Excited states and dynamics in materials and at interfaces underpin a wide range of important physical processes such as light harvesting, spectroscopy, and photochemistry. Modeling these processes from first-principles is challenging for density functional methods, however, largely due to deficiencies in exchange-correlation functionals and basis set limitations. In this talk I will discuss time-dependent density functional theory (TDDFT) approaches to optical and X-ray absorption of insulators using bulk-mimicking finite clusters. Combining range-separated hybrid functionals with non-Hermitian real-time propagation enables TDDFT to accurately capture Rydberg-like and post-ionization states without input from experiment. Application to valence and core-level spectra of various SiO2 polymorphs will be presented, along with methodological details including embedding, time propagation, and high-performance computing aspects.