Short-form video has emerged as a pivotal medium for the dissemination of emerging technologies, positioning technology enterprises as one of the main agents of science communication. Focusing on Unitree Robotics’ official Douyin account, this study draws on a sample of 81 videos and 4 114 comments to examine the interaction between corporate technological performance and public sociotechnical imaginaries. The study integrates multimodal discourse analysis with text mining to capture both the semiotic construction of technology and its discursive reinterpretation by the public. The findings reveal that embodied AI is rendered as a visualized, contextualized, and immersive technological performance through the orchestration of textual, visual, and auditory modes. Correspondingly, public discourse generates multiple, intersecting sociotechnical imaginaries structured around functional utility, techno-nationalism, and human-machine relations. These dynamics suggest that technological meaning is not unilaterally transmitted but continuously negotiated and stabilized through platform-mediated interactions between technology enterprises and the public. This study reveals that technological performance goes far beyond mere technology demonstration. It opens up a crucial space for the public to engage in the social construction of technology and to shape sociotechnical imaginaries. Consequently, these practices reconfigure science communication from one-way dissemination toward the participatory co-production of meaning.