video-summarizer

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Video Summarizer

视频总结工具

Overview

概述

Download videos from any platform and generate a complete resource package including:

Original video file (mp4)
Audio file (mp3)
Subtitle file (with timestamps, vtt/srt format)
Summary file (summary.md)

Supports all 1800+ websites supported by yt-dlp.

支持从任意平台下载视频，并生成包含以下内容的完整资源包：

原始视频文件（mp4格式）
音频文件（mp3格式）
字幕文件（带时间戳，vtt/srt格式）
总结文件（summary.md）

支持yt-dlp兼容的所有1800+网站。

Trigger Conditions

触发条件

When the user:

Provides a video link and asks for a summary
Says "summarize this video", "what's in this video"
Asks to "extract video content", "transcribe video"
Says "download this video"
Provides a link from YouTube/Bilibili/Twitter/Vimeo/TikTok etc.

当用户：

提供视频链接并要求生成总结
说出“总结这个视频”、“这个视频讲了什么”
要求“提取视频内容”、“转录视频”
说出“下载这个视频”
提供来自YouTube/Bilibili/Twitter/Vimeo/TikTok等平台的链接

Supported Platforms

支持的平台

YouTube (youtube.com, youtu.be)
Bilibili (bilibili.com, b23.tv)
Twitter/X (x.com, twitter.com)
Vimeo (vimeo.com)
TikTok (tiktok.com)
Instagram (instagram.com)
Twitch (twitch.tv)
And 1800+ other platforms (all sites supported by yt-dlp)

YouTube (youtube.com, youtu.be)
Bilibili (bilibili.com, b23.tv)
Twitter/X (x.com, twitter.com)
Vimeo (vimeo.com)
TikTok (tiktok.com)
Instagram (instagram.com)
Twitch (twitch.tv)
以及其他1800+平台（所有yt-dlp支持的网站）

Output Structure

输出结构

All files are saved to

downloads/<video-title>/

in the current working directory:

./downloads/
└── <video-title>/
    ├── video.mp4          # Original video
    ├── audio.mp3          # Extracted audio
    ├── subtitle.vtt       # Subtitles with timestamps
    ├── transcript.txt     # Plain text transcript (no timestamps)
    └── summary.md         # Structured summary

所有文件将保存到当前工作目录的

downloads/<video-title>/

路径下：

./downloads/
└── <video-title>/
    ├── video.mp4          # 原始视频
    ├── audio.mp3          # 提取的音频
    ├── subtitle.vtt       # 带时间戳的字幕
    ├── transcript.txt     # 纯文本转录内容（无时间戳）
    └── summary.md         # 结构化总结

Workflow

工作流程

Step 1: Install Dependencies

步骤1：安装依赖

Run the install script to check and install all dependencies:

bash

bash "$SKILL_DIR/scripts/install_deps.sh"

This installs: uv (Python package manager), ffmpeg, yt-dlp, and checks Python version. faster-whisper will be automatically managed by uv.

运行安装脚本检查并安装所有依赖项：

bash

bash "$SKILL_DIR/scripts/install_deps.sh"

该脚本会安装：uv（Python包管理器）、ffmpeg、yt-dlp，并检查Python版本。faster-whisper将由uv自动管理。

Step 2: Get Video Info and Create Output Directory

步骤2：获取视频信息并创建输出目录

bash

undefined

bash

undefined

Get video title (sanitize special characters for folder name)

获取视频标题（清理特殊字符以用作文件夹名称）

TITLE=$(yt-dlp --print "%(title)s" "VIDEO_URL" | sed 's/[/:*?"<>|]/_/g' | cut -c1-80) DURATION=$(yt-dlp --print "%(duration)s" "VIDEO_URL")

Create output directory

创建输出目录

OUTPUT_DIR=./downloads/"$TITLE" mkdir -p "$OUTPUT_DIR"

undefined

OUTPUT_DIR=./downloads/"$TITLE" mkdir -p "$OUTPUT_DIR"

undefined

Step 3: Download Video and Audio

步骤3：下载视频和音频

bash

undefined

bash

undefined

Download video (mp4 format, best quality up to 1080p)

