Taking L2 video learning with subtitles as the research context, this study adopts a 2 (Perceptual Style: Visual/Auditory) × 2 (L2 Proficiency: Intermediate-High/Intermediate-Low) between-subjects experimental design. Combining comprehension tests, eye-tracking, questionnaires and interviews, it systematically investigates how perceptual styles and L2 proficiency jointly affect learners’ subtitle learning outcomes, cognitive processing and subjective experiences. The results reveal significant main effects and interaction effects of both variables. Visual learners gain greater advantages in subtitle processing, with an especially prominent compensatory effect at the low-proficiency stage. L2 proficiency exerts a stronger influence on learning performance and cognitive efficiency, and higher proficiency can substantially eliminate gaps caused by perceptual styles. Low-proficiency and auditory learners rely more heavily on subtitles, accompanied by heavier cognitive load and lower processing efficiency. This study uncovers the bidirectional interactive mechanism: proficiency moderates perceptual styles while styles compensate for insufficient proficiency, providing empirical evidence for personalized L2 subtitle teaching design.