Rectified flow offers a simplified yet unified perspective on flow- and diffusion-based generative modeling and has been widely applied to state-of-the-art image, audio, and video generation. It was derived by simplifying diffusion models to the extreme both conceptually and algorithmically, thereby uncovering a principled framework for rectifying transport maps to learn straightened ODE dynamics for fast discretized inference.
This series of tutorials on rectified flow addresses topics that are often sources of confusion and clarifies the connections with other methods.
For those eager to dive deeper, we provide:
- A comprehensive codebase for practical exploration.
- Lecture notes containing detailed theoretical derivations.
If you have questions regarding the blog posts, codebase, or notes, please feel free to reach out via this email.
