AI YouTube Shorts Generator

AI-powered tool to automatically generate engaging YouTube Shorts from long-form videos. Uses GPT-4o-mini and Whisper to extract highlights, add subtitles, and crop videos vertically for social media.

Features

• 🎬 Flexible Input: Supports both YouTube URLs and local video files
• 🎤 GPU-Accelerated Transcription: CUDA-enabled Whisper for fast speech-to-text
• 🤖 AI Highlight Selection: GPT-4o-mini automatically finds the most engaging 2-minute segments
• ✅ Interactive Approval: Review and approve/regenerate selections with 15-second auto-approve timeout
• 📝 Auto Subtitles: Stylized captions with Franklin Gothic font burned into video
• 🎯 Smart Cropping:
  • Face videos: Static face-centered crop (no jerky movement)
  • Screen recordings: Half-width display with smooth motion tracking (1 shift/second max)
• 📱 Vertical Format: Perfect 9:16 aspect ratio for TikTok/YouTube Shorts/Instagram Reels
• ⚙️ Automation Ready: CLI arguments, auto-quality selection, timeout-based approvals
• 🔄 Concurrent Execution: Unique session IDs allow multiple instances to run simultaneously
• 📦 Clean Output: Slugified filenames (e.g., my-video-title_short.mp4) and automatic temp file cleanup

Quick Start (No Setup)

Want better results without the setup? The AI Clipping API offers improved clip selection, faster processing, and no dependencies to manage.

---

Installation (Self-Hosted)

Prerequisites

• Python 3.10+
• FFmpeg with development headers
• NVIDIA GPU with CUDA support (optional, but recommended for faster transcription)
• ImageMagick (for subtitle rendering)
• OpenAI API key

Steps

1. Clone the repository:

   git clone https://github.com/SamurAIGPT/AI-Youtube-Shorts-Generator.git
   cd AI-Youtube-Shorts-Generator

2. Install system dependencies:

   Ubuntu/Debian:

   sudo apt install -y ffmpeg libavdevice-dev libavfilter-dev libopus-dev \
     libvpx-dev pkg-config libsrtp2-dev imagemagick

   macOS:

   brew install ffmpeg imagemagick

   Windows:

   • Install FFmpeg and add to PATH
   • Install ImageMagick

3. Fix ImageMagick security policy (Linux only, required for subtitles):

   sudo sed -i 's/rights="none" pattern="@\*"/rights="read|write" pattern="@*"/' /etc/ImageMagick-6/policy.xml

4. Create and activate virtual environment:

   python3.10 -m venv venv
   source venv/bin/activate

5. Install Python dependencies:

   pip install -r requirements.txt

6. Set up environment variables:

   Create a .env file in the project root:

   OPENAI_API=your_openai_api_key_here

CPU-Only Installation

If you don't have an NVIDIA GPU, see INSTALL_CPU.md for CPU-only setup instructions.

Docker Installation

# Build and run with Docker Compose
docker-compose up --build

# Or build manually
docker build -t ai-shorts-generator .
docker run -v $(pwd)/.env:/app/.env -v $(pwd)/videos:/app/videos ai-shorts-generator

Usage

With YouTube URL (Interactive)

./run.sh
# Then enter YouTube URL when prompted
# You'll be able to select video resolution (5s timeout, auto-selects highest)

With YouTube URL (Command-Line)

./run.sh "https://youtu.be/VIDEO_ID"

With Local Video File

./run.sh "/path/to/your/video.mp4"

Batch Processing Multiple URLs

Create a urls.txt file with one URL per line, then:

# Process all URLs sequentially with auto-approve
xargs -a urls.txt -I{} ./run.sh --auto-approve {}

Or without auto-approve (will prompt for each):

xargs -a urls.txt -I{} ./run.sh {}

Resolution Selection

When downloading from YouTube, you'll see:

Available video streams:
  0. Resolution: 1080p, Size: 45.2 MB, Type: Adaptive
  1. Resolution: 720p, Size: 28.1 MB, Type: Adaptive
  2. Resolution: 480p, Size: 15.3 MB, Type: Adaptive

Select resolution number (0-2) or wait 5s for auto-select...
Auto-selecting highest quality in 5 seconds...

