IThis study engineers and rigorously validates a transformerdiffusion architecture that co-synthesises ultra-high-dynamic-range visuals, object-based ambisonic soundfields and broadband vibrotactile waveforms while maintaining inter-channel onset dispersion below 10 ms, thereby enabling an authentically embodied, cross-sensory screen aesthetic. Thirty-two adults participated in a within-subjects trial, viewing AI-enhanced and baseline clips inside a 270 cylindrical CAVE outfitted with 20 000-lumen RGB-laser projection, a 64-speaker wave-field-synthesis array and a 256-actuator haptic floor. Objective responses were captured through gaze-entropy, galvanic-skin-response amplitude and EEG beta-band desynchronisation, whereas subjective aesthetics were rated on a nine-factor inventory with Cronbachs = 0.92. The AI condition amplified the composite cross-sensory immersion index by 47.3 %, curtailed onset error from 23.5 ms to 8.7 ms and boosted aesthetic appraisal by 1.12 standard deviations. Hierarchical mixed-effects modelling ( = 0.57, p < 0.001) revealed that 71 % of the aesthetic uplift was mediated by temporal congruence, underscoring millisecond synchrony as a pivotal mechanism linking physical stimulus design to perceptual reward. Beyond confirming the causal potency of AI-driven synchrony, the work supplies reproducible calibration protocols, quantitative immersion metrics and a generalisable analytic toolkit for designers of next-generation multisensory content and display infrastructure.
Liang et al. (Tue,) studied this question.