Live streaming has become a major component of the e-commerce economy, influencing consumer purchasing behavior. However, the optimal strategies for streamers to increase viewers’ purchase intention remain unclear, with most studies focusing on qualitative aspects. This research aims to explore the effects of facial cues, including smile and eye gaze, on purchase intention using a quantitative approach during a dynamic process. Guided by Brownian lens and SOR theories, 36 participants were exposed to 16 video stimuli with controlled facial cues, followed by surveys to assess perceptions and purchase intent. Eye-tracking data captured subconscious attention. The analysis, using GLMM and non-parametric ANOVA, revealed that both smile expression and eye gaze significantly influenced purchase intention, with gender, age, and education affecting outcomes. Additionally, higher engagement was observed in the latter stages of viewing, with more frequent eye gaze shifts indicating increased attention. The findings highlight the importance of customized interactions in live-streaming, offering implications for optimizing streamer performance and informing human-computer mobile interaction design to enhance consumer engagement and purchase behavior.