In the context of digital transformation in tourism, tourism live streaming (TLS) has emerged as a novel virtual tourism experience, enabling participants to interact in real time with hosts/KOLs and scenic destinations. Drawing on dual-process theories of emotion, this study develops and validates a conceptual model of emotional and sentimental experience in TLS, integrating both explicit (verbal, self-reported) and implicit (facial expression-based) expression processes. Using a convergent mixed-method design, multimodal data from 41 tourism live streaming informants were collected, including facial emotion recognition via FaceReader software and semi-structured interviews. The results show that implicit expressions effectively captured short-lived and often negative emotions, such as anger or sadness, which were frequently underrepresented in verbal responses. In contrast, participants' self-report tended to highlight positive emotional memories, such as satisfaction, happiness, and inspiration. Dynamic temporal analyses further revealed distinct emotional patterns across three key phases of livestreaming experience: an anticipatory beginning, a peak interaction phase with high emotional volatility, and a reflective ending. These patterns suggest that emotional experience in TLS is phase-dependent and socially constructed through both interaction quality and content resonance. Furthermore, sentimental outcomes are reflected in core dimensions such as "enjoyment and autonomy", "inspiration and arousal", and "existence and identity".