Functional near-infrared spectroscopy (fNIRS) research is hindered by the scarcity of high-quality, publicly available datasets, limiting rigorous benchmarking of signal processing algorithms and quality-control metrics. Existing synthetic approaches either rely on fixed, literature-based parameters without empirical grounding, yield only partial ground truth, or require large training corpora. This work presents a pipeline for generating realistic synthetic fNIRS datasets that addresses these gaps through three contributions: subject-specific parameter extraction from real resting-state recordings using spectral parameterization (FOOOF), fully synthetic signal generation in the frequency domain ensuring complete ground truth over all signal components, and parametric motion artifact injection at the raw intensity level enabling controlled modulation of scalp coupling index (SCI) distributions.