Talking to the screen: Influencer nonverbal cues and observed platform engagement in Douyin travel short-form videos

Short-form travel videos have become important digital service touchpoints, yet little is known about how influencers' auditory and visual nonverbal cues jointly relate to cumulative observed likes. Drawing on processing fluency and cognitive load theory, we computationally extracted acoustic and gestural features from 495 Douyin travel videos posted by 60 influencers and estimated Poisson fixed-effects models of within-creator variation. Harmonics-to-noise ratio (HNR)-based acoustic signal quality exhibited a U-shaped association with observed likes, with the fitted minimum occurring at an intermediate HNR level. Faster speaking was associated with fewer likes. Illustrator gestures had no uniform main association, but their association with likes became less favorable at higher vocal pitch and faster speaking rates, with the latter pattern showing greater sensitivity across robustness checks. Highlighter gestures were more favorably associated with likes at higher vocal intensity, although this evidence was comparatively limited. Overall, audiovisual moderation was selective across cue combinations rather than uniform. These observational findings identify bounded associations between nonverbal cues and cumulative observed likes and provide empirically grounded hypotheses for controlled experiments on audio-gesture combinations.