Youtube Transcript

This skill helps download transcripts (subtitles/captions) from YouTube videos using yt-dlp.

When to Use This Skill

Activate this skill when the user:

1. Provides a YouTube URL and wants the transcript
2. Asks to "download transcript from YouTube"
3. Wants to "get captions" or "get subtitles" from a video
4. Asks to "transcribe a YouTube video"
5. Needs text content from a YouTube video

How It Works

Priority Order:

1. Check if yt-dlp is installed - install if needed
2. List available subtitles - see what's actually available
3. Try manual subtitles first (--write-sub) - highest quality
4. Fallback to auto-generated (--write-auto-sub) - usually available
5. Last resort: Whisper transcription - if no subtitles exist (requires user confirmation)
6. Confirm the download and show the user where the file is saved
7. Optionally clean up the VTT format if the user wants plain text

Installation Check

IMPORTANT: Always check if yt-dlp is installed first:

which yt-dlp || command -v yt-dlp

If Not Installed

Attempt automatic installation based on the system:

macOS (Homebrew):

brew install yt-dlp

Linux (apt/Debian/Ubuntu):

sudo apt update && sudo apt install -y yt-dlp

Alternative (pip - works on all systems):

pip3 install yt-dlp
# or
python3 -m pip install yt-dlp

If installation fails: Inform the user they need to install yt-dlp manually and provide them with installation instructions from https://github.com/yt-dlp/yt-dlp#installation

Check Available Subtitles

ALWAYS do this first before attempting to download:

yt-dlp --list-subs "YOUTUBE_URL"

This shows what subtitle types are available without downloading anything. Look for:

1. Manual subtitles (better quality)
2. Auto-generated subtitles (usually available)
3. Available languages

Download Strategy

Option 1: Manual Subtitles (Preferred)

Try this first - highest quality, human-created:

yt-dlp --write-sub --skip-download --output "OUTPUT_NAME" "YOUTUBE_URL"

Option 2: Auto-Generated Subtitles (Fallback)

If manual subtitles aren't available:

yt-dlp --write-auto-sub --skip-download --output "OUTPUT_NAME" "YOUTUBE_URL"

Both commands create a .vtt file (WebVTT subtitle format).

Option 3: Whisper Transcription (Last Resort)

ONLY use this if both manual and auto-generated subtitles are unavailable.

Step 1: Show File Size and Ask for Confirmation

# Get audio file size estimate
yt-dlp --print "%(filesize,filesize_approx)s" -f "bestaudio" "YOUTUBE_URL"

# Or get duration to estimate
yt-dlp --print "%(duration)s %(title)s" "YOUTUBE_URL"

IMPORTANT: Display the file size to the user and ask: "No subtitles are available. I can download the audio (approximately X MB) and transcribe it using Whisper. Would you like to proceed?"

Wait for user confirmation before continuing.

Step 2: Check for Whisper Installation

command -v whisper

If not installed, ask user: "Whisper is not installed. Install it with pip install openai-whisper (requires ~1-3GB for models)? This is a one-time installation."

Wait for user confirmation before installing.

Install if approved:

pip3 install openai-whisper

Step 3: Download Audio Only

yt-dlp -x --audio-format mp3 --output "audio_%(id)s.%(ext)s" "YOUTUBE_URL"

Step 4: Transcribe with Whisper

# Auto-detect language (recommended)
whisper audio_VIDEO_ID.mp3 --model base --output_format vtt

# Or specify language if known
whisper audio_VIDEO_ID.mp3 --model base --language en --output_format vtt

Model Options (stick to base for now):

1. tiny - fastest, least accurate (~1GB)
2. base - good balance (~1GB) ← USE THIS
3. small - better accuracy (~2GB)
4. medium - very good (~5GB)
5. large - best accuracy (~10GB)

Step 5: Cleanup

After transcription completes, ask user: "Transcription complete! Would you like me to delete the audio file to save space?"

