This project provides a lightweight, dependency-free C/C++ implementation for running various diffusion models including Stable Diffusion, Flux, Wan2.1, Qwen-Image, and Z-Image. It uses the ggml library (similar to llama.cpp) to enable efficient inference on CPU and GPU, supporting text-to-image, image-to-image, and video generation tasks. The project includes an embedded web UI for convenient model interaction.