vidcom

---

Video Compilations

---

📋 Table of Contents

• Overview
• Features
• Installation
• Usage
• Configuration
• Documentation
• Contributing
• Roadmap
• Support
• License
• Acknowledgements

---

🔍 Overview

VIDCOM is a high-performance video compilation pipeline designed for gaming content creators. It automatically detects highlights (kills, headshots, clutches) from gaming streams and compiles them into YouTube Shorts-ready vertical videos.

Why vidcom?

• Pure C Implementation - Minimal dependencies, maximum performance with ONNX Runtime
• GPU Accelerated - NVENC/NVDEC encoding with CUDA inference for real-time processing
• Game-Specific Detection - Optimized ROI regions for Fortnite, Valorant, CS2, Overwatch, Apex
• End-to-End Pipeline - From raw stream to uploaded YouTube Short in one command

---

✨ Features

• 🎮 YOLO Highlight Detection - ML-based detection of kills, headshots, assists, and action moments
• 🚀 GPU Acceleration - CUDA inference + NVENC encoding for real-time processing
• 📦 Multi-Game Support - Fortnite, Valorant, CS2, Overwatch, Apex Legends
• 🔧 Automated Pipeline - Fetch → Detect → Encode → Upload workflow
• ⚡ Batch Processing - Process entire streams or multiple videos at once
• 🎬 YouTube Integration - Direct upload to YouTube with metadata

---

📥 Installation

Prerequisites

• NVIDIA GPU with CUDA 11.8+ support
• FFmpeg 6.0+ with NVENC support
• ONNX Runtime 1.20+ (bundled)
• Linux (Ubuntu 22.04+ recommended)

Quick Start

# Clone the repository
git clone https://github.com/xaoscience/vidcom.git
cd vidcom

# Install dependencies
./scripts/build-deps.sh

# Build
make

# Verify installation
./build/vidcom help

Building from Source

# Install build dependencies (Ubuntu/Debian)
sudo apt install build-essential libavcodec-dev libavformat-dev \
    libavutil-dev libswscale-dev libsqlite3-dev

# Build with CUDA support
make CUDA=1

# Or CPU-only build
make CUDA=0

---

🚀 Usage

Highlight Detection

Detect highlights (kills, headshots, etc.) in gaming footage:

# Basic usage
./build/vidcom highlights video.mp4

# With game-specific optimization
./build/vidcom highlights stream.mp4 --game fortnite --confidence 0.6

# Verbose output
./build/vidcom highlights gameplay.mp4 --game valorant -v

Batch Processing with Auto-Detection

# Auto-detect and process highlights
./scripts/process-batch.sh stream.mp4 --auto --game fortnite --max-clips 5

# Upload to YouTube
./scripts/process-batch.sh stream.mp4 --auto --game valorant --upload --privacy unlisted

Manual Manifest Processing

Create a manifest file (highlights.jsonl):

{"input": "stream.mp4", "start": "01:23:45", "duration": 55, "title": "Epic Kill", "tags": "fortnite,gaming"}
{"input": "stream.mp4", "start": "02:10:30", "duration": 45, "title": "Clutch Win", "tags": "fortnite,clutch"}

Process the manifest:

./scripts/process-batch.sh highlights.jsonl --upload

Training a Custom Model

1. Collect training data:

./scripts/collect-training-data.sh videos/ --game fortnite --fps 2

2. Label frames with LabelImg or CVAT

3. Train the model:

./scripts/train-highlight-model.sh --data datasets/highlight_detection/data.yaml --epochs 100

---

⚙️ Configuration

Configuration File

Edit config/vidcom.conf:

[general]
verbose = 1
use_gpu = true
device_id = 0

[highlight_detection]
# Path to YOLO ONNX model
highlight_model = models/highlight_yolov8n.onnx

# Detection confidence (0.0-1.0)
confidence_threshold = 0.5

# Game type for optimized detection
game = fortnite

# Segment timing
padding_before = 4.0
padding_after = 2.0
merge_threshold = 3.0

[encoding]
shorts_width = 1080
shorts_height = 1920
quality = 20
max_duration = 60

Supported Games

Game	Flag	Detection Region
Fortnite	--game fortnite	Top-right kill feed
Valorant	--game valorant	Center kill confirmation
CS2	--game csgo2	Top-right killfeed
Overwatch	--game overwatch	Center elimination popup
Apex Legends	--game apex	Top-right kill feed

Highlight Classes

Class	Description
KILL	Standard elimination
HEADSHOT	Headshot indicator
ASSIST	Assist notification
DOWN	Enemy knocked (BR games)
MULTI_KILL	Double/triple/quad kills
CLUTCH	1vX clutch situations
ACTION	High-action moments

---

📚 Documentation

Document	Description
📖 Getting Started	Quick start guide
📋 API Reference	Complete API documentation
🔧 Configuration	Configuration options
❓ FAQ	Frequently asked questions

---

🤝 Contributing

Contributions are welcome! Please read our Contributing Guidelines before submitting PRs.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

See also: Code of Conduct | Security Policy

---

🗺️ Roadmap

• {{COMPLETED_FEATURE_1}}
• {{COMPLETED_FEATURE_2}}
• {{PLANNED_FEATURE_1}}
• {{PLANNED_FEATURE_2}}
• {{PLANNED_FEATURE_3}}

See the open issues for a full list of proposed features and known issues.

---

💬 Support

• 📧 Email: {{SUPPORT_EMAIL}}
• 💻 Issues: GitHub Issues
• 💬 Discussions: GitHub Discussions
• 📝 Wiki: GitHub Wiki

---

📄 License

Distributed under the GPL-3.0 License. See LICENSE for more information.

---

🙏 Acknowledgements

• {{ACKNOWLEDGMENT_1}}
• {{ACKNOWLEDGMENT_2}}
• {{ACKNOWLEDGMENT_3}}

---

⬆️ Back to Top