video-to-shorts

A tool for automatically converting long-form videos into engaging short-form content using AI-powered scene analysis.

Features

• 🎬 Automatically analyze video content using OpenAI's GPT-4 Vision
• ✂️ Intelligently trim, speed up, or keep scenes based on content
• 🎯 Customizable analysis criteria and confidence thresholds
• 📱 Output optimized for vertical video platforms
• 🚀 Process multiple videos in parallel using all CPU cores
• ⚡ Parallel frame analysis within each video
• 🎯 Process specific portions of videos with start/end times
• 🔧 Flexible configuration via JSON files or command line

Examples

• Lott Gaming - my gaming YouTube channel

Getting Started

Prerequisites

• Node.js 18+
• FFmpeg (automatically installed on most platforms)
• OpenAI API key

Installation

Using yarn:

yarn global add @shipworthy/video-to-shorts

Or using pnpm:

pnpm add -g @shipworthy/video-to-shorts

Quick Start

Checkout the CLI documentation for more detailed usage instructions and options.

Process a directory of videos or a single video:

# Process a directory
video-to-shorts /path/to/videos

# Process a single video
video-to-shorts /path/to/video.mp4

Advanced Usage

Checkout the CLI documentation for more detailed usage instructions and options.

# Process videos with debug logging
video-to-shorts ./videos --debug

# Process 2 videos concurrently (max: number of CPU cores)
video-to-shorts ./videos -j 2

# Analyze every 60th frame, 3 frames in parallel
video-to-shorts ./videos -f 60 -p 3

# Process specific portion of videos
video-to-shorts ./videos -s 00:30 -e 02:00

# Specify custom output directory
video-to-shorts ./videos -o ./shorts

# Process a video with start/end times in filename
video-to-shorts ./00:00:30-00:02:00.mp4

Configuration

The analyzer configuration can be provided in two ways:

1. As a JSON file (video-to-shorts.json):

{
  "keep": "Content is interesting, important, or high quality",
  "remove": "Content is unwanted, irrelevant, or of poor quality",
  "speedUp": "Content is repetitive, slow, or less interesting but still relevant",
  "confidenceThreshold": 0.7,
  "openaiApiKey": "your-api-key"  // Optional, can also be set via environment variable or CLI flag
}

2. As a JSON string directly in the command line:

video-to-shorts ./videos -c '{"keep": "Content is engaging", "confidenceThreshold": 0.8}'

For more detailed usage instructions and options, see the CLI documentation.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

For detailed contributing guidelines and instructions on updating the changelog, please see our CONTRIBUTING.md.

Development

This is a monorepo using Yarn workspaces. The main packages are:

• @shipworthy/video-processor: Core video processing and analysis logic
• @shipworthy/video-to-shorts: CLI interface

Setup

1. Clone the repository

git clone https://github.com/lottamus/video-to-shorts.git
cd video-to-shorts

2. Install dependencies:

yarn install

Building Packages

# Build all packages
yarn build

# Build a specific package
yarn build --filter @shipworthy/video-processor

Running Tests

# Run all tests
yarn test

# Run tests for a specific package
yarn test --filter @shipworthy/video-processor

Running CLI Locally

This command will build and run the CLI.

# Build & run the CLI for a single video with debug logging
yarn start -- ./videos/your-video.mp4 -d

# Build & run the CLI for a directory of videos with debug logging
yarn start -- ./videos -d

Changelog

For changes that affect functionality or the user experience of the project, please update the changelog by running:

yarn changelog

License

By using this project, you agree your creations and contributions will be licensed under the same terms, MIT, as the project.

Thank you for your contributions! 🚀