A rectified flow-based generative model that improves generation quality and straightens ODE solution trajectories by incorporating real images via slerp into the reflow training process.