If yes:

rm audio_VIDEO_ID.mp3

Getting Video Information

Extract Video Title (for filename)

yt-dlp --print "%(title)s" "YOUTUBE_URL"

Use this to create meaningful filenames based on the video title. Clean the title for filesystem compatibility:

1. Replace / with -
2. Replace special characters that might cause issues
3. Consider using sanitized version: $(yt-dlp --print "%(title)s" "URL" | tr '/' '-' | tr ':' '-')

Post-Processing

Convert to Plain Text (Recommended)

YouTube's auto-generated VTT files contain duplicate lines because captions are shown progressively with overlapping timestamps. Always deduplicate when converting to plain text while preserving the original speaking order.

python3 -c "
import sys, re
seen = set()
with open('transcript.en.vtt', 'r') as f:
    for line in f:
        line = line.strip()
        if line and not line.startswith('WEBVTT') and not line.startswith('Kind:') and not line.startswith('Language:') and '-->' not in line:
            clean = re.sub('<[^>]*>', '', line)
            clean = clean.replace('&amp;', '&').replace('&gt;', '>').replace('&lt;', '<')
            if clean and clean not in seen:
                print(clean)
                seen.add(clean)
" > transcript.txt

Complete Post-Processing with Video Title

# Get video title
VIDEO_TITLE=$(yt-dlp --print "%(title)s" "YOUTUBE_URL" | tr '/' '_' | tr ':' '-' | tr '?' '' | tr '"' '')

# Find the VTT file
VTT_FILE=$(ls *.vtt | head -n 1)

# Convert with deduplication
python3 -c "
import sys, re
seen = set()
with open('$VTT_FILE', 'r') as f:
    for line in f:
        line = line.strip()
        if line and not line.startswith('WEBVTT') and not line.startswith('Kind:') and not line.startswith('Language:') and '-->' not in line:
            clean = re.sub('<[^>]*>', '', line)
            clean = clean.replace('&amp;', '&').replace('&gt;', '>').replace('&lt;', '<')
            if clean and clean not in seen:
                print(clean)
                seen.add(clean)
" > "${VIDEO_TITLE}.txt"

echo "✓ Saved to: ${VIDEO_TITLE}.txt"

# Clean up VTT file
rm "$VTT_FILE"
echo "✓ Cleaned up temporary VTT file"

Output Formats

1. VTT format (.vtt): Includes timestamps and formatting, good for video players
2. Plain text (.txt): Just the text content, good for reading or analysis

Tips

1. The filename will be {output_name}.{language_code}.vtt (e.g., transcript.en.vtt)
2. Most YouTube videos have auto-generated English subtitles
3. Some videos may have multiple language options
4. If auto-subtitles aren't available, try --write-sub instead for manual subtitles

Complete Workflow Example

VIDEO_URL="https://www.youtube.com/watch?v=dQw4w9WgXcQ"

# Get video title for filename
VIDEO_TITLE=$(yt-dlp --print "%(title)s" "$VIDEO_URL" | tr '/' '_' | tr ':' '-' | tr '?' '' | tr '"' '')
OUTPUT_NAME="transcript_temp"

# ============================================
# STEP 1: Check if yt-dlp is installed
# ============================================
if ! command -v yt-dlp &> /dev/null; then
    echo "yt-dlp not found, attempting to install..."
    if command -v brew &> /dev/null; then
        brew install yt-dlp
    elif command -v apt &> /dev/null; then
        sudo apt update && sudo apt install -y yt-dlp
    else
        pip3 install yt-dlp
    fi
fi

# ============================================
# STEP 2: List available subtitles
# ============================================
echo "Checking available subtitles..."
yt-dlp --list-subs "$VIDEO_URL"

# ============================================
# STEP 3: Try manual subtitles first
# ============================================
echo "Attempting to download manual subtitles..."
if yt-dlp --write-sub --skip-download --output "$OUTPUT_NAME" "$VIDEO_URL" 2>/dev/null; then
    echo "✓ Manual subtitles downloaded successfully!"
    ls -lh ${OUTPUT_NAME}.*
else
    # ============================================
    # STEP 4: Fallback to auto-generated
    # ============================================
    echo "Manual subtitles not available. Trying auto-generated..."
    if yt-dlp --write-auto-sub --skip-download --output "$OUTPUT_NAME" "$VIDEO_URL" 2>/dev/null; then
        echo "✓ Auto-generated subtitles downloaded successfully!"
        ls -lh ${OUTPUT_NAME}.*
    else
        # ============================================
        # STEP 5: Last resort - Whisper transcription
        # ============================================
        echo "⚠ No subtitles available for this video."

