A sophisticated dual-video switching application built for fun and learning advanced C++ programming with Claude Code. This project demonstrates hardware-accelerated video processing, multiple rendering backends, and clean software architecture patterns.
This application was developed as a learning exercise to explore:
- Advanced C++ programming techniques
- FFmpeg integration and video processing
- Hardware acceleration (NVIDIA NVDEC, DirectX, CUDA)
- Modern graphics APIs (DirectX 11, OpenGL 4.6)
- Clean architecture and design patterns
- AI-assisted development with Claude Code
- Multi-Resolution Support - Videos don't need identical resolutions
- Hardware Acceleration - NVIDIA NVDEC with automatic software fallback
- Codec Support - H.264 and H.265 in MP4 containers
- Advanced Switching - Multiple algorithms with different performance characteristics
- Dual Backend Support - DirectX 11 or OpenGL (compile-time selection)
- Hardware Acceleration - D3D11VA (DirectX) or CUDA interop (OpenGL)
- YUV Rendering - Direct YUV rendering with hardware color conversion
- Dynamic Windows - Resizable windows with renderer adjustment
- Fullscreen Mode - F11 toggle with state preservation
- Multi-Monitor Support - Window size limited by display resolution
- Extensible Input - Modular trigger framework
./dual_stream video1.mp4 video2.mp4# Different switching algorithms
./dual_stream video1.mp4 video2.mp4 --algorithm predecoded
./dual_stream video1.mp4 video2.mp4 -a keyframe-sync
# Playback speed control
./dual_stream video1.mp4 video2.mp4 --speed 2.0
# Debug mode
./dual_stream video1.mp4 video2.mp4 --debug- 1/2 Keys - Switch between videos
- F11 - Toggle fullscreen
- ESC - Exit application
- OS: Windows 10/11 (64-bit)
- Compiler: Visual Studio 2019+ or MinGW-w64
- CMake: 3.15+
- FFmpeg: 7.1.1 (automatically downloaded by CMake)
- GPU: NVIDIA RTX series (recommended) or DirectX 11 compatible
mkdir build
cd build
# Both DirectX 11 and OpenGL renderers are always available
cmake ..
cmake --build . --config Release| Algorithm | Latency | Memory Usage | Quality | Use Case |
|---|---|---|---|---|
| Immediate (Default) | ~1-5ms | 1x | Good | General purpose |
| Predecoded | ~0ms | 2x | Perfect | Zero-latency switching |
| Keyframe Sync | Variable | 1x | Perfect | Artifact-free transitions |
src/
├── core/ # Foundation services (CLI, logging, FFmpeg init)
├── ui/ # Window management and input handling
├── video/ # Video processing pipeline
│ ├── demux/ # Container parsing (MP4)
│ ├── decode/ # Hardware/software decoding
│ ├── switching/ # Video switching strategies
│ └── triggers/ # Input trigger system
└── rendering/ # Multi-backend rendering (D3D11/OpenGL)
The application uses modern C++ design patterns including:
- Strategy pattern for switching algorithms
- Factory pattern for renderer/trigger creation
- RAII for resource management
- Interface segregation for extensibility
- Container Parsing: ~5% CPU
- Hardware Decoding: ~10% GPU (NVDEC)
- Format Conversion: ~5% GPU (shaders)
- Rendering: ~5% GPU (3D pipeline)
- Container Parsing: ~5% CPU
- Software Decoding: ~40% CPU
- Format Conversion: ~20% CPU
- Rendering: ~5% GPU
Each component has detailed technical documentation:
This project provided hands-on experience with:
- Modern C++17/20 features
- Memory management and RAII
- Template metaprogramming
- Design patterns implementation
- FFmpeg library integration
- Hardware-accelerated video decoding
- YUV color space handling
- Container format parsing
- DirectX 11 API usage
- OpenGL 4.6 core profile
- CUDA-OpenGL interoperability
- Shader programming
- Clean architecture principles
- Dependency injection
- Strategy and factory patterns
- Extensible plugin systems
- CMake build systems
- Cross-platform development
- AI-assisted programming
- Documentation-driven development
This project was developed using Claude Code, Anthropic's AI coding assistant. The AI helped with:
- Architecture design and code reviews
- Complex FFmpeg integration
- Hardware acceleration implementation
- Cross-platform compatibility
- Comprehensive documentation
The experience demonstrates how AI can accelerate learning and enable building sophisticated applications that would typically require years of domain expertise.
This is a personal learning project created for educational purposes. Feel free to use it as a reference for your own multimedia programming journey!
Built with curiosity, powered by Claude Code 🚀