Pseudonymised, analysis-ready participant-level dataset underpinning the study "Big Five Personality Traits as Moderators of Depression-Facial Action Unit Relationships in Naturalistic Free-Speech Video". It covers 103 Arabic-speaking university students (aged 18-25) recruited at Tartous University, Syria. Each participant recorded an approximately one-minute free-speech video on a self-selected topic and completed the PHQ-9 depression screening questionnaire and the Arabic short form of the Big Five Inventory-2 (BFI-2-XS). The deposit comprises: (i) mean facial action unit (AU) intensities for 20 AUs extracted from the video recordings using the py-feat toolbox, together with three derived composite measures (Depressive Facial Index, Positive Affect Suppression, Facial Expressivity Variability); (ii) PHQ-9 total scores; (iii) Big Five domain scores; (iv) age band and sex/gender; (v) valid frame counts per recording; (vi) a data dictionary and README describing all variables, processing steps and access conditions; and (vii) non-disclosive aggregate outputs supporting verification of the analyses reported in the associated manuscript. The original facial video and audio recordings are NOT included in this deposit and will not be shared in any form. Data files are provided in CSV format (UTF-8). This dataset is related to, but distinct from, the ROSE acoustic-feature record (https://doi.org/10.5525/gla.researchdata.2332). The participant-level FVPD file is proposed as a separate deposit for RDM assessment under its own consent documentation. Neither raw video nor raw audio is included.