A study published in *Scientific Reports* introduces BabyFlow, a 3D modeling framework for realistic and expressive infant faces. The model uses 23 anatomical landmarks, with 19 central landmarks used for facial expression recognition. The expression space is built from 47 action units (AUs) from the Facial Action Coding System (FACS) available in the FaceWarehouse dataset, with composite expressions created as combinations of these AUs. Validation of expression transfer assessed anatomical plausibility through AU-based consistency, verifying that spatial activation patterns of AUs are preserved across domains. The study found that while the magnitude of deformations is lower in BabyFlow, this is consistent with differences in facial geometry and scale between infants and adults.