This study, through multi-case empirical analysis, systematically explores the collaborative mechanism of gestures and visual graphics in the phonetic teaching of Chinese as a second language (L2). The study adopted the interpretivism - constructivism paradigm for in-depth observation (N=7 teachers, N=22 learners), combined with video analysis, semi-structured interviews and neurocognitive experiments (EEG/fMRI). The research found that: (1) The effectiveness of gestures is restricted by cultural embodiment. For instance, Vietnamese learners misinterpret the upward gesture as an insulting action; (2) Static graphics are prone to trigger cognitive simplification traps, while the dynamic integration of "gesture-graphics" increases pronunciation accuracy by 22% (*p*<.01); (3) Teachers need to adjust the modal combination in real time based on the phonetic features (such as the dynamic selection of fricatives, the comparison of pitch values, and the use of spectrograms) and the learner's characteristics (L1 background, cognitive style). Based on this, the research proposes the Contextualized Multimodal Integration Framework (SMIF), which includes three major principles: the law of cultural embodiment priority, the law of movement-visual synergy, and the law of emotional empowerment priority, providing a theoretical paradigm and practical path for the design of Chinese phonetics teaching.