下载视频（mp4格式，最高1080p最佳画质）

yt-dlp -f "bestvideo[height<=1080][ext=mp4]+bestaudio[ext=m4a]/best[height<=1080][ext=mp4]/best"
--merge-output-format mp4
-o "$OUTPUT_DIR/video.%(ext)s" "VIDEO_URL"

Extract audio (mp3 format)

提取音频（mp3格式）

yt-dlp -x --audio-format mp3 -o "$OUTPUT_DIR/audio.%(ext)s" "VIDEO_URL"

undefined

yt-dlp -x --audio-format mp3 -o "$OUTPUT_DIR/audio.%(ext)s" "VIDEO_URL"

undefined

Step 4: Get Subtitles

步骤4：获取字幕

Priority order:

Try downloading manual subtitles (best quality)

bash

yt-dlp --write-subs --sub-lang zh,en,zh-Hans,zh-Hant --skip-download \
  -o "$OUTPUT_DIR/subtitle" "VIDEO_URL"

Try downloading auto-generated subtitles

bash

yt-dlp --write-auto-subs --sub-lang zh,en --skip-download \
  -o "$OUTPUT_DIR/subtitle" "VIDEO_URL"

Use faster-whisper transcription when no subtitles available

bash

uv run "$SKILL_DIR/scripts/parallel_transcribe.py" \
  --input "$OUTPUT_DIR/audio.mp3" \
  --output-dir "$OUTPUT_DIR" \
  --model small \
  --language auto

The script automatically:

Splits long audio files at silence points
Uses multiple CPU cores for parallel transcription
Outputs both
```
subtitle.vtt
```
and
```
transcript.txt
```

Transcription Options:

Option	Default	Description
`--model`	small	tiny/base/small/medium/large-v3
`--language`	auto	Language code or 'auto'
`--workers`	CPU/2	Number of parallel workers
`--min-segment`	60	Min duration (sec) to enable splitting

优先级顺序：

尝试下载手动添加的字幕（画质最佳）

bash

yt-dlp --write-subs --sub-lang zh,en,zh-Hans,zh-Hant --skip-download \
  -o "$OUTPUT_DIR/subtitle" "VIDEO_URL"

尝试下载自动生成的字幕

bash

yt-dlp --write-auto-subs --sub-lang zh,en --skip-download \
  -o "$OUTPUT_DIR/subtitle" "VIDEO_URL"

当无可用字幕时使用faster-whisper进行转录

bash

uv run "$SKILL_DIR/scripts/parallel_transcribe.py" \
  --input "$OUTPUT_DIR/audio.mp3" \
  --output-dir "$OUTPUT_DIR" \
  --model small \
  --language auto

该脚本会自动：

在静音点分割长音频文件
使用多个CPU核心进行并行转录
同时输出
```
subtitle.vtt
```
和
```
transcript.txt
```

转录选项:

选项	默认值	描述
`--model`	small	tiny/base/small/medium/large-v3
`--language`	auto	语言代码或'auto'
`--workers`	CPU/2	并行工作线程数
`--min-segment`	60	启用分割的最小音频时长（秒）

Step 5: Generate Plain Text Transcript

步骤5：生成纯文本转录内容

If subtitles were downloaded (not transcribed), convert to plain text:

bash

if [[ ! -f "$OUTPUT_DIR/transcript.txt" ]]; then
  SUBTITLE_FILE=$(ls "$OUTPUT_DIR"/*.vtt "$OUTPUT_DIR"/*.srt 2>/dev/null | head -1)
  if [[ "$SUBTITLE_FILE" == *.vtt ]]; then
    sed '/^[0-9]/d; /^$/d; /-->/d; /^WEBVTT/d; /^Kind:/d; /^Language:/d; /^NOTE/d' \
      "$SUBTITLE_FILE" > "$OUTPUT_DIR/transcript.txt"
  elif [[ "$SUBTITLE_FILE" == *.srt ]]; then
    sed '/^[0-9]/d; /^$/d; /-->/d' "$SUBTITLE_FILE" > "$OUTPUT_DIR/transcript.txt"
  fi
fi

