Core FM Generative Evaluation: Essential Derived Data and Analysis Code

Version 3.0.0 updates the reproducibility package to reflect the final Core FM implementation and the complete analytical workflow reported in the current study. This version includes updated derived real-audio features, synthetic-evaluation outputs, periodic-signal validation, singer-grouped classification results, recalibrated cVAE v4 and WGAN-GP outputs, closed-loop recomputation results, revised figures and tables, updated Python and R analysis code, and the current reference implementation of the Core Frequency-Modulation Regularity Index (Core FM). The package documents the Core FM formulation based on dynamic stability, nonlinear regularity, phase-slip continuity, and rolling consistency, with weights of 0.25, 0.20, 0.10, and 0.45, respectively. It also documents proportional weight redistribution when individual components are unavailable, the treatment of phase-slip continuity across exported scales, and the score-adjustment rules used by the current VibratoScope implementation. This record contains essential derived acoustic-feature tables, controlled synthetic vibrato contours, synthetic WAV-derived evaluation outputs, perfectly periodic signal-validation results, calibrated generated vibrato contours, mixed-effects model outputs, singer-grouped baseline-classification results, conditional variational autoencoder and Wasserstein generative adversarial network outputs, closed-loop recomputation results, final tables, figures, consistency checks, and analysis scripts supporting a computational evaluation of Core FM in vocal vibrato. The deposit is restricted to derived acoustic features, synthetic and generated contours, statistical and computational outputs, documentation, and code. It is intended to support independent reproduction and evaluation of the reported analyses while respecting copyright and platform restrictions related to the source recordings.