video Title}_summary.mp3  # Audio summary (or .wav)

After generation, opens the folder (not individual files) so you can access everything.

Dependencies

Required:

• summarize CLI — brew install steipete/tap/summarize
• Python 3.8+

Optional (better quality):

• pandoc — DOCX output: brew install pandoc
• ffmpeg — MP3 audio: brew install ffmpeg
• Kokoro TTS — High-quality voices: see https://github.com/hexgrad/kokoro

Tips

• For long videos (>30 min), increase sub-agent timeout to 900s
• Speaker detection works best with clear interview/podcast formats
• Single-speaker videos (tutorials, lectures) skip speaker labels automatically
• Timestamps link directly to YouTube at that moment