一个AI代理技能，可以从视频平台提取字幕/文字稿，并生成带有关键帧截图的结构化摘要笔记。支持**Cursor**和**Claude代码**。
概览
一个AI代理技能，可以从视频平台提取字幕/文字稿，并生成带有关键帧截图的结构化摘要笔记。可与和 一起使用。 - 支持自动平台检测 - ：平台API → yt-dlp 字幕 → Whisper 语音识别 - ：Bilibili、抖音和小红书均无需登录或 Cookie 即可使用 - ： 自动提取视频帧并嵌入摘要中 - ： 提取结果缓存本地以避免重复下载 - ： 结构化标记笔记，包含章节、高亮和标签 1. 将技能放入正确的目录（参见安装） 2.对于你需要的平台（参见依赖） 3.进入光标或Claude——AI会自动检测并总结 这个技能是一个 pip 包。它是 Cursor/Claude 通过 SKILL.md 约定发现的文件目录。你通过克隆或下载该文件夹到特定位置来安装它。 或者下载ZIP并解压到~/.cursor/skills/multi-video-summarizer/。
概述（AI 翻译）
中文文档
多视频摘要器
一个AI代理技能，可以从视频平台提取字幕/文字稿，并生成带有关键帧截图的结构化摘要笔记。支持Cursor和Claude代码。
支持平台
| 平台 | URL 示例 | 提取方法 | 额外依赖 |
|---|---|---|---|
| Bilibili（B站） | bilibili.com/video/BVxxx |
公共API（WBI签名） | 无（仅标准库） |
| YouTube | youtube.com/watch?v=xxx， youtu.be/xxx |
youtube-transcript-api |
pip install youtube-transcript-api |
| 抖音 | douyin.com/， v.douyin.com/xxx |
直接API + yt-dlp | pip install yt-dlp |
| 小红书 | xiaohongshu.com/， xhslink.com/ |
直接页面解析（无cookie） | 无提取;低语转录 |
| 抖音 | tiktok.com/@user/video/xxx |
yt-dlp | pip install yt-dlp |
| 1800+ 其他站点任何由 yt-dlp 支持的 URL | yt-dlp | pip install yt-dlp |
特色
- 多平台字幕提取，并具备自动平台检测功能
- 三层备份：平台API → yt-dlp 字幕→ Whisper 语音识别
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
Multi-Video-Summarizer
An AI agent skill that extracts subtitles/transcripts from video platforms and generates structured summary notes with keyframe screenshots. Works with Cursor and Claude Code.
Supported Platforms
| Platform | URL Examples | Extraction Method | Extra Dependencies |
|---|---|---|---|
| Bilibili (B站) | bilibili.com/video/BVxxx |
Public API (WBI signing) | None (stdlib only) |
| YouTube | youtube.com/watch?v=xxx, youtu.be/xxx |
youtube-transcript-api |
pip install youtube-transcript-api |
| Douyin (抖音) | douyin.com/, v.douyin.com/xxx |
Direct API + yt-dlp | pip install yt-dlp |
| Xiaohongshu (小红书) | xiaohongshu.com/, xhslink.com/ |
Direct page parsing (no cookie) | None for extraction; Whisper for transcription |
| TikTok | tiktok.com/@user/video/xxx |
yt-dlp | pip install yt-dlp |
| 1800+ other sites | Any URL supported by yt-dlp | yt-dlp | pip install yt-dlp |
Features
- Multi-platform subtitle extraction with automatic platform detection
- 3-layer fallback: platform API → yt-dlp subtitles → Whisper speech recognition
- Cookie-free for major platforms: Bilibili, Douyin, and Xiaohongshu all work without login or cookies
- Keyframe screenshots: automatically extracts video frames and embeds them in summaries
- Caching: extracted results are cached locally to avoid redundant downloads
- BibiGPT-style output: structured markdown notes with sections, highlights, and tags
Quick Start
- Download the skill into the correct directory (see Installation)
- Install dependencies for the platforms you need (see Dependencies)
- Paste a video URL into Cursor or Claude — the AI will automatically detect and summarize it
Installation
This skill is not a pip package. It is a directory of files that Cursor/Claude discovers via the SKILL.md convention. You install it by cloning or downloading the folder to a specific location.
For Cursor IDE
# Clone into Cursor's skills directory
git clone https://github.com/keepongo/video-summarizer.git \
~/.cursor/skills/multi-video-summarizer
Or download the ZIP and extract to ~/.cursor/skills/multi-video-summarizer/.
Windows path: %USERPROFILE%\.cursor\skills\multi-video-summarizer\
For Claude Code / Claude Desktop
git clone https://github.com/keepongo/video-summarizer.git \
~/.claude/skills/multi-video-summarizer
Windows path: %USERPROFILE%\.claude\skills\multi-video-summarizer\
How the AI Discovers the Skill
Cursor and Claude automatically scan their skills directories for folders containing a SKILL.md file. When you paste a video URL (e.g., bilibili.com, youtube.com, douyin.com), the AI reads SKILL.md, which tells it how to run video_subtitle.py and format the output. No manual activation is needed.
Dependencies
推荐工具
换一个关键词，或者移除筛选条件。
安装
npx skillfish add keepongo/video-summarizer