youtube-digest

What This Skill Does

The youtube-digest skill is a powerful tool designed to turn long-form YouTube video content into digestible, actionable knowledge. By utilizing a transcript-first workflow, this skill extracts metadata, subtitles, and audio transcripts to provide summaries, translations, and timestamped key insights. It effectively bridges the gap between passive video viewing and active information retrieval, allowing users to extract value without needing to watch entire videos.

Installation

To install this skill, run the following command in your terminal within the OpenClaw environment: clawhub install openclaw/skills/skills/benheee/youtube-digest Ensure you have yt-dlp, ffmpeg, and deno installed on your system, as these are prerequisites for the tool's core functionality.

Use Cases

• Research Efficiency: Quickly parse long educational or tutorial videos to find specific methods or technical solutions.
• Language Learning: Use the translation capabilities to understand videos in different languages with accurate transcript mapping.
• Content Creation: Extract notable quotes and structured insights to repurpose video content into blog posts, social media updates, or newsletters.
• Meeting Summarization: Quickly review recorded meetings or lecture sessions shared via YouTube to identify key action items and open questions.

Example Prompts

1. "Summarize this video [URL] into a 3-line executive summary and provide a bulleted list of actionable steps."
2. "Watch the video [URL] and tell me what the speaker says about AI model architecture at the 5-minute mark, and translate that section into Chinese."
3. "Create a list of key timestamps and a summary of the main arguments presented in [URL]. Please note if the summary relies on auto-generated subtitles."

Tips & Limitations

For optimal results, prioritize URLs that include manually uploaded subtitles rather than auto-generated ones. If you encounter errors, verify your network connectivity or use the --proxy flag. Note that the skill currently relies on text-based analysis; for deep visual comprehension, ensure ffmpeg is properly configured to extract key frames. Always check the summary.json file generated in your output directory to understand the quality of the extraction before proceeding with your final analysis.