We present an emotion-centric video foundation model trained with fine-grained captions and rationales via affective-tree reasoning guidance, achieving high-level emotional intelligence for video understanding.