OMG (Occlusion-friendly Multi-concept Generation) is a research framework enabling high-quality image generation with multiple characters and styles simultaneously. It builds on diffusion models and supports integration with LoRA adapters for character and style customization, as well as InstantID for multi-identity face generation from single reference images. The framework handles complex scenes where characters may occlude each other.