youtube-digest

To summarize YouTube videos into Markdown notes, provide a video URL to extract subtitles, correct proper nouns, and generate structured outputs. The system produces a concise summary, key insights, and an optional learning quiz formatted in Markdown.

Subtitle extraction for YouTube transcripts works by pulling available subtitles and prioritizing manual captions over auto-generated ones. A VTT parser normalizes the raw transcript text, fixing named entities for consistency before summarization.

Yes, you need yt-dlp installed to extract YouTube video transcripts, as it is the required dependency for retrieving subtitles. The summarization and quiz generation agents then process the retrieved subtitle data.

Yes, you can generate a learning quiz from a YouTube video transcript by using the --quiz flag. This triggers a dedicated agent to create basic to intermediate questions that reinforce learning from the extracted video content.

The best way to capture key insights from long YouTube videos is to extract subtitles and generate a Markdown digest. This approach produces a concise summary and core insights without requiring you to watch the entire video.

Proper nouns are often incorrect in auto-generated YouTube transcripts due to speech recognition limitations. This tool corrects proper nouns and named entities during the VTT parsing stage to ensure consistency in the final Markdown output.