At present, discourse analysis is widely used in news reports and TV talk shows in China, but there is a lack of discourse analysis research on live streaming. Against this backdrop, this study adopts a multimodal discourse analysis approach, drawing upon the comprehensive theoretical framework for multimodal discourse analysis and Kress and Van Leeuwen's seminal visual grammar theory as the theoretical basis, taking quality and quantity method to analyze data. Our research aims to investigate the intricate interplay between anchors and the amalgamation of diverse modalities in the process of constructing their image. Furthermore, we endeavor to analyze potential strategies that can be employed to enhance the efficacy of live broadcasting and augment viewers' engagement and consumption. In this paper, we decompose the multimodal discourse of the selected live video of “Oriental selection” into verbal and non-verbal modes, and analyze the process and role of how anchors use multimodality to form interactions and promote construction.