youtube-anycaption-summarizer

What This Skill Does

The YouTube AnyCaption Summarizer is a robust, self-contained OpenClaw agent skill designed to bridge the gap between inaccessible video content and actionable documentation. Unlike standard transcription tools that rely solely on YouTube's provided metadata or captions, this skill implements a multi-tiered extraction strategy. It first attempts to fetch manual closed captions (CC), proceeds to auto-generated subtitles, and finally falls back to local Whisper transcription if no usable text is available. The skill outputs clean, formatted markdown transcripts alongside polished, session-ready summaries, ensuring consistent documentation across diverse video sources.

Installation

Setting up the skill requires a few system dependencies to ensure the local audio processing engine is functional. Open your terminal and execute the following commands to install the necessary binaries and download the Whisper model:

brew install yt-dlp ffmpeg whisper-cpp
MODELS_DIR="$HOME/.openclaw/workspace"
MODEL_PATH="$MODELS_DIR/ggml-medium.bin"
mkdir -p "$MODELS_DIR"
if [ ! -f "$MODEL_PATH" ]; then
  curl -L https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-medium.bin \
    -o "$MODEL_PATH.part" && mv "$MODEL_PATH.part" "$MODEL_PATH"
fi

After installation, you can install the skill via the OpenClaw command interface: clawhub install openclaw/skills/skills/arthurli202602-commits/youtube-anycaption-summarizer. This process is non-destructive and verifies your system environment before finalizing.

Use Cases

This skill is highly effective for technical research and professional content management. Users can process founder videos and operator walkthroughs into searchable knowledge bases. It excels in batch-processing long-form technical explainers that require precise transcriptions and high-level summaries. Furthermore, it supports authenticated access for private or internal-only YouTube videos via cookie configuration, making it indispensable for private tutorial libraries and corporate training materials that are not accessible to public-facing scrapers.

Example Prompts

1. "Summarize this video: https://youtube.com/watch?v=EXAMPLE_ID and provide a detailed markdown transcript for my technical documentation."
2. "Process this playlist of three tutorial videos; generate a single summary document and separate transcripts for each."
3. "Summarize this private company update video using my browser cookies for authentication: [URL]"

Tips & Limitations

To ensure optimal performance, ensure that your yt-dlp is updated regularly, as YouTube frequently changes its playback architecture. While the local Whisper fallback is highly reliable, it is resource-intensive; for long videos, ensure you have sufficient disk space and memory. If you require multi-language support, verify that your selected Whisper model variant supports the target source language. Always use the --models-dir flag if you prefer to keep your library outside of the default OpenClaw workspace.