Skip to content
Back to search

video-quotes

MarketplaceWorks with Claude CodeGitHub

Claude Code skill: YouTube interview to corrected transcript, quote page and word-accurate clips.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/aokellermann--video-quotes/badge.svg)](/listings/aokellermann--video-quotes)

HTML

<a href="/listings/aokellermann--video-quotes"><img src="/api/listings/aokellermann--video-quotes/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

As a plugin:

/plugin marketplace add aokellermann/video-quotes
/plugin install video-quotes@video-quotes

Or as a personal skill, by linking the skill folder into your skills directory:

git clone https://github.com/aokellermann/video-quotes ~/repos/video-quotes
ln -s ~/repos/video-quotes/skills/video-quotes ~/.claude/skills/video-quotes

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 14 lines
  1. {
  2. "name": "video-quotes",
  3. "owner": {
  4. "name": "Antony Kellermann"
  5. },
  6. "description": "Claude Code skill: YouTube interview to corrected transcript, quote page and word-accurate clips.",
  7. "plugins": [
  8. {
  9. "name": "video-quotes",
  10. "source": "./",
  11. "description": "YouTube interview -> corrected transcript, quote page with timestamps, and word-accurate clips."
  12. }
  13. ]
  14. }
README.md · README · 116 lines
  1. # video-quotes
  2. A [Claude Code](https://claude.com/claude-code) skill that turns a YouTube interview or talk into:
  3. - a **corrected transcript** with speaker labels (YouTube's auto-captions fixed for misheard names and terms,
  4. filler words and stutters removed),
  5. - a **quote page**: the best quotes on a theme you choose, grouped, each linked to its moment in the video, plus a
  6. summary of what else the conversation covers,
  7. - optional **video clips** of each quote, cut at word boundaries and checked so no word is clipped.
  8. The page is a single HTML file (published as a claude.ai Artifact when Claude has that tool), light and dark themes
  9. included.
  10. ## Use
  11. Ask Claude Code things like:
  12. > get me the transcript for https://www.youtube.com/watch?v=…
  13. >
  14. > pull the best quotes about AI risk from it, with clips
  15. >
  16. > do the same for these two videos
  17. The skill's instructions (`skills/video-quotes/SKILL.md`) walk Claude through fetching captions, correcting them,
  18. labelling speakers, choosing quotes, cutting clips and checking every clip's first and last words.
  19. ## Install
  20. As a plugin:
  21. ```
  22. /plugin marketplace add aokellermann/video-quotes
  23. /plugin install video-quotes@video-quotes
  24. ```
  25. Or as a personal skill, by linking the skill folder into your skills directory:
  26. ```bash
  27. git clone https://github.com/aokellermann/video-quotes ~/repos/video-quotes
  28. ln -s ~/repos/video-quotes/skills/video-quotes ~/.claude/skills/video-quotes
  29. ```
  30. ## Requirements
  31. You need two programs: **uv** (always) and **ffmpeg** (for clips; it includes `ffprobe`). Everything else installs
  32. itself on first use.
  33. ### 1. uv
  34. [uv](https://docs.astral.sh/uv/getting-started/installation/) runs `vq.py` and installs its Python dependencies
  35. (youtube-transcript-api, faster-whisper, numpy) and a suitable Python automatically. You don't need Python installed.
  36. | System | Command |
  37. |---|---|
  38. | macOS / Linux | `curl -LsSf https://astral.sh/uv/install.sh \| sh` |
  39. | macOS (Homebrew) | `brew install uv` |
  40. | Windows (PowerShell) | `powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 \| iex"` |
  41. | Windows (winget) | `winget install --id=astral-sh.uv -e` |
  42. Open a new terminal afterwards so `uv` is on your `PATH`.
  43. ### 2. ffmpeg
  44. Only needed for video clips (transcripts and quote pages work without it).
  45. | System | Command |
  46. |---|---|
  47. | macOS (Homebrew) | `brew install ffmpeg` |
  48. | Ubuntu / Debian | `sudo apt install ffmpeg` |
  49. | Fedora | `sudo dnf install ffmpeg` (enable [RPM Fusion](https://rpmfusion.org/Configuration) first) |
  50. | Arch | `sudo pacman -S ffmpeg` |
  51. | Windows (winget) | `winget install --id=Gyan.FFmpeg -e` |
  52. ### 3. Check
  53. ```bash
  54. uv --version && ffmpeg -version | head -1 && ffprobe -version | head -1
  55. ```
  56. Claude runs the same check before starting and tells you what's missing.
  57. ### Installed automatically on first use
  58. - **Python packages** for `vq.py`, by uv, into its cache (a few hundred MB, mostly faster-whisper).
  59. - **yt-dlp**, the latest release, run through `uvx` when clips need the video.
  60. - **The whisper speech model** (`small.en`, ~500 MB) from Hugging Face, the first time clips are cut. It runs on
  61. the CPU, no GPU needed, at roughly 3 minutes of audio per minute.
  62. ### Other things to know
  63. - **Disk**: each downloaded video is ~300 MB at 720p (kept in the work folder you choose; delete it when done).
  64. - **Network**: use a home or office connection. YouTube blocks caption and video requests from many cloud and
  65. datacenter IPs, so this generally won't work on a cloud VM or in a hosted sandbox.
  66. - **Publishing**: in Claude Code with the claude.ai Artifact tool, pages are published as private Artifacts; without
  67. it you get `page.html` with a `clips/` folder to open locally or host anywhere.
  68. ## The script
  69. `skills/video-quotes/vq.py` can also be used on its own:
  70. ```
  71. uv run -q vq.py fetch VIDEO -d DIR captions -> paras.txt (numbered speaker turns), ts.txt, meta.json
  72. uv run -q vq.py fix DIR FIXES.json [--tidy] apply caption corrections; --tidy drops uh/um and stutters
  73. uv run -q vq.py label DIR SPEC.json speaker labels -> tx.html
  74. uv run -q vq.py page DIR PAGE.json -o OUT quote page
  75. uv run -q vq.py video VIDEO -d DIR 720p mp4
  76. uv run -q vq.py words DIR --from S --to E whisper word timestamps for a stretch
  77. uv run -q vq.py clips DIR CLIPS.json cut clips at word boundaries
  78. uv run -q vq.py check DIR transcribe each clip's first/last 3 s
  79. ```
  80. The JSON formats are described in `SKILL.md`.
  81. ## License
  82. MIT, see `LICENSE`.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.