        # Get file size
        FILE_SIZE=$(yt-dlp --print "%(filesize_approx)s" -f "bestaudio" "$VIDEO_URL")
        DURATION=$(yt-dlp --print "%(duration)s" "$VIDEO_URL")
        TITLE=$(yt-dlp --print "%(title)s" "$VIDEO_URL")

        echo "Video: $TITLE"
        echo "Duration: $((DURATION / 60)) minutes"
        echo "Audio size: ~$((FILE_SIZE / 1024 / 1024)) MB"
        echo ""
        echo "Would you like to download and transcribe with Whisper? (y/n)"
        read -r RESPONSE

        if [[ "$RESPONSE" =~ ^[Yy]$ ]]; then
            # Check for Whisper
            if ! command -v whisper &> /dev/null; then
                echo "Whisper not installed. Install now? (requires ~1-3GB) (y/n)"
                read -r INSTALL_RESPONSE
                if [[ "$INSTALL_RESPONSE" =~ ^[Yy]$ ]]; then
                    pip3 install openai-whisper
                else
                    echo "Cannot proceed without Whisper. Exiting."
                    exit 1
                fi
            fi

            # Download audio
            echo "Downloading audio..."
            yt-dlp -x --audio-format mp3 --output "audio_%(id)s.%(ext)s" "$VIDEO_URL"

            # Get the actual audio filename
            AUDIO_FILE=$(ls audio_*.mp3 | head -n 1)

            # Transcribe
            echo "Transcribing with Whisper (this may take a few minutes)..."
            whisper "$AUDIO_FILE" --model base --output_format vtt

            # Cleanup
            echo "Transcription complete! Delete audio file? (y/n)"
            read -r CLEANUP_RESPONSE
            if [[ "$CLEANUP_RESPONSE" =~ ^[Yy]$ ]]; then
                rm "$AUDIO_FILE"
                echo "Audio file deleted."
            fi

            ls -lh *.vtt
        else
            echo "Transcription cancelled."
            exit 0
        fi
    fi
fi

# ============================================
# STEP 6: Convert to readable plain text with deduplication
# ============================================
VTT_FILE=$(ls ${OUTPUT_NAME}*.vtt 2>/dev/null || ls *.vtt | head -n 1)
if [ -f "$VTT_FILE" ]; then
    echo "Converting to readable format and removing duplicates..."
    python3 -c "
import sys, re
seen = set()
with open('$VTT_FILE', 'r') as f:
    for line in f:
        line = line.strip()
        if line and not line.startswith('WEBVTT') and not line.startswith('Kind:') and not line.startswith('Language:') and '-->' not in line:
            clean = re.sub('<[^>]*>', '', line)
            clean = clean.replace('&amp;', '&').replace('&gt;', '>').replace('&lt;', '<')
            if clean and clean not in seen:
                print(clean)
                seen.add(clean)
" > "${VIDEO_TITLE}.txt"
    echo "✓ Saved to: ${VIDEO_TITLE}.txt"

    # Clean up temporary VTT file
    rm "$VTT_FILE"
    echo "✓ Cleaned up temporary VTT file"
else
    echo "⚠ No VTT file found to convert"
fi

echo "✓ Complete!"

Note: This complete workflow handles all scenarios with proper error checking and user prompts at each decision point.

Error Handling

Common Issues and Solutions:

1. yt-dlp not installed

1. Attempt automatic installation based on system (Homebrew/apt/pip)
2. If installation fails, provide manual installation link
3. Verify installation before proceeding

2. No subtitles available

1. List available subtitles first to confirm
2. Try both --write-sub and --write-auto-sub
3. If both fail, offer Whisper transcription option
4. Show file size and ask for user confirmation before downloading audio

3. Invalid or private video

1. Check if URL is correct format: https://www.youtube.com/watch?v=VIDEO_ID
2. Some videos may be private, age-restricted, or geo-blocked
3. Inform user of the specific error from yt-dlp

4. Whisper installation fails

1. May require system dependencies (ffmpeg, rust)
2. Provide fallback: "Install manually with: pip3 install openai-whisper"
3. Check available disk space (models require 1-10GB depending on size)

5. Download interrupted or failed

1. Check internet connection
2. Verify sufficient disk space
3. Try again with --no-check-certificate if SSL issues occur

6. Multiple subtitle languages

1. By default, yt-dlp downloads all available languages
2. Can specify with --sub-langs en for English only
3. List available with --list-subs first

Best Practices:

1. ✅ Always check what's available before attempting download (--list-subs)
2. ✅ Verify success at each step before proceeding to next
3. ✅ Ask user before large downloads (audio files, Whisper models)
4. ✅ Clean up temporary files after processing
5. ✅ Provide clear feedback about what's happening at each stage
6. ✅ Handle errors gracefully with helpful messages