# VerticalX (VX) - Open Source Video Reframing Engine

> Free, open-source automated engine to convert 16:9 widescreen horizontal videos (YouTube podcasts, webinars, interviews, vlogs) into high-performing 9:16 portrait vertical clips for TikTok, Instagram Reels, and YouTube Shorts. Powered by local AI face tracking, shot detection, and OpenCV computer vision.

## Key Features
- **100% Offline & Private**: Runs completely on local hardware. Zero cloud video uploads.
- **No Subscriptions / Watermarks**: Free open-source software under the MIT License.
- **Smart Focus Face Tracking**: Dynamically tracks active speakers across widescreen space with Exponential Moving Average (EMA) camera motion smoothing.
- **6 Cinematic Layout Modes**:
  1. `Smart Focus` (AI speaker tracking for solo presenters/vlogs)
  2. `Dual Stack` (Split-screen top/bottom for 2-person podcast interviews)
  3. `Blurred Background` (Gaussian-blurred background padding)
  4. `9:16 Center Crop` (Static center viewport crop)
  5. `Fit (Black Padding)` (Preserves 100% of widescreen frame)
  6. `Fit & Zoom In` (Custom zoom adjustment)
- **Automatic Scene Cut Detection**: Uses PySceneDetect to segment long videos into individual clips automatically.
- **Auto-Subtitles via Whisper AI**: OpenAI Whisper integration generates frame-accurate WebVTT vertical subtitles.
- **Hardware Acceleration**: Probes Apple Silicon GPU (`h264_videotoolbox`) and NVIDIA GPU (`h264_nvenc`) automatically with CPU fallback (`libx264`).

## Quickstart
```bash
# Clone repository
git clone https://github.com/vinaykolupula/VerticalX.git
cd VerticalX

# Install dependencies
pip install -r requirements.txt

# Run Web Studio Editor
python web_ui/app.py
```

## Documentation Links
- [GitHub Repository](https://github.com/vinaykolupula/VerticalX)
- [Architecture Spec](https://github.com/vinaykolupula/VerticalX/blob/main/architecture.md)
- [Full Technical LLM Docs](https://verticalx.vinaykolupula.workers.dev/llms-full.txt)