InstaFlow is a method for one-step image generation that distills multi-step diffusion models into a single step using Rectified Flow, a continuous-time generative model framework. It builds on Stable Diffusion and achieves comparable image quality to multi-step methods while dramatically reducing inference time. The project provides code, pre-trained models, Hugging Face demos, and Gradio interfaces for local usage.