Yang Song

Song helped lay the groundwork for diffusion models that generate images by gradually removing noise.^[1] He later co-developed consistency models, which can generate an image in one step, speeding up a process that normally takes many passes.^[2]