The sound interface is ubiquitous yet often remains inconspicuous. Moving beyond the visual-centric paradigm, this study adopts a procedural perspective of the interface, focusing on voice interfaces embedded in domestic smartphones and embodied as intelligent assistants. It explores their formation, maintenance, and the complex interface effects triggered by sound media. Through discursive interface analysis and semi-structured interviews, the study finds that within the “black box”, the interface modifies users’ spoken language based on a cybernetic technical logic, simulating interpersonal communication while masking the programmed nature of interactions. In the interactive practices, users engage in continuous efforts to address interface failures or breakdowns and to reconcile the contradictions arising from the dual nature of voice—as both an interactive tool and a medium of communication. Furthermore, as a relational medium that constructs the relationship between humans and the technological environment, sound activates an auditory turn in interface interaction and a paradox in technological perception. While users rationally assess the semantic limitations of voice assistants, they also develop imaginative conceptions of technology as possessing a sense of life. This study conceptualizes voice-dominated human-computer interaction as “acoustic interface communication”, emphasizing the urgent need to recognize the unique significance of sound in human-machine interfaces.