The problem

Manually summarizing long YouTube videos or podcasts is time-consuming and prone to detail loss. Raw transcripts are often cluttered with auto-caption errors, lack structure, and miss clickable navigation.

What it does

• Extracts transcripts and metadata using yt-dlp and jq.
• Chunks long videos to ensure no section is skipped during analysis.
• Generates structured Markdown notes with clickable, precise timestamps.
• Identifies and flags phonetic auto-caption errors and speaker claims.
• Extracts mentioned tools, people, and specific data points with proper attribution.

Frameworks & tools

Requires a Bash environment with yt-dlp, jq, and awk installed.

Why this beats prompting it yourself

Standard LLM prompts often hallucinate details or hit context limits on long videos, leading to skipped sections. This skill uses a multi-step chunking strategy and cross-references metadata to ensure every minute of the video is accounted for without losing technical accuracy.

Use cases

• Convert technical conference talks into searchable documentation.
• Summarize long-form podcasts into actionable key takeaways and tool lists.
• Create timestamped study guides from educational lectures.
• Audit video claims by extracting specific data points and their original context.

Known limitations

Only analyzes spoken audio and captions; cannot "see" on-screen visuals, slides, or demos. Requires videos to have available transcripts (manual or automatic).