YouTube Transcript

Download YouTube video transcripts with automatic frame extraction at visual references.

Metadata

Usage

/youtube-transcript https://www.youtube.com/watch?v=VIDEO_ID

What It Does

1. Downloads transcript - Auto-generated or manual captions
2. Detects visual references - Phrases like “as you can see”, “look at this”
3. Extracts frames - Screenshots at key moments
4. Presents combined output - Transcript with embedded images

Example Output

[00:00] Introduction to the topic...
[01:23] Now let me show you the architecture diagram
        [Frame extracted: architecture-01-23.png]
[02:45] As you can see here, the data flows from...
        [Frame extracted: diagram-02-45.png]

Visual Reference Detection

The skill detects phrases in English and German:

English:

• “as you can see”
• “look at this”
• “here’s the diagram”
• “on this slide”
• “let me show you”

German:

• “wie Sie sehen”
• “schauen Sie hier”
• “auf dieser Folie”

Use Cases

• Conference talks - Extract slides and diagrams
• Tutorials - Capture UI screenshots with instructions
• Code walkthroughs - Save code snippets shown on screen
• Presentations - Get slides without screen recording

Requirements

Dependencies are installed automatically when you select this skill:

• yt-dlp - Downloads videos and transcripts
• ffmpeg - Extracts frames from video

Limitations

• Requires captions (auto-generated or manual)
• Frame quality depends on video quality
• Large videos take longer to process

Related

• Record 021 - Implementation details