Stanford Emotional Narratives Dataset

A publicly available dataset of short video clips of people telling emotionally significant personal stories, created by Ong et al. at Stanford (2019) to support multimodal emotion-recognition research. Each video is annotated continuously for valence, arousal, and observer-rated emotional content, aligned to the spoken transcript. SEND is widely used in affective captioning and expressive-media research because the natural, unscripted speech exhibits the spontaneous prosodic and emotional variation that laboratory-acted datasets typically lack. Accessibility researchers use SEND to evaluate how well captioning, haptic, or expressive-typography systems convey real-world emotional content to d/DHH viewers.