Describir: Feasibility of Conditional Variational Autoencoders for Phase-Averaged Synthetic Time Series