YouTube AI Digest
https://github.com/yizhiyanhua-ai/youtube-ai-digest- Category
- Developer Tools
- Rank
- No. 1999Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- yizhiyanhua-ai
- GitHub
- 56 stars
- Date
About
Claude Code skill that browses, summarizes, and captures AI-related YouTube videos.
What it does
You choose YouTube channels, then ask Claude Code to find recent AI videos, retrieve available subtitles, and produce structured Markdown notes. The supporting scripts filter candidates by title and description keywords, preserve timestamps, fetch thumbnails, and attach video metadata.
Why it's ranked here
This is a useful, focused workflow for turning a hand-picked channel list into readable AI briefings. Its strongest distinction is transparency: discovery, transcript retrieval, summarization, and report generation remain separate steps. However, Claude supplies the actual summaries, so the Python tooling alone is not a complete digest engine.
What's good
Discovery supports configurable time windows, custom keyword filters, and an unfiltered mode. Results sort by upload date and view count. Transcript retrieval tries public access first, then Chrome and Firefox cookies, and exports timestamped text plus JSON. Reports combine metadata, a thumbnail, a supplied summary, and a transcript excerpt.
Tradeoffs
The AI filter is simple substring matching across titles and short descriptions, so relevance can be noisy. Channel fetching examines only ten playlist entries and relies on approximate upload dates. Transcript parsing drops repeated text globally and ignores lines containing markup. Browser-cookie retries assume Chrome or Firefox, while reports include only the first 3,000 transcript characters.
How to use it well
It suits Claude Code users who follow a stable set of AI channels and want quick daily scanning before choosing videos to watch. Tune the channel list and keywords, inspect the candidates, then summarize selected transcripts or create Markdown reports. It does not replace broad web research, semantic video search, or an independent summarization service.
Technical notes+
README.md and README.en.md document Claude Code plugin installation, manual cloning, and a Python 3.9+ runtime with yt-dlp. scripts/fetch_videos.py shells out to yt-dlp in flat-playlist mode, caps each channel at ten entries, applies approximate-date and keyword filtering, then serializes results. scripts/get_transcript.py requests English and Chinese VTT subtitles, retries with Chrome and Firefox cookies, parses timestamped cues, and writes text plus JSON. scripts/generate_report.py fetches metadata and a thumbnail through yt-dlp, then writes a per-video Markdown report with at most 3,000 transcript characters. SKILL.md assigns Chinese summarization to Claude rather than implementing model inference in Python. metadata.json lists macOS and Linux, while data/channels.json supplies fifteen default channels.
Observed
- License
- MIT License
- Primary language
- Python
- Runtime
- Python 3.9 or newer
- Install surface
- Claude Code plugin, global skill installation, or manual repository clone
- Interface
- Claude Code skill with command-line Python scripts
- External tooling
- yt-dlp is used for video metadata, subtitles, and thumbnails
- Platform support
- macOS and Linux
- Configuration
- JSON channel list with fifteen supplied YouTube channels
Read from README.md, scripts/fetch_videos.py, scripts/get_transcript.py, scripts/generate_report.py, SKILL.md, README.en.md, metadata.json, data/channels.json, .claude-plugin/marketplace.json.
What it can do
Browse YouTube for AI-related videos
Search terms or keywords related to AI → List of relevant AI YouTube videos
Summarize AI YouTube video content
YouTube video URL or video identifier → Text summary of video content
Extract key information from AI videos
AI-related YouTube video → Structured data with main points and insights
Capture and save video metadata
YouTube video URL → Video title, description, duration, and other metadata
Generate digestible reports from AI video content
One or multiple AI YouTube videos → Formatted digest report with key takeaways
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.