• Enter a number to select that resolution immediately
• Wait 5 seconds to auto-select highest quality (1080p)
• Invalid input falls back to highest quality

How It Works

1. Download/Load: Fetches from YouTube or loads local file
2. Resolution Selection: Choose video quality (5s timeout, auto-selects highest)
3. Extract Audio: Converts to WAV format
4. Transcribe: GPU-accelerated Whisper transcription (~30s for 5min video)
5. AI Analysis: GPT-4o-mini selects most engaging 2-minute segment
6. Interactive Approval: Review selection, regenerate if needed, or auto-approve in 15s
7. Extract Clip: Crops selected timeframe
8. Smart Crop:
   • Detects faces → static face-centered vertical crop
   • No faces → half-width screen recording with motion tracking
9. Add Subtitles: Burns Franklin Gothic captions with blue text/black outline
10. Combine Audio: Merges audio track with final video
11. Cleanup: Removes all temporary files

Output: {video-title}_{session-id}_short.mp4 with slugified filename and unique identifier

Interactive Workflow

After AI selects a highlight, you'll see:

============================================================
SELECTED SEGMENT DETAILS:
Time: 68s - 187s (119s duration)
============================================================

Options:
  [Enter/y] Approve and continue
  [r] Regenerate selection
  [n] Cancel

Auto-approving in 15 seconds if no input...

• Press Enter or y to approve
• Press r to regenerate a different selection (can repeat multiple times)
• Press n to cancel
• Wait 15 seconds to auto-approve (perfect for automation)

Configuration

Subtitle Styling

Edit Components/Subtitles.py - search for TextClip:

• Font: font='Franklin-Gothic' (requires Franklin Gothic installed, or change to any system font)
• Size: fontsize=80
• Color: color='#2699ff' (blue)
• Outline: stroke_color='black', stroke_width=2

To list available fonts:

convert -list font | grep -i "font:"

Highlight Selection Criteria

Edit Components/LanguageTasks.py:

• Prompt: Modify the system variable to adjust what's "interesting, useful, surprising, controversial, or thought-provoking"
• Model: Change model="gpt-4o-mini" in ChatOpenAI() call
• Temperature: Adjust temperature=1.0 (higher = more creative)

Motion Tracking

Edit Components/FaceCrop.py - search for use_motion_tracking:

• Update frequency: update_interval = int(fps) - currently 1 shift/second
• Smoothing: 0.90 * smoothed_x + 0.10 * target_x - 90% previous, 10% new
• Motion threshold: motion_threshold = 2.0

Face Detection

Edit Components/FaceCrop.py - search for detectMultiScale:

• Sensitivity: minNeighbors=8 - Higher = fewer false positives
• Minimum size: minSize=(30, 30) - Minimum face size in pixels

Video Quality

Edit Components/Subtitles.py - search for write_videofile:

• Bitrate: bitrate='3000k'
• Preset: preset='medium' (options: ultrafast, fast, medium, slow, veryslow)

Output Files

Final videos are named: {video-title}_{session-id}_short.mp4

Example: my-awesome-video_a1b2c3d4_short.mp4

• Slugified title: Lowercase, hyphens instead of spaces
• Session ID: 8-character unique identifier for traceability
• Resolution: Matches source video height (720p → 404x720, 1080p → 607x1080)

Concurrent Execution

Run multiple instances simultaneously:

./run.sh "https://youtu.be/VIDEO1" &
./run.sh "https://youtu.be/VIDEO2" &
./run.sh "/path/to/video3.mp4" &

Each instance gets a unique session ID and temporary files, preventing conflicts.

Troubleshooting

CUDA/GPU Issues

# Verify CUDA libraries
export LD_LIBRARY_PATH=$(find $(pwd)/venv/lib/python3.10/site-packages/nvidia -name "lib" -type d | paste -sd ":" -)

The run.sh script handles this automatically.

No Subtitles

Ensure ImageMagick policy allows file operations:

grep 'pattern="@\*"' /etc/ImageMagick-6/policy.xml
# Should show: rights="read|write"

Face Detection Issues

• Video needs visible faces in first 30 frames
• For screen recordings, automatic motion tracking applies
• Low-resolution videos may have less reliable detection

Looking for Better Results?

The AI Clipping API uses an improved algorithm that produces higher-quality clips with better highlight detection.

Contributing

Contributions are welcome! Please fork the repository and submit a pull request.

License

This project is licensed under the MIT License.

Related Projects

• AI Influencer Generator
• Text to Video AI
• Faceless Video Generator
• AI B-roll Generator
• No-code YouTube Shorts Generator