Abstract: In recent years, diffusion models such as denoising diffusion probabilistic models and flow matching have made significant progress in speech synthesis. However, these models exhibit a ...