Providing computationally efficient and robust high-resolution projections of coastal wave directional spectra is a key challenge in coastal prediction under future scenarios. On one hand, the available wave spectra outputs from CMIP6 global climate models (GCMs) are limited and scarce; on the other hand, dynamic downscaling models are often computationally expensive, making it impractical to explore future scenarios in a probabilistic and detailed manner. To address these issues and overcome existing limitations, the present study proposes a novel and innovative framework to obtain the full directional wave spectra at a specific location using representative bulk parameters. For this purpose, clustering techniques and the analysis of their conditioned probabilities are established and validated over a common historical period. Once the spectra are generated and bias-corrected using a robust directional technique, the hybrid additive model BinWaves [1] is employed to efficiently downscale the waves to the coast, providing highly detailed and reliable information for analyzing probabilistic future impacts under different scenarios.