YouTube Clipper Skill

AI-powered YouTube video clipper for Claude Code. Download videos, generate semantic chapters, clip segments, translate subtitles to bilingual format, and burn subtitles into videos.

English | 简体中文

Features • Installation • Usage • Requirements • Configuration • Troubleshooting

---

Features

• AI Semantic Analysis - Generate fine-grained chapters (2-5 minutes each) by understanding video content, not just mechanical time splitting
• Precise Clipping - Use FFmpeg to extract video segments with frame-accurate timing
• Bilingual Subtitles - Batch translate subtitles to Chinese/English with 95% API call reduction
• Subtitle Burning - Hardcode bilingual subtitles into videos with customizable styling
• Content Summarization - Auto-generate social media content (Xiaohongshu, Douyin, WeChat)

---

Installation

Option 1: npx skills (Recommended)

npx skills add https://github.com/op7418/Youtube-clipper-skill

This command will automatically install the skill to ~/.claude/skills/youtube-clipper/.

Option 2: Manual Installation

git clone https://github.com/op7418/Youtube-clipper-skill.git
cd Youtube-clipper-skill
bash install_as_skill.sh

The install script will:

• Copy files to ~/.claude/skills/youtube-clipper/
• Install Python dependencies (yt-dlp, pysrt, python-dotenv)
• Check system dependencies (Python, yt-dlp, FFmpeg)
• Create .env configuration file

---

Requirements

System Dependencies

Python Packages

These are automatically installed by the install script:

• yt-dlp - YouTube downloader
• pysrt - SRT subtitle parser
• python-dotenv - Environment variable management

Important: FFmpeg libass Support

macOS users: The standard ffmpeg package from Homebrew does NOT include libass support (required for subtitle burning). You must install ffmpeg-full:

# Remove standard ffmpeg (if installed)
brew uninstall ffmpeg

# Install ffmpeg-full (includes libass)
brew install ffmpeg-full

Verify libass support:

ffmpeg -filters 2>&1 | grep subtitles
# Should output: subtitles    V->V  (...)

---

Usage

In Claude Code

Simply tell Claude to clip a YouTube video:

Clip this YouTube video: https://youtube.com/watch?v=VIDEO_ID

or

剪辑这个 YouTube 视频：https://youtube.com/watch?v=VIDEO_ID

Workflow

1. Environment Check - Verifies yt-dlp, FFmpeg, and Python dependencies
2. Video Download - Downloads video (up to 1080p) and English subtitles
3. AI Chapter Analysis - Claude analyzes subtitles to generate semantic chapters (2-5 min each)
4. User Selection - Choose which chapters to clip and processing options
5. Processing - Clips video, translates subtitles, burns subtitles (if requested)
6. Output - Organized files in ./youtube-clips/<timestamp>/

Output Files

For each clipped chapter:

./youtube-clips/20260122_143022/
└── Chapter_Title/
    ├── Chapter_Title_clip.mp4              # Original clip (no subtitles)
    ├── Chapter_Title_with_subtitles.mp4    # With burned subtitles
    ├── Chapter_Title_bilingual.srt         # Bilingual subtitle file
    └── Chapter_Title_summary.md            # Social media content

---

Configuration

The skill uses environment variables for customization. Edit ~/.claude/skills/youtube-clipper/.env:

Key Settings

# FFmpeg path (auto-detected if empty)
FFMPEG_PATH=

# Output directory (default: current working directory)
OUTPUT_DIR=./youtube-clips

# Video quality limit (720, 1080, 1440, 2160)
MAX_VIDEO_HEIGHT=1080

# Translation batch size (20-25 recommended)
TRANSLATION_BATCH_SIZE=20

# Target language for translation
TARGET_LANGUAGE=中文

# Target chapter duration in seconds (180-300 recommended)
TARGET_CHAPTER_DURATION=180

For full configuration options, see .env.example.

---

Examples

Example 1: Extract highlights from a tech interview

Input:

Clip this video: https://youtube.com/watch?v=Ckt1cj0xjRM

Output (AI-generated chapters):

1. [00:00 - 03:15] AGI as an exponential curve, not a point in time
2. [03:15 - 06:30] China's gap in AI development
3

...