The field of time series analysis is moving towards leveraging diffusion models to improve data generation, style transfer, and estimation of fundamental quantities such as local intrinsic dimension and transfer entropy. Diffusion models have shown great promise in disentangling content and style in time series data, enabling effective style transfer and data augmentation. Additionally, they have been used to capture the multi-scaled structure of real-world time series, outperforming traditional methods. The connection between score matching and local intrinsic dimension has also been explored, providing a scalable and competitive estimator. Noteworthy papers include: DiffStyleTS, which introduces a diffusion-based framework for style transfer in time series. WaveletDiff, which trains diffusion models on wavelet coefficients to generate high-quality time series data. A Connection Between Score Matching and Local Intrinsic Dimension, which shows that the denoising score matching loss is a highly competitive LID estimator. TENDE, which proposes a novel approach for estimating transfer entropy using score-based diffusion models.