如果是下载的字幕（而非转录生成），将其转换为纯文本：

bash

if [[ ! -f "$OUTPUT_DIR/transcript.txt" ]]; then
  SUBTITLE_FILE=$(ls "$OUTPUT_DIR"/*.vtt "$OUTPUT_DIR"/*.srt 2>/dev/null | head -1)
  if [[ "$SUBTITLE_FILE" == *.vtt ]]; then
    sed '/^[0-9]/d; /^$/d; /-->/d; /^WEBVTT/d; /^Kind:/d; /^Language:/d; /^NOTE/d' \
      "$SUBTITLE_FILE" > "$OUTPUT_DIR/transcript.txt"
  elif [[ "$SUBTITLE_FILE" == *.srt ]]; then
    sed '/^[0-9]/d; /^$/d; /-->/d' "$SUBTITLE_FILE" > "$OUTPUT_DIR/transcript.txt"
  fi
fi

Step 6: Generate Summary File

步骤6：生成总结文件

Read prompt template from
```
$SKILL_DIR/reference/summary-prompt.md
```

Replace placeholders:

{{TITLE}}

{{PLATFORM}}

{{URL}}

{{DURATION}}

{{LANGUAGE}}

{{DOWNLOAD_TIME}}

{{TRANSCRIPT}}

Generate summary and save to
```
$OUTPUT_DIR/summary.md
```

从
```
$SKILL_DIR/reference/summary-prompt.md
```
读取提示模板

替换占位符：

{{TITLE}}

、

{{PLATFORM}}

、

{{URL}}

、

{{DURATION}}

、

{{LANGUAGE}}

、

{{DOWNLOAD_TIME}}

、

{{TRANSCRIPT}}

生成总结并保存到
```
$OUTPUT_DIR/summary.md
```

Platform-Specific Handling

平台专属处理

Bilibili

bash

undefined

bash

undefined

Prioritize Chinese subtitles

优先选择中文字幕

yt-dlp --sub-lang zh-Hans,zh-Hant,zh ...

If login required

如果需要登录

yt-dlp --cookies-from-browser chrome "VIDEO_URL"

undefined

yt-dlp --cookies-from-browser chrome "VIDEO_URL"

undefined

Platforms Requiring Login

需要登录的平台

bash

yt-dlp --cookies-from-browser chrome "VIDEO_URL"

bash

yt-dlp --cookies-from-browser chrome "VIDEO_URL"

or firefox

或使用firefox

yt-dlp --cookies-from-browser firefox "VIDEO_URL"

undefined

yt-dlp --cookies-from-browser firefox "VIDEO_URL"

undefined

Error Handling

错误处理

Cannot Get Subtitles

无法获取字幕

Use the parallel transcription script (Step 4, option 3).

使用并行转录脚本（步骤4，选项3）。

Video Too Long (>1 hour)

视频过长（超过1小时）

Ask user if they only need partial content
The parallel script handles long files automatically

询问用户是否只需要部分内容
并行脚本会自动处理长文件

Unsupported Platform

不支持的平台

bash

yt-dlp --list-extractors | grep -i "platform-name"

bash

yt-dlp --list-extractors | grep -i "platform-name"

Notes

注意事项

Storage: Files saved to
```
./downloads/
```
in current working directory
Copyright: For personal learning use only
Network: Some platforms may require proxy
First Run: Whisper model download required (~244MB for small)
Parallel Processing: Long audio (>60s) auto-splits at silence points

存储：文件保存到当前工作目录的
```
./downloads/
```
路径下
版权：仅用于个人学习使用
网络：部分平台可能需要使用代理
首次运行：需要下载Whisper模型（small版本约244MB）
并行处理：长音频（超过60秒）会自动在静音点分割