yt-summarize skill

---
name: youtube-summarize
description: Summarize YouTube videos by extracting transcripts and captions. Use when you need to get a quick summary of a video, extract key points, or analyze video content without watching it.
metadata: {"openclaw":{"requires":{"bins":["yt-dlp"]},"install":[{"id":"python","kind":"pip","package":"yt-dlp","bins":["yt-dlp"],"label":"Install yt-dlp (pip)"}]}}
---

# YouTube Video Summarizer

## Setup

Install yt-dlp:
```bash
pip install yt-dlp
```

## Extract Transcript

Get auto-generated subtitles:
```bash
yt-dlp --write-auto-sub --sub-lang en --skip-download --sub-format vtt -o "%(title)s" "VIDEO_URL"
```

Get manual subtitles (if available):
```bash
yt-dlp --write-sub --sub-lang en --skip-download --sub-format vtt -o "%(title)s" "VIDEO_URL"
```

List available subtitles:
```bash
yt-dlp --list-subs "VIDEO_URL"
```

## Extract as Plain Text

Download and convert to text:
```bash
yt-dlp --write-auto-sub --sub-lang en --skip-download --sub-format vtt -o "transcript" "VIDEO_URL" && \
sed -e '/^$/d' -e '/^[0-9]/d' -e '/-->/d' -e 's/<[^>]*>//g' transcript.en.vtt | sort -u > transcript.txt
```

## Quick Transcript to Stdout

```bash
yt-dlp --write-auto-sub --sub-lang en --skip-download --sub-format json3 -o - "VIDEO_URL" 2>/dev/null | \
python3 -c "
import sys, json
data = json.load(sys.stdin)
for event in data.get('events', []):
    for seg in event.get('segs', []):
        if text := seg.get('utf8', '').strip():
            print(text, end=' ')"
```

## Get Video Metadata

```bash
yt-dlp --dump-json "VIDEO_URL" | python3 -c "
import sys, json
d = json.load(sys.stdin)
print(f\"Title: {d['title']}\")
print(f\"Channel: {d['channel']}\")
print(f\"Duration: {d['duration']//60}:{d['duration']%60:02d}\")
print(f\"Views: {d.get('view_count', 'N/A'):,}\")
print(f\"Upload: {d.get('upload_date', 'N/A')}\")
print(f\"Description:\n{d.get('description', '')[:500]}...\")"
```

## Summarization Workflow

1. Extract transcript:
```bash
yt-dlp --write-auto-sub --sub-lang en --skip-download -o "video" "VIDEO_URL"
```

2. Clean VTT to plain text:
```bash
python3 -c "
import re
with open('video.en.vtt', 'r') as f:
    content = f.read()
# Remove VTT headers and timestamps
content = re.sub(r'WEBVTT.*?\n\n', '', content, flags=re.DOTALL)
content = re.sub(r'\d+:\d+:\d+\.\d+ --> \d+:\d+:\d+\.\d+.*?\n', '', content)
content = re.sub(r'<[^>]+>', '', content)
lines = [l.strip() for l in content.split('\n') if l.strip()]
unique = []
for l in lines:
    if l not in unique[-1:]:
        unique.append(l)
print(' '.join(unique))" > transcript.txt
```

3. Send to LLM for summarization (the transcript is now ready for Claude to analyze)

## Multi-language Support

Extract subtitles in other languages:
```bash
# Russian
yt-dlp --write-auto-sub --sub-lang ru --skip-download "VIDEO_URL"

# Spanish
yt-dlp --write-auto-sub --sub-lang es --skip-download "VIDEO_URL"

# Multiple languages
yt-dlp --write-auto-sub --sub-lang "en,ru,es" --skip-download "VIDEO_URL"
```

## Chapter Extraction

Get video chapters (if available):
```bash
yt-dlp --dump-json "VIDEO_URL" | python3 -c "
import sys, json
d = json.load(sys.stdin)
for ch in d.get('chapters', []):
    start = int(ch['start_time'])
    print(f\"{start//60}:{start%60:02d} - {ch['title']}\")"
```

## Common Options

| Option | Description |
|--------|-------------|
| `--sub-lang en` | Subtitle language (en, ru, es, de, fr, etc.) |
| `--write-auto-sub` | Get auto-generated captions |
| `--write-sub` | Get manual subtitles |
| `--sub-format vtt` | Output format (vtt, srt, json3) |
| `--skip-download` | Don't download video |

## Notes

- Auto-generated subtitles may have errors
- Not all videos have subtitles available
- Some videos have subtitles disabled by uploader
- Use `--sub-lang` with appropriate language code
- Transcripts work best for spoken content (lectures, podcasts, tutorials)