YouTube to Shorts Automation System

Complete Bash + Python automation system to convert YouTube videos into viral short clips with interactive subtitles.

📁 Directory Structure

~/automation/
├── scripts/              # All automation scripts
│   ├── download_video.sh           # Download YouTube videos
│   ├── transcribe_video.sh         # Transcribe videos using whisper.cpp
│   ├── romanize_subtitles.py       # Convert Urdu/Hindi to Roman script
│   ├── suggest_clip.py             # AI-powered viral clip suggestions
│   ├── create_short.sh             # Create short clips using ffmpeg
│   ├── create_interactive_subtitles.py  # Add interactive subtitles
│   └── process_all.sh              # Complete pipeline
├── download/             # Raw downloaded videos (MP4)
├── transcribe/           # [Optional] Audio files (WAV)
├── subtitles/            # Generated subtitles (.srt/.txt) and clip suggestions (.json)
├── shorts/               # Final short videos with interactive subtitles
├── requirements.txt      # Python dependencies
├── setup.sh             # Installation script
└── README.md            # This file

🚀 Quick Start

1. Setup Dependencies

cd ~/automation
./setup.sh

2. Process a YouTube Video (Complete Pipeline)

# First activate the virtual environment (if using Arch Linux)
source ~/automation/venv/bin/activate

./scripts/process_all.sh "https://www.youtube.com/watch?v=YOUR_VIDEO_ID"

That's it! The system will:

1. Download the video
2. Transcribe it using Whisper
3. Romanize non-English text
4. Suggest viral clip segments using AI
5. Create short clips
6. Add interactive subtitles

📋 Individual Script Usage

Download Video Only

./scripts/download_video.sh "https://www.youtube.com/watch?v=VIDEO_ID"

Transcribe Videos Only

./scripts/transcribe_video.sh

Romanize Subtitles Only

# Activate virtual environment first (Arch Linux)
source ~/automation/venv/bin/activate
./scripts/run_python.sh romanize_subtitles.py

Suggest Viral Clips Only

# Activate virtual environment first (Arch Linux)
source ~/automation/venv/bin/activate
./scripts/run_python.sh suggest_clip.py

Create Short Clips Only

./scripts/create_short.sh

Add Interactive Subtitles Only

# Activate virtual environment first (Arch Linux)
source ~/automation/venv/bin/activate
./scripts/run_python.sh create_interactive_subtitles.py

🔧 Configuration

OpenRouter API

The system uses DeepSeek v3 (free) model via OpenRouter for:

• Language detection and romanization
• Viral clip suggestions

API key is configured in:

• scripts/romanize_subtitles.py
• scripts/suggest_clip.py

Whisper Model

The system automatically detects and uses the best available model:

• medium (1.5 GB) - Best accuracy (if available)
• base.en (74 MB) - Good accuracy, English only
• base (74 MB) - Good accuracy, multilingual

Priority order: medium → base.en → base

You already have both ggml-medium.bin and ggml-base.en.bin models, so the system will use the medium model for better transcription accuracy!

📊 Output Files

Download Directory

• Video Title.mp4 - Original downloaded video

Subtitles Directory

• Video Title.txt - Text transcript (romanized if needed)
• Video Title.srt - SRT subtitle file with timestamps
• Video Title.clip.json - AI-suggested clip with timestamps and reason

Shorts Directory

• Video Title_short.mp4 - Short clip without subtitles
• Video Title_with_subtitles.mp4 - Final video with interactive subtitles

🎯 Features

Smart Language Detection

• Automatically detects Urdu/Hindi scripts
• Romanizes non-English text for better accessibility
• Preserves English content unchanged

AI-Powered Clip Suggestions

• Analyzes transcript content for viral potential
• Suggests optimal 30-90 second segments
• Provides reasoning for each suggestion
• Looks for emotional moments, insights, and quotable content

Interactive Subtitles

• Word-level timing (when available)
• Highlighted current words
• Professional styling with stroke borders
• Optimized for mobile viewing

Error Handling

• Comprehensive logging with timestamps
• Graceful failure handling
• Detailed error messages
• Progress tracking

🛠️ Dependencies

System Requirements

• Arch Linux (or other Linux distributions)
• Python 3.8+
• FFmpeg
• Git

Installed by Setup Script

• yt-dlp - YouTube video downloader
• whisper.cpp - Fast speech recognition
• ffmpeg - Video processing
• Python packages: requests, moviepy, langdetect

⚡ Performance Tips

For Faster Processing

1. Use smaller Whisper models for speed
2. Process shorter videos (< 30 minutes)
3. Use SSD storage for faster I/O

For Better Quality

1. Use larger Whisper models (medium or large)
2. Ensure good audio quality in source videos
3. Review AI suggestions before creating clips

🐛 Troubleshooting

Common Issues

"whisper.cpp not found"

cd ~
git clone https://github.com/ggerganov/whisper.cpp.git
cd whisper.cpp
make
bash ./models/download-ggml-model.sh base.en

"No MP4 files found"

Ensure videos are downloaded to ~/automation/download/ directory.

"MoviePy import error"

pip install --user moviepy

"OpenRouter API error"

Check your internet connection and API key in the Python scripts.

Logs

All scripts provide detailed logging with timestamps. Check console output for specific error messages.

📈 Optimization for Different Content Types

Educational Content

• Longer clips (60-90s) for complete thoughts
• Focus on key insights and actionable advice

Entertainment Content

• Shorter clips (30-60s) for quick engagement
• Highlight funny or surprising moments

Interview Content

• Find quotable statements and revelations
• Look for emotional or controversial moments

🔒 Security Notes

• API key is stored in script files (consider environment variables for production)
• Downloaded content respects YouTube's terms of service
• No personal data is transmitted to AI services

📱 Social Media Ready

Output videos are optimized for:

• TikTok (vertical format compatible)
• YouTube Shorts
• Instagram Reels
• Twitter/X videos

🤝 Contributing

Feel free to enhance the scripts:

1. Add support for more languages
2. Improve subtitle styling
3. Add more AI models
4. Optimize performance

📄 License

This automation system is for personal and educational use. Respect YouTube's terms of service and content creators' rights.