The aim of this study was to analyze the repeatability in a stereophotogrammetry digital system used for the evaluation of facial morphology. Thirty healthy Brazilian, 18 to 45 years old (26.71 ± 6.53), had 11 reference landmarks marked on their faces by the same examiner and were photographed with an interval of 1 week by the VECTRA M3. Nine angular measurements (nasolabial, mentolabial, nasofrontal, maxillofacial, nasal, maxillary, mandibular, facial convexity, full facial convexity) and 2 linear measurements (middle facial height and lower facial height) were taken. Repeatability was analyzed by the mean absolute differences, relative error of magnitude, technical error of measurement, intraclass correlation coefficient, and Bland–Altman analysis. Paired t test sought any systematic errors between the acquisitions. Associations among body mass index, age, and the error in the measurements were made using the Pearson correlation coefficient. For the technical error of measurement, the nasolabial and mentolabial angles showed values >2° (clinical limit set for this study). For the relative error of measurement, most of the measures were rated between good and excellent. The maxillofacial angle was only moderate. There was a systematic error for middle facial height. The nasolabial, mentolabial, facial convexity, full facial convexity, maxillofacial, and nasofrontal angles presented intraclass correlation coefficient values rated as excellent. The relationship between age/body mass index and the error found between measurements was not confirmed. The nasolabial and mentolabial angles should be interpreted with caution due to the variability showed. The results found stereophotogrammetry to be repeatable, giving accurate measures within the references established for this study.